In terms of RUclips analytics, this is my most successful RUclips video to date. In terms of content efficacy, I dropped the ball on more than a few of these topics. So, be sure to read through the comments as there is some very valuable information. Thanks to everyone contributing!
19:42 - That is not something we are aware of cause we had not used AC for decades just to avoid the costs that come along with that. We have gotten AC in the past 2 decades but only those in cars which usually do not resonate hearable. I bet this is the american perspective which even does not fit in central europe. kWh / household consumption is far lower in the EU and awareness of energy costs too cause they are higher. The higher the costs the more the people are aware of consumption and efficiency. Therefore you will not find trucks.
That crane video finally made me understand. I use input shaping when I drive, or more specifically, when I brake. When coming to a stop, I release the brake to prevent the uncomfortable jerk you can get.
I think you're referring to a different kind of jerk. I believe they're referring to the uncomfortable sudden stop or start that comes from quickly pressing or releasing the brake. Changing the speed at which you press or release the brake would reduce this jerk. A jerk while normally using the brakes would point to an issue, obviously
@@noah_s906When you brake to a complete stop, the car will keep going forward (just a couple of centimeters) at the moment of "full stop". It will also lean forward. Then the springs will return the car back and you will get this bad jerk. If you brake real hard, you can actually hit the seat with the back of your head on "the return". If you release the brake just before coming to a full stop, this will not happen.
No. What you apply when braking is called S_CURVE_ACCELERATION. This is a change in the acceleration value at the beginning and end of the section. Input Shaping is more like the work of Linear Advance, when the extruder pulls the filament back and forth so that there is no swelling on the print.
A pretty good video for beginners. But there is one little fault in the PSD graphs (min 26.30-28). You compiled multible resonace tests together what will lead to an incorrect PSD graph and subsequentely less optimal inputshaper. It is a common issue. To fix it reboot klipper after compliling one set of results.
Spencer, first of all - you did a really great job with this video and you earned a subscriber! Almost brought a tear to my eye as an engineer. Some "into the weeds" details are given below, which may help as a second layer to what you already said, for posterity: In terms of dampening and tolerance to error, MZV is looser than ZV. And EI is even looser still. As such, a lot of people prefer to use EI on the Y axis for a bed slinger, because the moving mass will inherently increase as the print progresses. This gives you some flexibility at the expense of the theoretically ideal shaping scheme. For the X axis, you obviously picked up on this already - but something should be addressed if you have 2 major resonances modes as shown. Is your filament spool on the top of gantry? I would be looking for decoupled masses like that, etc. On my machine, moving the spool to a holder off to the side really cleaned up my shaper results and I am back to 1 discrete peak. Somewhat related to the above - a trick in the Klipper community is to 'pin down' the Y axis when you are measuring the X axis, and vice-versa when measuring the Y axis. This is usually done mechanically with some sort of clamping or interference fit piece to stop it moving in an orthogonal axis during the measurement. Especially useful on a bed slinger.
As a physicist, I’m humbled that I made an engineer proud😂 I’m definitely new to video making and honestly the entire field of 3D printing, DIY etc… I expect to make loads of mistakes and learn a ton! I remember hearing of your first point somewhere. I wish I would’ve mentioned it! I have no clue where the second peak came from honestly. In my last video I stripped everything from the printer. Drawers, electronics, everything! No spool attached, no uncoupled masses (as far as I can tell). I assumed that maybe the z-axis was super unstable or something. As for the 3rd point, is it a good idea to clamp the masses down? In my head, they affect each other while printing so wouldn’t you want to have them affecting each other while measuring? Thanks so much for your comments!
@@SpencersDesk For tbs. just remove the whole toolhead, and mount the accelerometer directly to the backplate. Then you will know if its the motion system itself, or something on the toolhead that is making the extra peak.
@@SpencersDesk that's a really fair point about the last point, i.e. 'pinning down' the orthogonal axis to isolate just the tested axis. It makes sense that they are not restrained during printing, so this deviates from real conditions. This would be a great question for the Klipper community. If I had to speculate - perhaps it's something to do with how Klipper implements the negative interference. Both axes are obviously prescribed different (and inherently discrete) modes, so the "composite" commands to the steppers may work the best when these are also purely discrete in the testing? Again... 'speculation' being the operative word! Another theory could be related to the 2nd point in my original post - maybe people were getting dodgy results due to interference from other SMD systems on the machine? Whilst the steppers should be frozen in place on the other axes during this test, the reality is of course that there is some decoupled mass movement. For what it's worth, I've never actually pinned down the other axis and have never had a problem. I'm pretty sure that the largest source of error is the quality of the ADXL345 itself. When I query the current accelerations and noise, it's not very consistent. But it generally aligns with what I am expecting, comparing to global gravity constant in Z. But usually +/- 10% around this supposed constant! Cheap Amazon parts...
woah. the diagnosing printer part was very illuminating. a lot of vids I've seen just chuck the shaper on and move on after setting the profiles. what you mentioned, but didn't explain, is why one would have input shaper permanently on.
Hey, input shaper changes how your printer moves as discussed in the video. When you run the input shaper tests, you get parameters that will best help your printer. You then add these to your printer and enable input shaper. Without input shaper activated, you’d see no change in your print results as the printer wouldn’t be moving differently. Hope that answers your question?
I'm so glad the SV06 Klipper upgrade did all this work for me. Just consult the manual for mount locations, print a clip for the bed then screw the provided sensor to the toolhead and run the configurator for X, then attach the sensor to the bed clip, put it where it goes, run the configurator for Y and done.
I discovered you yesterday, I'm looking in today and I see you're off to a great start. I really like it, keep it up! I've been following a number of 3D print youtubers for a few years now, but I haven't come across such a nice analysis of resonances yet. I would highly recommend, don't say what you think, for example about those A, B, C subsets of SPI, but get the facts and say the facts, or don't say anything at all. ;) I appreciate how you mentioned the obsolescence of the naming of MISO and MOSI without mentioning the background of it all, and just put forward the solution as it's being named now. Similar names have been part of the electrical and PC jargon for decades and it brings some confusion now, but I appreciate that people are thinking about it and looking for solutions.
Well, it's a good start to a nice analysis, lol! I really appreciate the feedback. Sometimes it's hard to get accross the voice saying (it's good enough). I want these videos to get to the point where they're fairly airtight and don't involve guessing! Yeah, the MISO MOSI thing never stuck out to me until some high school students were touring a lab I worked in. We had "MASTER" and "SLAVE" written on a Raspberry Pi CM4 IO Board and one of the kids asked why we had that. A little uncomfortable, haha. I'm not bothered by a change to it so why not try to help the trend. And again, COPI (co-pee) and CIPO (kippo) sound way funnier anyway. Thank you for the support!
Elsewhere on youtube, I saw someone resting their printer on a heavy cement tile (about a couple of inches thick) itself resting on some foam, claiming some improvement with input shaper and overall accelerations. Anyway, I really enjoyed this video, and hope you make many more. Thanks!
Hey, I definitely want to do that at some point. A video about reducing/fixing bad graphs might be really fun. It’ll definitely be a continuing theme on this channel for a little bit. Thanks for the kind comment (:
Great video! One important note about shorting out / damaging your electronics. The only thing that is ever critical with those communication signals is power and ground. If you triple check those, you can have any other mistake with clock and data lines - nothing bad will happen. It just won't work. You can try another combination and correct it. I'd mention that clearly not to scare people. Paying attention to that simple rule I never destroyed anything in 10 years of tinkering with electronics (at least when dealing with simple connections like that). Keep on making great videos :)
That's a great point! I just know from my own experience that once, a mixture of exhaustion, frustration, and impatience made me make some very silly errors and I destroyed a Pi. So, the extra layer of protection is nice! Thanks for your support!
An informative video. Great watch. Off the top of my head, your shelf material is likely to be flatter than the floor. So the printer base will be equally supported on each corner/foot. So the printer may wobble more even on a hard floor. On the other hand a lot of shelving isn't level, indeed some brackets encourage a slight rearward slope, which may give uneven results. It might be revealing to experiment with thicker or heavier shelving material, or even rubber mating. There may be a sweet spot between damping or encouraging wobble!?
Thank you! I remember someone (Tom from Made With Layers?) did a video on it. I think he landed on using a concrete paver stone as the base then using some foam padding at each corner. It gave the printer a sturdy base and the foam uncoupled the printer from the surface. Definitely something I plan on looking into in the future!
Also worth mentioning, that input shaper on small parts such as gears can result in artifacts! So in case you are printing something small and precise set low speeds and maybe disable IS.
Jut as an additional input: Marlin also sports input shaping since a few minor versions. I had very nice results by printing a test-print and measured the position of best edges and calculated the frequency. Added this f to Marlin config and perfect :) I like to have no additional Raspi and have all done in Marlin. Also a pro: No sensor needed ;)
Thank you for explaining this so well! I was able to get two accelerometers connected to one Pico :) Being able to increase acceleration (to 5k) improved my print time a lot (takes 63% of the time it used to).
This is a fantastic video. If you're up for it I'd love to see you experimenting further with frame brackets, braces etc. to see how rigid you can get it and how it affects the resonance tests. This is really interesting to me because my other passion is automotive. And what you're effectively doing here is building a race car. By making it as stiff and rigid as humanely possible, you're allowing any feedback (resonance) to be picked up by the driver (accelerometer) which can be compensated for with technique (input shaping) to go as fast as possible (same same). If the car (printer) has loose or flexible components, then it acts in unpredictable ways which can be hard to compensate for.
Thank you so much! Don't you worry, I have many more plans for resonance! Yeah, that's a great analogy! Our goal is to have a delta function (kinda, not the whole infinite amplitude part)! My point being, we want to have a single frequency which we can just avoid creating! Thank you so much for your support friend
Your description of the wiring is good, but there are some efficiencies to simplify things.... there is no real need to use two different SPI busses for the sensors, whilst not an issue with the Pico example it's not really the most effecient thing when using SPI long term (that is to say, wastes pins that could be used for other things if you had a different project using SPI devices I the future). You can use the same power, data in (MOSI), data out (MISO), and clock lines for all SPI devices... just attach different controller pins to each sensors Chip Select pins. An SPI device will only communicate of the MISO line if it's CS pin is high (or low can't remember off top of head). So using different pins for CS allows the MCU to select which device it's talking to. In your Klipper example it's just a case of putting a different cs_pin in the config for each accelerometer. For other projects using SPI in the likes of arduino or other IDE (or coding language) again it's just a case of defining the pin used for CS for each device .
You're spot on with that! There are two reasons I recommended wiring separate buses. The first is, simplicity. I didn't want to go in depth to SPI and how it works and its capabilities if someone only wanted to get input shaper. I just tried to give enough to give some context in a digestibleish manner. The second reason is it "complicates" the wiring. Sometimes splicing two wires together can be a hassle, and the idea of having multiple chips on the same wires while another wire had to be different didn't seem like a digestible topic in the amount of time I had. If you were wiring directly to a Pi or something and wanted it to be a permanent implementation, I'd say your way is the best. For me, I'm not planning on using the Pico for anything else at the moment, so no worry of running out of pins. Thanks for your comment! There's so much great information coming from people, I wish I could pin multiple comments...
Thank you so much, I knew the theory more or less behind input shaping but i didnt knew how to read the graphs and/or choose what shaper I would use. Now lets see if i can get bet print quality and higher speeds
Great video! A suggested video turned out to be a real gem! I have been thinking about sand filling the frame of my Ender 3 to dampen resonance and increase mass. A common practice in heavier tools such as lathes, mills and bench drills. Even cars have resonance dampers for specific frequency mounted on suspensions, exhaust, engines and gearboxes. I'm sure theres a specific amount of mass which will be optimal for best perfomance on a 3D printer which would be seen on klippers resonance curves. Also in video Y axis had more resonance when placed on floor instead of the shelf. Got me thinking, what if printer was placed on rails? Which made printer movable. You earned a new subscriber just by this video! Keep it up
Thanks for your kind comment! I hadn't heard of filling the frame with sand but I like the idea! I'm curious how the printer would respond if it was sitting on a free moving platform (rails in the x and y direction). My intuition tells me that the lack of dampening wouldn't be great for it? I think skyscraper do something similar where their supports are allowed to move, but the motion is dampened. Cool idea!
Think it's the first time I see any explanation at all about smoothing. Klipper docs just throw the term, everywhere else is, well all over the place. Great content, you did a pretty good job introducing it. I'd suggest a follow up to go more in detail for the diagnostic part and understand all the chart lines. I would also point out that the input shaper is a great way to tune belt tension. It's very easy to get it wrong and can have a dramatic impact on print quality, acceleration and even layer shift odds at higher speeds.
Thanks! Yeah, I wish they had areas where they explained the actual kinematics equations, how they implement things, etc. While this video was a bit of a fail, I'm happy that people are at least thinking about it now! For belt tension, this is a lot easier on a CoreXY as you can see the peaks not aligning properly. For me, I'd have to test, adjust tension, test... repeat. They can help identify relative belt tension, but not absolute belt tension. So I think they can tell you if your relative tension is off, but not if you have the correct tension. Do you agree? In all honesty, I know very little about belt tension. I'm not sure how one does it properly!
@@SpencersDesk I don't think of it as a fail at all. Don't be hard on yourself. About belt tension, I have a cartesian printer (not bed slinger) and input shaping was instrumental to tune belt tension. You will see that the indicative max acceleration will increase or decrease depending of how far you are from the ideal applied tension. Of course this method only works if the print is already reasonably calibrated for other factors that may affect rigidity.
When you say Cartesian, do you mean CoreXY? So you suggest running the test over and over, adjusting belt tension in between? Until you find the highest suggested acceleration?
@@SpencersDesk, no, cartesian is one of the kinematic types you can use in klipper. It has independent X and Y axis, in contrast with corexy where they both work together. Yes, I start with some tension, measure resonances, increase tension and measure it again, until I find the sweet spot. Ideally you should take note of it and I think the easiest way to do it is to use a guitar tuner (yes, you heard me right, you can play your printer like a musical instrument), it produces different sound notes, when you fiddle with it that are related to the tension applied. Edit: It is not about finding the suggested accelaration, but instead, observing the suggested acceleration and observing when it starts to decrease (too much tension).
Hi. Just a 'litlle' hint to 'ALLl' You Tube presentators (NOTE! Just in general): When you hold your item in your hand to show the thing and explane something of it at the same time: Please hold Your hands 'STILL', don't 'swift' 'flap' 'move'... or whats the description of my meaning (I'm a FIN so my gramma in english doesn't cover so well the thought), I hope you know what I mean. I certainly like the stuff on Your video ;)
I think that's a cool idea. Honestly, cooling for PLA is strange if you are printing at low speeds, no crazy overhangs, stuff like that. But (as you can see in my benchy's horizontal lines) when you begin printing at higher speeds, cooling can become a very difficult task. Beyond that, the amount of cooling you have can seriously alter the mechanical properties of your printed parts. I'd really like a setup similar to Tom (Made With Layers) so I can begin to test things like this!
I don’t have something off the cuff, just things I’ve seen. On speed benchy printers, I’ve seen “auxiliary cooling” where there are ducts on the side of the printer pointing towards the middle of the build plate rather than the ducts being on the print head. I’ve even seen a hose being connected to the print head and a turbine off printer blasting air through the hose. As to the mechanical properties changing, this is a pretty common trend in material science. There are all these cool graphs and plots that show how changing how fast your materials cool changes what crystalline structures they form. Rule of thumb is to go hot and slow for strong parts. Helps the parts fuse together. Going fast with lots of cooling makes parts less strong as the layers don’t have time to melt together.
I was always surprised at how little resonance my printer had during all the tests, considering it was an Ender series. Worked out over time it must be due to it being the Ender 5, which has a cube frame and doesn't use the bed slinger method for the Y. I still ran the input shaper. It did improve how quiet the printer and a slight improvement in speed, but I think the Input Shaper is most useful for the "bed slinger" style printers. This video appealed to me both on the geeky nerdy level with the data, but also on a level where it was simple enough to understand and not have to think about htings too deeply.
Nice to see you again! Input shaper will have the great effect when your printer is at higher speeds. If you’re weren’t pushing the ender with high speeds and accelerations, it’s possible there wouldn’t be too much of a difference! Thanks! I’m definitely trying to tow the line between exposing people to new ideas while still keeping the experience enjoyable.
Thanks for the video. Just a bit of errata with the pin labeling for the pico within the klipper config file. While the cs wire is connected to the physically second pin on the pico, it is labeled as gpio1, not gpio2, so the cs line should read cs_pin: pico: gpio1
Build my first Prusa MK1 9 years ago. Then build same concept on my own steel cube frame - big improvement. Now own Bambu Lab X1 - Carbon and no matter what I do with frame will never match performance of X1. Sophisticated software which check rezonans on 1st calibration. Extruder adaptation. 1st layer scanner. Carbon X rails with external extruder makes it very fast.
Print quality definitely starts with the construction of the printer. What is the 1st layer scanner? Is it similar to that "Beacon" ABL sensor? Thanks for your comment!
You're getting your vyper to cook at this speed wow, im going to up the speeds of mine i thought from the thumbnail you were going to talk more core xy than bed slinging. I have been printing at 75mm/s at 3000 accel
In all honesty, I’m gonna try to push it a little faster later on. I still haven’t maxed out the flow rate of my mosquito! Look up 2 minute bed slinger benchy on RUclips and check it out! Someone did a 1:59 benchy on the bed slinger! Thanks for your comment!
Hi Spencer, nice video. The theory was neatly explained. But... (right? ha ha) you created an expectancy 0:24 about showing something for people familiar with input shaping that would "change how we approach 3D printing" - so I stayed... (I wanted to know "why I am NOT getting the most out of Input Shaper") and... where is it? If that's the charts and understanding how they work... well, that's nothing people familiar with it don't know already. By the way, having two distinct peaks at your X axis means the printer needs mechanical attention. A mechanical healthy printer has a more distinct single peak, not two as yours are showing. Finally, the printed results are still quite wobbly - a LOT of 3D Printers can do better than that without input shaping at all. Hope you can improve the X-axis and try it again. Good luck with the channel!
Thanks! Unfortunately, the theory isn't quite correct, so there's still a lot for me to learn. What I did learn changed how I approached 3D printing personally. I never knew the curves had a lot more information that could help you diagnose printer issues. Loose belts, loose screws, problematic wire looms, motion issues, etc... This will change how I test whether future projects are worth it or not. I've implemented input shaper a time or two and was familiar with the process, but didn't know about this side of resonance curves. So, I was "familiar" with input shaper. 99% of the printing community (myself included) has no idea what is going on underneath the hood. This is me looking underneath the hood of a car and attempting to explain how it works after watching a few Top Gear episodes. As other comments pointed out, my last charts are inaccurate due to the way they were generated, so I'm not sure if the second peak is even still present. This would of course influence the input shaper recommendation and possibly lead to smoother prints. I will however say, the before and after result I show in the thumbnail, moving twice as fast and with over twice the acceleration, on a bed slinger, are very good results. The ringing towers are made to accentuate resonance issues, were in silk PLA, and exposed to very harsh light, showing every little artifact there was. Anyway, I will be making a video in the future to replace this one that is based in fact and not my guesses, and I hope it will serve as a useful tool for others. Thanks for watching
This is a bit harsh IMO, it may not be relevant or additional information to you, but going on what is on YT this video indeed puts out additional information.
Great video from the start but you flew by the results, (the 3d printed parts). For someone who delved into details I was expecting for a detailed analisys of the parts. It deserves a brief follow up video imo. Subscribed.
Hey, thanks for the feedback! In all honesty, it was a combination of being lazy, tired, and thinking nobody would actually want to watch it, haha. But, I promise I won't be bringing that into my future videos. There will definitely be a followup video in the future that will better address this complex topic! Thanks for your support!
I would like to mention for other people you can just get the newer adxl345 with usb c and you dont need to do all of the fancy wiring as it is just ran usb c to usb A via Raspberry Pi or whatever Ive ran it through a raspberry, and laptops. Also if you get the BTT adxl345 you can screw it in where the nozzle goes as they provide a hole and a screw to tie into the hotend directly.
Hey, good point! I forget if I discovered the adxl/rpi combo before or after the video, but they seem so cool! It is still nice to have some understanding of how things work behind the scenes though!
Hey, that's a great video. It would be possible, that the PSD peaks you saw on the bed were the back an forth movement (typically printer y axis). I'm somewhat convinced, that the "loudspeaker" theme is not the obvious answer to that measurement. Maybe there was a mix-up of printer vs accel axes - or what could happen is, that your mount vibrated a lot, since the breakout board hangs off of the printbed itself. Just as a reference, a flat aluminum sheet of decent thickness has audible resonances in the highest of Hz, I would assume kHz - range. It's very stiff compared to your belts and the large frame with bolted connections etc :)..
I know that the ADXL was aligned with the printer properly, but maybe it could've been messed up in the configuration? I definitely agree that the z-axis is the likely culprit. In the future I want to add those rods that help support it and see if that solves the issue! Thanks for the info!
Request: clearer side-by-sides. Both diagrams and print tests Diagrams: pick out just the 2 relevant columns from the CSVs, label them properly (e.g. "floor" and "shelf", instead of "before" and "after"), put them in the same diagram instead of next to each other Print tests: photographs side by side, same angle on both objects, clear labels Thank you for the video, I know it's a lot of work, I recommend using a checklist for the production. There are plenty out there for you to combine and adapt
Great feedback! Thank you! Yes I’m still learning a lot about the production process, how I want videos to look, etc. this is very helpful to hear. While normally I would do the combined graphs, I think I’ll keep them separate. They’re just busy as is, and auto generated. I’ll definitely label them better in the future. Thanks for your support!
Please consider the remark this video after you learn more about input_shaper results interpretation. Issues Example at 27:09: -The left graphs has a results of 1e3 instead 1e4 of greater...that's mean that likely a mechanical part is binding during the test. -in Right graphs top, you can see that two CSV files was use to make the result (you have made a average of two results) don't forget to delete previous results before making a new graphs
Thanks for pointing that out! I definitely felt like I rushed it a bit. I’m still grappling with how much of what to include. When I first sat down to edit it, it was about 12 minutes long. I felt like I skipped over so much though and it turned into 30 minutes. I guess I was afraid that viewership would tank if I just started comparing benchys. I’ll make sure to do a better job in future videos and at least do a bare minimum amount. Thanks for the support!!
@@SpencersDesk When it comes to educational content, don't worry too much about how long the video is (other than ad rolls if that is something you care about). Honestly, the longer ones generally do better when it comes to learning, you just want to make sure you include all the important bits. You're doing good, man. Your channel will take off soon
You should use the recommended input shapers for your outside perimeters, you can bump up your speeds and accelerations for most other print features, just test your max speeds and accel don't cause stepper skipping.
I found that title a bit misleading, a good intro to input shaping, but I was left hungry for more info on interpreting the graphs and the meaning of those numbers. I've come across statements saying things like "if your recommended shaper isn't ZV or MZV there's something wrong with your printer" - what do you make of that? Another thing I'd like to mention regarding that Z resonance you're complaining about: I've had similar problems due to the bed belt not being parallel to the Y axis, the angle it had on both ends of the carriage was introducing a vertical component to the forces, I modified the carriage model so the belts became parallel which fixed the problem.
I think the title is fair, as most are just inputting the recommended shaper and moving on. Introducing the idea that you can improve the resonances is what I was trying to convey. I would like to go more into the graphs, but I'm not confident that I could do it justice yet. I'd like to get some more experience and do more testing before I make an in depth video on it! As to the quote... I have no idea why someone would say that. ZV is the "ideal" shaper as it lets you achieve the highest accelerations and does the least amount of smoothing. As another user mentioned, if you have a bed slinger, zv is fine on the x-axis, but ei is recommended for the bed. For my graphs, as others mentioned, I kind of screwed them up as I was blending results from multiple tests. I can't confidently say the adjustments I did didn't remove that extra peak. As far as I know, my y-axis belt looks fine? It's parallel to the extrusion it moves over. I'll take a closer look though! Thanks for the feedback!
vibr refers to the residual vibration % should that particular model be applied--lower is better. sm is how much smoothing will result with that model or how much input shaper will cause the actual movement to deviate from its requested path in order to achieve the vibration reduction--lower is better and higher leads to rounded corners, loss of fine details, etc. only use this to rank the available models, outside of this, the number has no purpose. accel is maximum acceleration for that model beyond which there will be ADDITIONAL smoothing effect. it does not mean your machine is actually capable at that acceleration--you have to print the model to inspect for ringing for this. smoothing is probably not a big deal if you print artistic parts but if you are modelling walls with circular holes that need to be circular, smoothing will make them ovals which is going to be terribly annoying for functional parts. klipper's recommendation for you is based on an internal calculation to rank the models based on these values, details of which i dont know of, but is usually close to spot on. ZV ("zero vibration") is going to the the best model for an ideal printer (one resonant frequency with infinite power spectral density). since such a printer can't exist, other models will probably perform better and this is due to the printer having multiple resonant frequencies that ZV cannot handle but other more complex models can. higher frequencies are better because this means the shaper can work at higher frequencies to counter it. low frequency is worse because if the shaper can only work at lower frequencies, your accelerations will also be lowered. that said there is little you can do to majorly alter the major resonant frequencies of your printer without serious modification. but you can improve it a little with screw tightening: tightening screws everywhere is a good idea in general: it can really eliminate certain small peaks as shown in the video. for example if the fan shroud screw is loose, this means the mass of the shroud becomes its own oscillator causing a small peak other than the main one (the printer frame, the toolhead). tightening it down eliminates the peak as it truly becomes one with the toolhead, creating a new, larger mass that resonates at one, new frequency, probably a higher one. resonant frequency is proportional to the root of stiffness so if you tighten all screws in the frame, the major resonant frequency contributed by the frame should shift right as shown in the video. it will also reduce internal damping which will squeeze the peak making it taller and thinner (become more like a tuning fork, which is easier for input shaping to counter). less damping is not good but if you tighten the peak, the shaper will work better at its sole job which is to add damping. the peaks which didn't move are probably not related to stiffness from screws. just a warning on screw tightening in the frame. this can create skew in the frame for any printer making it lose squareness. where you sit the printer also matters. if the table or shelf has a different resonant frequency as the printer, it is going to show up as its own peak. meanwhile the solid ground has infinite mass but also infinite stiffness, meaning it does nothing to the printer--it doesn't actually make the printer worse per se---it just reveals the true resonance of the printer itself where previously the table or shelf you had it on was simultaneously providing noise as well as damping (in most cases the damping effect more than compensates for the fact that it introduces a noise frequency which sometimes at least one of the shaper models can also compensate for). but ideally the best place to put the printer is on a stable and stiff furniture (no noise) but with damping from stiff foam. you don't want the printer to sit on foam directly because the frame needs a solid and flat surface to prevent flexing from its own vibrations. so what people have come up with is to place it on a stone/granite/cinder block slab with stiff foam below and the sturdy table underneath. this gives you a massive, stiff and flat surface for the printer while still providing damping. last note: input shaper will alter the behavior of PA/LA. do input shaper early or else redo PA tuning after.
I recently got a Mini and it does move a lot faster than I expected! It's cool as long as you never change anything about the Prusa machines. Otherwise, there's nothing you can really do to adjust the input shaper parameters
I've got an Orbiter V2.0 on my current setup. I'm currently planning a video that looks further into flow and what the biggest factors are for it (temperature, extruder motor current, etc...)
@@SpencersDesk as far as the pico, I did Bltouch, accelerometer and relays, but then dug more into finding ways to utilize spare gpio pins on the actual printer MCU and started making a video on that. Then I got sucked down a rabbit hole of converting my artillery sidewinder X3 plus to Klipper after they finally decided to send me the official information on the pinout and bootloader offset I needed. Now I need to circle back around to the gpio stuff and finish up that video.
Very cool! Make sure to share the configuration you use with the community! I think you can share the example config on Klipper and they will ad it in!
Hey Spencer, great video, I appreciate the detail in the electronics section, and the recommendation to use a sacrificial pico; that will no doubt save heaps of frustration! I'd have loved to see a more detailed review and comparison of the results - seemed like you introduced and then skipped it? is there a section missing in the edit?
Thanks! In all honesty, the video was originally 12 minutes long. I decided that it wasn't good enough for me and spent an extra 4 days filming and editing. Towards the end of it, I had a mindset of "no one is going to see it so just get it done". Clearly I was wrong, haha. I'd just assumed no one would care by the end of it I guess. Don't worry, videos from here on out will not be approached with that attitude!
I have the FLSUN PRO Delta style 3D printer. I just ordered everything you stated in the video. I'm hoping this will work with a Delta type 3D printer. Are you aware if this will work with my printer?
Input shaper is just one part of good print quality. If you print fast, the nozzle and linear advance has a much bigger impact on what looks like ringing or ghosting. Accelerometer input shaping can also give pretty poor real world results, as it doesn't measure the tip of the nozzle. The amlitude is always to low or even completely off depending where the accelerometer is mounted. Therefor the dampening values are set wrong over an average of printing speeds.
Hey, thanks for your thoughts. I totally agree, input shaper is just one of the 10e6 moving parts that determine the quality our printers can deliver. Your comment comes across as if you're saying input shaping for 3D printing is useless. I'd argue that the accelerometers do a very decent job for our use case. While it would be ideal to mount an ADXL in the exact same point in space that the nozzle is at, we just make the assumption that the print head is rigid enough so that resonance at one point is similar to any other point. Also, I don't think that resonance depends on print speeds. Resonance is a result of the construction of your printer, how rigid it is, what materials it is composed of, the surface the printer is sitting on. From what I can tell, the same input shaper works across different speeds
@@SpencersDesk IS is important. I don't know why my comment came across that way. I meant exactly what I said. IS is just one part of a good print. However I would not use an accelerometer to dial in IS alone. The dampening values of IS can throw off linear advance with different nozzles. The amplitude of the resonance is very important at the nozzle and not the accelerometer. Therefor dampening values are to low or too high. Depending on speed printed. This can than be amplified by the pressure advance settings. You can see ringing artifacts produced on all auto calibrating printers - with accelerometers, unless tuned to the exact print speed and pressure behavior of the nozzle/filament. Not even considering the impact of acceleration and jerk settings on IS.
I never heard of this before and it was very interesting. As a total neewbie, I didnt quite get, how this gets applied. If I understand correctly, this get measured once for the printer and then acts as a correction function. But as what point is it used. For example I use oftoprint with a raspberry pi on my artillery sidewinder. Is this correction something that gets applied in the slicer or does it need a totaly different way of controlling the printer directly with the pi?
Hey, great question! Klipper itself dictates how the printer will move. The slicer says how fast, and what direction to go, but the actual motion is determined by Klipper. So, Klipper takes the gcode you give it, and determines how the printer will move. If you’re not on Klipper, then other firmwares can also do this. For example, prusa machines have input shaper built into their firmware (depending on which version you have) Hope that answers your question!
Hey, yes sort of. The way I described it was a little wrong. We aren’t “canceling” the oscillations but instead “preventing” them. Noise canceling is definitely the canceling version!
@@SpencersDesk - hope so! My PiOS install is pretty gimpy, so I might flatten it and just start over. Not that hard to put in KIAUH and MC, and all that now that I've done it once. Pi won't even see my machine now, so maybe I gimped the USB ports, or... who knows!! :D
Quick questio, but kinda complex. Is there any way to combine input shaping results from the accelerometer on both the head and bed for one axis and input shape based on both resonance values combined? I know the values are far lower for X on bed for example, but would there be value in compensating for that as well? Just a thought.
I’m not 100% sure I understand your question. Do you mean, take measurements on both the head and the bed separately then combine them together? Or are you referring to the head having resonance in the x, y, and z and compensating for each of those?
@@SpencersDesk Yes taking readings on both and combining. As in, combine the X reading for both bed and head measurements and apply it to X movement and same for Y reading and the bed. It's probably just not significant enough to worry about.
Great Video. Thanks. What about the cross correlation with Pressure Advance? When I did the input shaping I had to repeat my Pressure advance calibration. And I found pressure advance to be very sensitive to the acceleration I am using. Did see any impact to your pressure advance setring when you have changed the acceletions? After input shaping I was able to run with 3500 mm/s2, but only for the generic calibration test. When printing real parts the print quality was6not good, and I think because of the cross correlation between acceleration and pressure advance.
Hey, great thing to point out! I did recalibrate pressure advance after all of my benchy’s. I should’ve pointed that out. I wouldn’t call it a cross correlation as it’s kind of a one way relatiknship, but yes, changing the accelerations requires you to recalibrate PA. Thanks!
Very helpful vid. I was thinking i'm lazy and it would be great if I can just leave the pi pico on the machine, but my raspberry pi3a runs hot and I'm concerned about the power draw. So I was wondering if I could connect my pico through a U2C [usb to can already takes power from the power supply so shouldn't strain raspi] converter with canboot and plug pico into that. I'm assuming the pin callout would look different, but also like would it be possible.
increase your accel per hz and you'll get cleaner graphs. the power spectral density is only scaled up to 1e3, which is usually an indicator that you're measuring a lot of non-data
Is that a parameter in the resonance testing sections? I’ll look into that. From what I’ve seen the 1e3 is pretty typical. What do you mean by non-data? If you mean noise, I don’t think the smaller peaks I’m seeing are noise
@@SpencersDesk yeah, it's in the tuning parameters, though it's used less. it's real data but the actual impact on printing is next to nothing, so while not noise, it's not as useful as the major peaks that will show up better with higher accel per hz. I guess you could call that noise if you wanted, but it's not reaaally noise in my estimation. it's still useful data, especially since you found loose bolts, but not what I would base my input shaping on. most of the machines i see have at least 1e4, but that could just be the types of machines I'm used to seeing
I see what you mean. The small peaks are real data, and have some cause, but you’re not taking them into consideration when determining an input shaper. But yeah, I think e3, e4 is normal range
I have not yet. I am planning on investigating whether the recommended input shaper is worth listening to/if adjusting its recommendation is any good, so hopefully I’ll have some answers for you!
there is or was a problem.. not sure if still present. the very first query to accelerometer after firmware restart on a pico will always fail. the second will succeed. dont give up to soon. (has todo with late spi initialization)
I'm not 100% sure but I don't remember this happening when I did it. I definitely should've mentioned that the first step is... try again. Thanks for the comment!
I would definitely love if you could dive a bit into the limited cartesian kinematics. From what I have seen it allows you to have a separate x and y axis acceleration which allows you to fully utilize the maximum input shaper speeds without being limited by a certain axis. For example, at around 23:25 If I wanted to use the input shaper EI with an acceleration of 3000, I wouldn't be able to utilize input shaper 3HUMP_EI to its full 3300 acceleration.
I may be misunderstanding your comment, but for Cartesian printers, you are able to define different input shapers for the x and y axis. The same goes for Delta and CoreXY or CoreXZ printers. I don't believe there is any printer out there which behaves the same in both the x and y direction. So, you will need to assign different input shapers to each one. You definitely can have 5,000mm/s^2 on one axis and 1,000mm/s^2 on the other, it shouldn't matter
@SpencersDesk my apologies, haven't done a ton of research on it. It more on setting independent acceleration limits for x and y axes. I can't sent links in comments but if you Google "limited cartesian klipper" it's the first link!
Hey, I checked out some links and it seems to say that Klipper, while letting you define separate values for the 2 axes, still only uses the lower value. I saw that someone developed a method to change this? But it isn't officially supported yet.
@@SpencersDesk Gotcha! Yeah someone made their own python script for the kinematics. Would you see any benefit of doing this or should I do more research before implementing?
I can't speak to it honestly. If you're not satisfied with your printer's current state then I'd say go for it. But, I don't know how much benefit it brings. Maybe others can speak to the difference it makes?
cannot belive there is a signgle thing thats this complicated that people are going to go through if its not their job and if they are not qualified. 3d print community is sth wild really. Like everyone are programmers.
I appreciate your work on the topic. Unfortunately, this only applies to one printer model. For me - a Delta owner - this information is only a theory.
Yes! Prusa recently added input shaper to their printers. Bambu labs does it a bit better as I think they can measure the resonance when you set it up whereas Prusa does it at the factory. Maybe that makes it account for the work surface better? Or maybe the Prusa is hand tuned and better that way?
Only 7 minutes in but the entire time I was wondering how do you input shape on a bed slinger printer. With the bed moving in the y axis and the print head moving in the x axis where do you put the senors in order to capture both oscillations. Honestly I felt like input shaping was only for CoreXY printers where the print head moved in both the x and y direction. I'll continue to watch.
Hey! If you’re looking for a direct drive setup, I think this extruder is the one to go for! Made With Layers or CNC Kitchen did a video comparing a bunch of extruder motors which might be more helpful to you! I think in the future I’m going to opt for a Bowden setup and see how that treats me!
@@SpencersDesk Thank your for answer, Nice Content! I. My Bondtech clone is just fine with simple screw and nuts mod (without spring) but I need a more versatile Extruder, I checked The Orbiter 2.0 but I didn't like plastic gear idea in a filament Extruder... After lots of research I decided to buy MicroSwiss NG Linear Edition. Thank you.
I don't have a printer but how about this technique. Draw/print the line then stop the printer at the corner then proceed downward. Would that make for sharper corners?
If your printer was moving slowly yes. At higher speeds, the corners aren’t sharp anymore due to the printer not being perfectly rigid. The advantage of input shaper is that it can create these sharp corners at higher speeds!
Hey! I really like that question. I’d argue that oscillation along all 3 axes change. I didn’t think of it until you mentioned it. I would say that the oscillations don’t change an insane amount though. As they mainly come from the resonance frequencies of rigid bodies, I don’t think the change should be too great. So my short answer is yes, input shaper will be effective at all heights. In the future I plan to add supports to the zframe and I’ll be sure to test the resonances at different heights to get a more concrete answer to this!
Hey, nice catch! Unfortunately, I didn't think it mattered at the time and rolled with it. Others have informed me that's not the case. The probe point did not change between tests.
I just moved to linear rails (video posted today) ;) I definitely still have a lot of learning to do on managing resonnace. I definitely think it's silly to expect great print quality at high speeds, so I printed at high speeds knowing there was that trade off, as I was trying to accentuate the errors. Thanks for your comment!
Hey! Thanks!! If you’re referring to the orange cabling, I did that in my last video! If you search up PET wire loom on Amazon or something you can find it! Orange is my personal favorite color so that’s what I went with but black is the most common! I have a link to the orange loom I used in the last videos description!
@@SpencersDesk must have missed that part in your last video, I’ll take a look. Did you make your own cabling for the printer? Looks like the cabling fit it is super long. Only asking because I have a ender 3 pro and I don’t like how short the cabling is and I want to rewire the machine
I sure did! I rewired every single electrical component. Steppers, limit switches, print head, PSU... I made them each super long because I plan to one day print high temperature plastics with this printer. Moving the electronics out of the printer enclosure is a must so I went ahead and installed the extended wiring I will need. I will say, the steppers were a bit frustrating. On the Vyper, all the wires were black. This made figuring out how to wire them to the controller board a bit of a guess and check process. It also requires some extra tool for crimping the connectors. Let me know if you have any other questions about it!
Hey, I’m not sure. I’m assuming you’re using a Pi and Klipper, but the 8bit board may just be too slow. This is mainly useful at high speeds, which you may not be able to reach with the 8bit? Disclaimer, I’ve not used an 8bit and can’t be sure. I would recommend a new board! They’re not too expensive for a reasonable one! I’ve got a video where I replaced my board if you need some inspiration!
Hi, I’m not familiar with the E3 V3 KE but I just looked it up. It mentioned input shaper but I have no clue if it has an accelerometer or not. It may be like the Prusa machines where input shaping values were found at the factory and are built into the firmware already on the printer. You may already have it activated by nature of having the printer. Hope that helps!
@@SpencersDesk yes and no, I have an option to adjust values on the adjust page , however when I tap it, I get a prompt telling me that I don't have the available for mentioned "sensor", vibration sensor. Some call it a G sensor.
Does it really tell the printer to go BACK? I thought that it "cuts the corner", since frequencies are REMOVED and not added: going backwards involves adding high frequency movements. But in general the results are the ones you mention
Imagine using a Yo-Yo. When it gets to the bottom of the string, you do a jerking motion to get it to come back. It's similar to that, where it pulls a little extra back in the other direction to cancel out the move. It's more of a controlled stop then just slamming on the brakes. It isn't just cutting the corner, as you wouldn't need to measure frequencies to round a corner. Think of the ripples in a pond when there are splashes in two separate locations. The ripples cancel out in places where they have opposite amplitudes. Input shaper guesses at what the opposite of the result would look like, and does that to try and cancel it out.
@@SpencersDesk I understand what you say and I also see that that's what they do in the video of the object hanging from strings, but I thought that klipper just band-blocks the resonating frequencies and that implies smoothing the motion, not adding a high frequency compensation (the "pulling back"). After all, excessive compensation causes smoothing of the corners, which is a confirmation of my interpretation of the correction applied. Let me ask the devs on the klipper discourse. If you look for new threads in few hours you'll be able to participate to it.
I will say, by adding two identical waves that are phase offset by 180 will result in no oscillations. In other words, adding more, but opposite vibrations will lead to no vibrations. But yes, as others have confirmed in here, it is a changing of accelerations. The ZV smoother splits acceleration into 2 steps, while other shapers add more and more steps, becoming more robust against error and change, but cause more smoothing. I don’t thinks it’s a band pass of any sort. And sounds good! I can’t wait to see what comes of it!
@@SpencersDesk The main developer of the IS module answered on the Klipper Discourse, you find the thread in the section Features. You are right that there is no direct band-stop filtering, the filtering is the result of splitting the motion in 2/3/... moves which results in the same effect. However Klipper does not use negative coefficients (just like the videos you incorporated), so the carriage is never actually pulled back. Only the timing of the movements is adjusted.
Yes, through other's comments and some more reading I've figured that out. I guess in my head, by pull back, I meant a deceleration. But I definitely did not represent that in the video, haha. I will try again one day when I'm ready to create a more airtight video on the topic!
Kann ein Input Shaper auch für Audio-Wiedergabe (20Hz bis 20kHz) genutzt werden? Kann ein Verstärker mit vorgeschaltetem Messmikrofon die Trägheiten eines Lautsprechers korrigieren, dämpfen oder sogar kompensieren ?
Tut mir leid, ich spreche kein Deutsch, aber das ist von Google Translate. Denken Sie daran, dass Input Shaper ein Steuerungssystem mit offenem Regelkreis ist. Also modellieren wir unser System mit Messungen und wenden dann die Korrekturen an. Wir messen und reagieren nicht. Es ist eine sehr coole Frage, auf die ich keine Antwort habe. Ich stelle mir vor, dass der größte Teil des Bewegungsmanagements durch die Konstruktion des Lautsprechers erfolgt. Aber ich bin mir sicher, dass es proprietäre Möglichkeiten gibt, Lautsprecher anzusteuern?
I'm unlikely to do this. I'm not familiar with Marlin firmware. I'll be making a long video on how to get Klipper! If you make the switch, then you can come back here ;)
Hi, small suggestion for improvement on video production: The background music during the introduction part was really distracting to me, so it was harder to follow you. Firstly it was a bit too loud and secondly - way more important - it had a lot of sounds in the same frequency range as your voice. So as a result your voice is competing with the background music. To resolve this you could either choose music that avoids these frequencies altogether or eq them out. Other than that - great video! I'm going to re-assess my input shaping
Hey, thanks for that! Music is something I've struggled with. Completely removing it makes it feel like something is missing. I'll be taking this into account, thank you!
In terms of RUclips analytics, this is my most successful RUclips video to date. In terms of content efficacy, I dropped the ball on more than a few of these topics. So, be sure to read through the comments as there is some very valuable information. Thanks to everyone contributing!
I video this long could really do with chapters.
Sorry, I thought I added them! I forgot to add the 00:00 one. Should be working now
Be great if you could summarize the key points of correction here at the top as there are 191 comments now - that’s a lot to go through!
19:42 - That is not something we are aware of cause we had not used AC for decades just to avoid the costs that come along with that.
We have gotten AC in the past 2 decades but only those in cars which usually do not resonate hearable.
I bet this is the american perspective which even does not fit in central europe. kWh / household consumption is far lower in the EU and awareness of energy costs too cause they are higher. The higher the costs the more the people are aware of consumption and efficiency.
Therefore you will not find trucks.
@@typxxilpsSame thing applies to heat pumps, fwiw.
That crane video finally made me understand. I use input shaping when I drive, or more specifically, when I brake. When coming to a stop, I release the brake to prevent the uncomfortable jerk you can get.
Hey, great point! Shaping the brake input definitely makes for a nice stop!
if youre getting jerk when braking you need an alignment or new rotors/pads lol
I think you're referring to a different kind of jerk. I believe they're referring to the uncomfortable sudden stop or start that comes from quickly pressing or releasing the brake. Changing the speed at which you press or release the brake would reduce this jerk. A jerk while normally using the brakes would point to an issue, obviously
@@noah_s906When you brake to a complete stop, the car will keep going forward (just a couple of centimeters) at the moment of "full stop". It will also lean forward. Then the springs will return the car back and you will get this bad jerk. If you brake real hard, you can actually hit the seat with the back of your head on "the return". If you release the brake just before coming to a full stop, this will not happen.
No.
What you apply when braking is called S_CURVE_ACCELERATION.
This is a change in the acceleration value at the beginning and end of the section.
Input Shaping is more like the work of Linear Advance, when the extruder pulls the filament back and forth so that there is no swelling on the print.
A pretty good video for beginners. But there is one little fault in the PSD graphs (min 26.30-28). You compiled multible resonace tests together what will lead to an incorrect PSD graph and subsequentely less optimal inputshaper. It is a common issue. To fix it reboot klipper after compliling one set of results.
Good shout! I didn’t realize it would mess the graphs up, just thought it was the titles. I’ll make sure to double check what I got. Thanks!
Funny thing is, the Artillery X4 Pro fails to start a second resonance test without doing a reboot 😅
There’s a pending PR to fix this in the docs too
Finally someone explains what to actually do with the data. Helpfull video to improve my IS results. Thanks!
Spencer, first of all - you did a really great job with this video and you earned a subscriber! Almost brought a tear to my eye as an engineer. Some "into the weeds" details are given below, which may help as a second layer to what you already said, for posterity:
In terms of dampening and tolerance to error, MZV is looser than ZV. And EI is even looser still. As such, a lot of people prefer to use EI on the Y axis for a bed slinger, because the moving mass will inherently increase as the print progresses. This gives you some flexibility at the expense of the theoretically ideal shaping scheme.
For the X axis, you obviously picked up on this already - but something should be addressed if you have 2 major resonances modes as shown. Is your filament spool on the top of gantry? I would be looking for decoupled masses like that, etc. On my machine, moving the spool to a holder off to the side really cleaned up my shaper results and I am back to 1 discrete peak.
Somewhat related to the above - a trick in the Klipper community is to 'pin down' the Y axis when you are measuring the X axis, and vice-versa when measuring the Y axis. This is usually done mechanically with some sort of clamping or interference fit piece to stop it moving in an orthogonal axis during the measurement. Especially useful on a bed slinger.
As a physicist, I’m humbled that I made an engineer proud😂 I’m definitely new to video making and honestly the entire field of 3D printing, DIY etc… I expect to make loads of mistakes and learn a ton!
I remember hearing of your first point somewhere. I wish I would’ve mentioned it! I have no clue where the second peak came from honestly. In my last video I stripped everything from the printer. Drawers, electronics, everything! No spool attached, no uncoupled masses (as far as I can tell). I assumed that maybe the z-axis was super unstable or something. As for the 3rd point, is it a good idea to clamp the masses down? In my head, they affect each other while printing so wouldn’t you want to have them affecting each other while measuring?
Thanks so much for your comments!
@@SpencersDesk For tbs. just remove the whole toolhead, and mount the accelerometer directly to the backplate.
Then you will know if its the motion system itself, or something on the toolhead that is making the extra peak.
A very good next step! Thanks for the tip!
@@SpencersDesk that's a really fair point about the last point, i.e. 'pinning down' the orthogonal axis to isolate just the tested axis. It makes sense that they are not restrained during printing, so this deviates from real conditions.
This would be a great question for the Klipper community. If I had to speculate - perhaps it's something to do with how Klipper implements the negative interference. Both axes are obviously prescribed different (and inherently discrete) modes, so the "composite" commands to the steppers may work the best when these are also purely discrete in the testing? Again... 'speculation' being the operative word!
Another theory could be related to the 2nd point in my original post - maybe people were getting dodgy results due to interference from other SMD systems on the machine? Whilst the steppers should be frozen in place on the other axes during this test, the reality is of course that there is some decoupled mass movement.
For what it's worth, I've never actually pinned down the other axis and have never had a problem. I'm pretty sure that the largest source of error is the quality of the ADXL345 itself. When I query the current accelerations and noise, it's not very consistent. But it generally aligns with what I am expecting, comparing to global gravity constant in Z. But usually +/- 10% around this supposed constant! Cheap Amazon parts...
woah. the diagnosing printer part was very illuminating. a lot of vids I've seen just chuck the shaper on and move on after setting the profiles.
what you mentioned, but didn't explain, is why one would have input shaper permanently on.
Hey, input shaper changes how your printer moves as discussed in the video. When you run the input shaper tests, you get parameters that will best help your printer. You then add these to your printer and enable input shaper. Without input shaper activated, you’d see no change in your print results as the printer wouldn’t be moving differently. Hope that answers your question?
very underrated channel
Thank you!!
Agreed. Couldn't believe the subs count was so low!
For now :)
very new channel
Kupi resin printer🎉
I'm so glad the SV06 Klipper upgrade did all this work for me. Just consult the manual for mount locations, print a clip for the bed then screw the provided sensor to the toolhead and run the configurator for X, then attach the sensor to the bed clip, put it where it goes, run the configurator for Y and done.
Definitely a tad easier! Convenience isn’t talked about enough when it comes to 3D printing IMO
I discovered you yesterday, I'm looking in today and I see you're off to a great start. I really like it, keep it up! I've been following a number of 3D print youtubers for a few years now, but I haven't come across such a nice analysis of resonances yet.
I would highly recommend, don't say what you think, for example about those A, B, C subsets of SPI, but get the facts and say the facts, or don't say anything at all. ;)
I appreciate how you mentioned the obsolescence of the naming of MISO and MOSI without mentioning the background of it all, and just put forward the solution as it's being named now. Similar names have been part of the electrical and PC jargon for decades and it brings some confusion now, but I appreciate that people are thinking about it and looking for solutions.
Well, it's a good start to a nice analysis, lol! I really appreciate the feedback. Sometimes it's hard to get accross the voice saying (it's good enough). I want these videos to get to the point where they're fairly airtight and don't involve guessing!
Yeah, the MISO MOSI thing never stuck out to me until some high school students were touring a lab I worked in. We had "MASTER" and "SLAVE" written on a Raspberry Pi CM4 IO Board and one of the kids asked why we had that. A little uncomfortable, haha. I'm not bothered by a change to it so why not try to help the trend. And again, COPI (co-pee) and CIPO (kippo) sound way funnier anyway. Thank you for the support!
Elsewhere on youtube, I saw someone resting their printer on a heavy cement tile (about a couple of inches thick) itself resting on some foam, claiming some improvement with input shaper and overall accelerations. Anyway, I really enjoyed this video, and hope you make many more. Thanks!
Hey, I definitely want to do that at some point. A video about reducing/fixing bad graphs might be really fun. It’ll definitely be a continuing theme on this channel for a little bit. Thanks for the kind comment (:
Great video! One important note about shorting out / damaging your electronics. The only thing that is ever critical with those communication signals is power and ground. If you triple check those, you can have any other mistake with clock and data lines - nothing bad will happen. It just won't work. You can try another combination and correct it. I'd mention that clearly not to scare people. Paying attention to that simple rule I never destroyed anything in 10 years of tinkering with electronics (at least when dealing with simple connections like that). Keep on making great videos :)
That's a great point! I just know from my own experience that once, a mixture of exhaustion, frustration, and impatience made me make some very silly errors and I destroyed a Pi. So, the extra layer of protection is nice! Thanks for your support!
Love the video! Would have loved an in depth overview of the print results, I didn’t really understand the print quality comparison between them.
Thanks! Definitely something I'll be doing a better job of in future videos.
An informative video. Great watch.
Off the top of my head, your shelf material is likely to be flatter than the floor. So the printer base will be equally supported on each corner/foot. So the printer may wobble more even on a hard floor.
On the other hand a lot of shelving isn't level, indeed some brackets encourage a slight rearward slope, which may give uneven results. It might be revealing to experiment with thicker or heavier shelving material, or even rubber mating. There may be a sweet spot between damping or encouraging wobble!?
Thank you! I remember someone (Tom from Made With Layers?) did a video on it. I think he landed on using a concrete paver stone as the base then using some foam padding at each corner. It gave the printer a sturdy base and the foam uncoupled the printer from the surface. Definitely something I plan on looking into in the future!
Also worth mentioning, that input shaper on small parts such as gears can result in artifacts! So in case you are printing something small and precise set low speeds and maybe disable IS.
Jut as an additional input:
Marlin also sports input shaping since a few minor versions.
I had very nice results by printing a test-print and measured the position of best edges and calculated the frequency. Added this f to Marlin config and perfect :)
I like to have no additional Raspi and have all done in Marlin. Also a pro: No sensor needed ;)
Thank you for explaining this so well! I was able to get two accelerometers connected to one Pico :) Being able to increase acceleration (to 5k) improved my print time a lot (takes 63% of the time it used to).
Ah that makes me so happy. I’m glad you’ve got it going!!
This is a fantastic video. If you're up for it I'd love to see you experimenting further with frame brackets, braces etc. to see how rigid you can get it and how it affects the resonance tests.
This is really interesting to me because my other passion is automotive. And what you're effectively doing here is building a race car. By making it as stiff and rigid as humanely possible, you're allowing any feedback (resonance) to be picked up by the driver (accelerometer) which can be compensated for with technique (input shaping) to go as fast as possible (same same).
If the car (printer) has loose or flexible components, then it acts in unpredictable ways which can be hard to compensate for.
Thank you so much! Don't you worry, I have many more plans for resonance! Yeah, that's a great analogy! Our goal is to have a delta function (kinda, not the whole infinite amplitude part)! My point being, we want to have a single frequency which we can just avoid creating! Thank you so much for your support friend
Your description of the wiring is good, but there are some efficiencies to simplify things.... there is no real need to use two different SPI busses for the sensors, whilst not an issue with the Pico example it's not really the most effecient thing when using SPI long term (that is to say, wastes pins that could be used for other things if you had a different project using SPI devices I the future).
You can use the same power, data in (MOSI), data out (MISO), and clock lines for all SPI devices... just attach different controller pins to each sensors Chip Select pins. An SPI device will only communicate of the MISO line if it's CS pin is high (or low can't remember off top of head). So using different pins for CS allows the MCU to select which device it's talking to. In your Klipper example it's just a case of putting a different cs_pin in the config for each accelerometer. For other projects using SPI in the likes of arduino or other IDE (or coding language) again it's just a case of defining the pin used for CS for each device .
You're spot on with that! There are two reasons I recommended wiring separate buses. The first is, simplicity. I didn't want to go in depth to SPI and how it works and its capabilities if someone only wanted to get input shaper. I just tried to give enough to give some context in a digestibleish manner. The second reason is it "complicates" the wiring. Sometimes splicing two wires together can be a hassle, and the idea of having multiple chips on the same wires while another wire had to be different didn't seem like a digestible topic in the amount of time I had. If you were wiring directly to a Pi or something and wanted it to be a permanent implementation, I'd say your way is the best. For me, I'm not planning on using the Pico for anything else at the moment, so no worry of running out of pins. Thanks for your comment! There's so much great information coming from people, I wish I could pin multiple comments...
Thank you so much, I knew the theory more or less behind input shaping but i didnt knew how to read the graphs and/or choose what shaper I would use.
Now lets see if i can get bet print quality and higher speeds
I’m glad this helped!
Great video! A suggested video turned out to be a real gem!
I have been thinking about sand filling the frame of my Ender 3 to dampen resonance and increase mass. A common practice in heavier tools such as lathes, mills and bench drills.
Even cars have resonance dampers for specific frequency mounted on suspensions, exhaust, engines and gearboxes.
I'm sure theres a specific amount of mass which will be optimal for best perfomance on a 3D printer which would be seen on klippers resonance curves. Also in video Y axis had more resonance when placed on floor instead of the shelf. Got me thinking, what if printer was placed on rails? Which made printer movable.
You earned a new subscriber just by this video! Keep it up
Thanks for your kind comment! I hadn't heard of filling the frame with sand but I like the idea! I'm curious how the printer would respond if it was sitting on a free moving platform (rails in the x and y direction). My intuition tells me that the lack of dampening wouldn't be great for it? I think skyscraper do something similar where their supports are allowed to move, but the motion is dampened. Cool idea!
Very good and thourough explanation. Great watch!
Thanks! Really trying to up my game on the videos I make. Really jealous of your Voron btw!
Think it's the first time I see any explanation at all about smoothing. Klipper docs just throw the term, everywhere else is, well all over the place.
Great content, you did a pretty good job introducing it. I'd suggest a follow up to go more in detail for the diagnostic part and understand all the chart lines.
I would also point out that the input shaper is a great way to tune belt tension. It's very easy to get it wrong and can have a dramatic impact on print quality, acceleration and even layer shift odds at higher speeds.
Thanks! Yeah, I wish they had areas where they explained the actual kinematics equations, how they implement things, etc. While this video was a bit of a fail, I'm happy that people are at least thinking about it now!
For belt tension, this is a lot easier on a CoreXY as you can see the peaks not aligning properly. For me, I'd have to test, adjust tension, test... repeat. They can help identify relative belt tension, but not absolute belt tension. So I think they can tell you if your relative tension is off, but not if you have the correct tension. Do you agree? In all honesty, I know very little about belt tension. I'm not sure how one does it properly!
@@SpencersDesk I don't think of it as a fail at all. Don't be hard on yourself. About belt tension, I have a cartesian printer (not bed slinger) and input shaping was instrumental to tune belt tension. You will see that the indicative max acceleration will increase or decrease depending of how far you are from the ideal applied tension. Of course this method only works if the print is already reasonably calibrated for other factors that may affect rigidity.
When you say Cartesian, do you mean CoreXY? So you suggest running the test over and over, adjusting belt tension in between? Until you find the highest suggested acceleration?
@@SpencersDesk, no, cartesian is one of the kinematic types you can use in klipper. It has independent X and Y axis, in contrast with corexy where they both work together. Yes, I start with some tension, measure resonances, increase tension and measure it again, until I find the sweet spot. Ideally you should take note of it and I think the easiest way to do it is to use a guitar tuner (yes, you heard me right, you can play your printer like a musical instrument), it produces different sound notes, when you fiddle with it that are related to the tension applied.
Edit: It is not about finding the suggested accelaration, but instead, observing the suggested acceleration and observing when it starts to decrease (too much tension).
I see! So more like a CNC machine. Haha, I had never thought about using a guitar tuner! That is some solid information, thank you!
Hi. Just a 'litlle' hint to 'ALLl' You Tube presentators (NOTE! Just in general): When you hold your item in your hand to show the thing and explane something of it at the same time: Please hold Your hands 'STILL', don't 'swift' 'flap' 'move'... or whats the description of my meaning (I'm a FIN so my gramma in english doesn't cover so well the thought), I hope you know what I mean. I certainly like the stuff on Your video ;)
The examples for visualizing are awesome!🤩
I think you should "dive deeper" into cooling prints, which is kinda strange topic when it comes to pla.
can you elaborare ? I'm very interested for a tldr and/or ressources 😇 I curently just have the biggest fan I had hooked up tonmy printer
I think that's a cool idea. Honestly, cooling for PLA is strange if you are printing at low speeds, no crazy overhangs, stuff like that. But (as you can see in my benchy's horizontal lines) when you begin printing at higher speeds, cooling can become a very difficult task.
Beyond that, the amount of cooling you have can seriously alter the mechanical properties of your printed parts. I'd really like a setup similar to Tom (Made With Layers) so I can begin to test things like this!
I don’t have something off the cuff, just things I’ve seen. On speed benchy printers, I’ve seen “auxiliary cooling” where there are ducts on the side of the printer pointing towards the middle of the build plate rather than the ducts being on the print head. I’ve even seen a hose being connected to the print head and a turbine off printer blasting air through the hose.
As to the mechanical properties changing, this is a pretty common trend in material science. There are all these cool graphs and plots that show how changing how fast your materials cool changes what crystalline structures they form. Rule of thumb is to go hot and slow for strong parts. Helps the parts fuse together. Going fast with lots of cooling makes parts less strong as the layers don’t have time to melt together.
I was always surprised at how little resonance my printer had during all the tests, considering it was an Ender series. Worked out over time it must be due to it being the Ender 5, which has a cube frame and doesn't use the bed slinger method for the Y. I still ran the input shaper. It did improve how quiet the printer and a slight improvement in speed, but I think the Input Shaper is most useful for the "bed slinger" style printers.
This video appealed to me both on the geeky nerdy level with the data, but also on a level where it was simple enough to understand and not have to think about htings too deeply.
Nice to see you again! Input shaper will have the great effect when your printer is at higher speeds. If you’re weren’t pushing the ender with high speeds and accelerations, it’s possible there wouldn’t be too much of a difference!
Thanks! I’m definitely trying to tow the line between exposing people to new ideas while still keeping the experience enjoyable.
Thanks for the video. Just a bit of errata with the pin labeling for the pico within the klipper config file. While the cs wire is connected to the physically second pin on the pico, it is labeled as gpio1, not gpio2, so the cs line should read cs_pin: pico: gpio1
Thanks! And nice catch!
Build my first Prusa MK1 9 years ago. Then build same concept on my own steel cube frame - big improvement. Now own Bambu Lab X1 - Carbon and no matter what I do with frame will never match performance of X1. Sophisticated software which check rezonans on 1st calibration. Extruder adaptation. 1st layer scanner. Carbon X rails with external extruder makes it very fast.
Print quality definitely starts with the construction of the printer. What is the 1st layer scanner? Is it similar to that "Beacon" ABL sensor? Thanks for your comment!
This was an really well put together video!
Thank you so much! (:
That's a huge amount of info about all I want to know about, leaving a like and hope to see you get bigger on YT bcs you deserved it! Tysm
You're getting your vyper to cook at this speed wow, im going to up the speeds of mine i thought from the thumbnail you were going to talk more core xy than bed slinging. I have been printing at 75mm/s at 3000 accel
In all honesty, I’m gonna try to push it a little faster later on. I still haven’t maxed out the flow rate of my mosquito! Look up 2 minute bed slinger benchy on RUclips and check it out! Someone did a 1:59 benchy on the bed slinger! Thanks for your comment!
Hi Spencer, nice video. The theory was neatly explained. But... (right? ha ha) you created an expectancy 0:24 about showing something for people familiar with input shaping that would "change how we approach 3D printing" - so I stayed... (I wanted to know "why I am NOT getting the most out of Input Shaper") and... where is it? If that's the charts and understanding how they work... well, that's nothing people familiar with it don't know already. By the way, having two distinct peaks at your X axis means the printer needs mechanical attention. A mechanical healthy printer has a more distinct single peak, not two as yours are showing. Finally, the printed results are still quite wobbly - a LOT of 3D Printers can do better than that without input shaping at all. Hope you can improve the X-axis and try it again. Good luck with the channel!
Thanks! Unfortunately, the theory isn't quite correct, so there's still a lot for me to learn. What I did learn changed how I approached 3D printing personally. I never knew the curves had a lot more information that could help you diagnose printer issues. Loose belts, loose screws, problematic wire looms, motion issues, etc... This will change how I test whether future projects are worth it or not. I've implemented input shaper a time or two and was familiar with the process, but didn't know about this side of resonance curves. So, I was "familiar" with input shaper. 99% of the printing community (myself included) has no idea what is going on underneath the hood. This is me looking underneath the hood of a car and attempting to explain how it works after watching a few Top Gear episodes.
As other comments pointed out, my last charts are inaccurate due to the way they were generated, so I'm not sure if the second peak is even still present. This would of course influence the input shaper recommendation and possibly lead to smoother prints. I will however say, the before and after result I show in the thumbnail, moving twice as fast and with over twice the acceleration, on a bed slinger, are very good results. The ringing towers are made to accentuate resonance issues, were in silk PLA, and exposed to very harsh light, showing every little artifact there was.
Anyway, I will be making a video in the future to replace this one that is based in fact and not my guesses, and I hope it will serve as a useful tool for others. Thanks for watching
This is a bit harsh IMO, it may not be relevant or additional information to you, but going on what is on YT this video indeed puts out additional information.
Great video from the start but you flew by the results, (the 3d printed parts). For someone who delved into details I was expecting for a detailed analisys of the parts. It deserves a brief follow up video imo.
Subscribed.
Hey, thanks for the feedback! In all honesty, it was a combination of being lazy, tired, and thinking nobody would actually want to watch it, haha. But, I promise I won't be bringing that into my future videos. There will definitely be a followup video in the future that will better address this complex topic! Thanks for your support!
I would like to mention for other people you can just get the newer adxl345 with usb c and you dont need to do all of the fancy wiring as it is just ran usb c to usb A via Raspberry Pi or whatever Ive ran it through a raspberry, and laptops. Also if you get the BTT adxl345 you can screw it in where the nozzle goes as they provide a hole and a screw to tie into the hotend directly.
Hey, good point! I forget if I discovered the adxl/rpi combo before or after the video, but they seem so cool! It is still nice to have some understanding of how things work behind the scenes though!
I love the explanations you did, they where really easy to understand :)
I’m so glad that you enjoyed!!
Hey, that's a great video.
It would be possible, that the PSD peaks you saw on the bed were the back an forth movement (typically printer y axis). I'm somewhat convinced, that the "loudspeaker" theme is not the obvious answer to that measurement. Maybe there was a mix-up of printer vs accel axes - or what could happen is, that your mount vibrated a lot, since the breakout board hangs off of the printbed itself. Just as a reference, a flat aluminum sheet of decent thickness has audible resonances in the highest of Hz, I would assume kHz - range. It's very stiff compared to your belts and the large frame with bolted connections etc :)..
I know that the ADXL was aligned with the printer properly, but maybe it could've been messed up in the configuration? I definitely agree that the z-axis is the likely culprit. In the future I want to add those rods that help support it and see if that solves the issue! Thanks for the info!
Wow good stuff sir! I thought I pretty much knew input shaper… turns out I had no idea.
Request: clearer side-by-sides. Both diagrams and print tests
Diagrams: pick out just the 2 relevant columns from the CSVs, label them properly (e.g. "floor" and "shelf", instead of "before" and "after"), put them in the same diagram instead of next to each other
Print tests: photographs side by side, same angle on both objects, clear labels
Thank you for the video, I know it's a lot of work, I recommend using a checklist for the production. There are plenty out there for you to combine and adapt
Great feedback! Thank you! Yes I’m still learning a lot about the production process, how I want videos to look, etc. this is very helpful to hear. While normally I would do the combined graphs, I think I’ll keep them separate. They’re just busy as is, and auto generated. I’ll definitely label them better in the future. Thanks for your support!
Please consider the remark this video after you learn more about input_shaper results interpretation.
Issues Example at 27:09:
-The left graphs has a results of 1e3 instead 1e4 of greater...that's mean that likely a mechanical part is binding during the test.
-in Right graphs top, you can see that two CSV files was use to make the result (you have made a average of two results) don't forget to delete previous results before making a new graphs
I feel like you skipped over the physical results. They were only on screen for a couple seconds each. Algo is picking your videos up, keep going!
Thanks for pointing that out! I definitely felt like I rushed it a bit. I’m still grappling with how much of what to include. When I first sat down to edit it, it was about 12 minutes long. I felt like I skipped over so much though and it turned into 30 minutes. I guess I was afraid that viewership would tank if I just started comparing benchys. I’ll make sure to do a better job in future videos and at least do a bare minimum amount. Thanks for the support!!
@@SpencersDesk When it comes to educational content, don't worry too much about how long the video is (other than ad rolls if that is something you care about). Honestly, the longer ones generally do better when it comes to learning, you just want to make sure you include all the important bits. You're doing good, man. Your channel will take off soon
I have my cheap Creality sat on a wobbly table. I think it makes the print better...
There’s an argument for it. Think about the way they design skyscrapers to survive earthquakes. Not so different right?
You should use the recommended input shapers for your outside perimeters, you can bump up your speeds and accelerations for most other print features, just test your max speeds and accel don't cause stepper skipping.
Good shout! Just make sure the outer walls are printed before the inner walls!
I would've really liked to have seen before & after prints sitting next to each other.
I know :/ I was so tired by the end of this video that I got lazy and didn't add it in... I promise, future videos will be nothing like that!
I found that title a bit misleading, a good intro to input shaping, but I was left hungry for more info on interpreting the graphs and the meaning of those numbers.
I've come across statements saying things like "if your recommended shaper isn't ZV or MZV there's something wrong with your printer" - what do you make of that?
Another thing I'd like to mention regarding that Z resonance you're complaining about: I've had similar problems due to the bed belt not being parallel to the Y axis, the angle it had on both ends of the carriage was introducing a vertical component to the forces, I modified the carriage model so the belts became parallel which fixed the problem.
I think the title is fair, as most are just inputting the recommended shaper and moving on. Introducing the idea that you can improve the resonances is what I was trying to convey. I would like to go more into the graphs, but I'm not confident that I could do it justice yet. I'd like to get some more experience and do more testing before I make an in depth video on it!
As to the quote... I have no idea why someone would say that. ZV is the "ideal" shaper as it lets you achieve the highest accelerations and does the least amount of smoothing. As another user mentioned, if you have a bed slinger, zv is fine on the x-axis, but ei is recommended for the bed.
For my graphs, as others mentioned, I kind of screwed them up as I was blending results from multiple tests. I can't confidently say the adjustments I did didn't remove that extra peak. As far as I know, my y-axis belt looks fine? It's parallel to the extrusion it moves over. I'll take a closer look though! Thanks for the feedback!
That's intensive information about inputshaper.
vibr refers to the residual vibration % should that particular model be applied--lower is better. sm is how much smoothing will result with that model or how much input shaper will cause the actual movement to deviate from its requested path in order to achieve the vibration reduction--lower is better and higher leads to rounded corners, loss of fine details, etc. only use this to rank the available models, outside of this, the number has no purpose. accel is maximum acceleration for that model beyond which there will be ADDITIONAL smoothing effect. it does not mean your machine is actually capable at that acceleration--you have to print the model to inspect for ringing for this. smoothing is probably not a big deal if you print artistic parts but if you are modelling walls with circular holes that need to be circular, smoothing will make them ovals which is going to be terribly annoying for functional parts. klipper's recommendation for you is based on an internal calculation to rank the models based on these values, details of which i dont know of, but is usually close to spot on. ZV ("zero vibration") is going to the the best model for an ideal printer (one resonant frequency with infinite power spectral density). since such a printer can't exist, other models will probably perform better and this is due to the printer having multiple resonant frequencies that ZV cannot handle but other more complex models can.
higher frequencies are better because this means the shaper can work at higher frequencies to counter it. low frequency is worse because if the shaper can only work at lower frequencies, your accelerations will also be lowered. that said there is little you can do to majorly alter the major resonant frequencies of your printer without serious modification. but you can improve it a little with screw tightening:
tightening screws everywhere is a good idea in general:
it can really eliminate certain small peaks as shown in the video. for example if the fan shroud screw is loose, this means the mass of the shroud becomes its own oscillator causing a small peak other than the main one (the printer frame, the toolhead). tightening it down eliminates the peak as it truly becomes one with the toolhead, creating a new, larger mass that resonates at one, new frequency, probably a higher one.
resonant frequency is proportional to the root of stiffness so if you tighten all screws in the frame, the major resonant frequency contributed by the frame should shift right as shown in the video. it will also reduce internal damping which will squeeze the peak making it taller and thinner (become more like a tuning fork, which is easier for input shaping to counter). less damping is not good but if you tighten the peak, the shaper will work better at its sole job which is to add damping.
the peaks which didn't move are probably not related to stiffness from screws.
just a warning on screw tightening in the frame. this can create skew in the frame for any printer making it lose squareness.
where you sit the printer also matters. if the table or shelf has a different resonant frequency as the printer, it is going to show up as its own peak. meanwhile the solid ground has infinite mass but also infinite stiffness, meaning it does nothing to the printer--it doesn't actually make the printer worse per se---it just reveals the true resonance of the printer itself where previously the table or shelf you had it on was simultaneously providing noise as well as damping (in most cases the damping effect more than compensates for the fact that it introduces a noise frequency which sometimes at least one of the shaper models can also compensate for). but ideally the best place to put the printer is on a stable and stiff furniture (no noise) but with damping from stiff foam. you don't want the printer to sit on foam directly because the frame needs a solid and flat surface to prevent flexing from its own vibrations. so what people have come up with is to place it on a stone/granite/cinder block slab with stiff foam below and the sturdy table underneath. this gives you a massive, stiff and flat surface for the printer while still providing damping.
last note: input shaper will alter the behavior of PA/LA. do input shaper early or else redo PA tuning after.
the new prusa slicer version also has the input shaper implemented, turns my mini+ from 40mm/s² to 140mm/s² !
I recently got a Mini and it does move a lot faster than I expected! It's cool as long as you never change anything about the Prusa machines. Otherwise, there's nothing you can really do to adjust the input shaper parameters
What extruders you all use for these speeds? You'll hit volumetric flow (melt capacity) sooner than any mechanical speed limits.
I've got an Orbiter V2.0 on my current setup. I'm currently planning a video that looks further into flow and what the biggest factors are for it (temperature, extruder motor current, etc...)
I just did this a couple of weeks ago, along with hooking up a Bltouch to it 😀
Super cool! Any other mods you're planning on adding?
@@SpencersDesk as far as the pico, I did Bltouch, accelerometer and relays, but then dug more into finding ways to utilize spare gpio pins on the actual printer MCU and started making a video on that. Then I got sucked down a rabbit hole of converting my artillery sidewinder X3 plus to Klipper after they finally decided to send me the official information on the pinout and bootloader offset I needed. Now I need to circle back around to the gpio stuff and finish up that video.
Very cool! Make sure to share the configuration you use with the community! I think you can share the example config on Klipper and they will ad it in!
@@SpencersDesk I've got two pull requests sitting in their inbox
Hey Spencer, great video, I appreciate the detail in the electronics section, and the recommendation to use a sacrificial pico; that will no doubt save heaps of frustration! I'd have loved to see a more detailed review and comparison of the results - seemed like you introduced and then skipped it? is there a section missing in the edit?
Thanks! In all honesty, the video was originally 12 minutes long. I decided that it wasn't good enough for me and spent an extra 4 days filming and editing. Towards the end of it, I had a mindset of "no one is going to see it so just get it done". Clearly I was wrong, haha. I'd just assumed no one would care by the end of it I guess. Don't worry, videos from here on out will not be approached with that attitude!
interesting, i do this manually when operating a 25t gantry crane at work.
You likely do it any many parts of your life without even realizing it! Braking/accelerating a car, catching something, etcb
I have the FLSUN PRO Delta style 3D printer. I just ordered everything you stated in the video. I'm hoping this will work with a Delta type 3D printer. Are you aware if this will work with my printer?
Input shaper is just one part of good print quality. If you print fast, the nozzle and linear advance has a much bigger impact on what looks like ringing or ghosting. Accelerometer input shaping can also give pretty poor real world results, as it doesn't measure the tip of the nozzle. The amlitude is always to low or even completely off depending where the accelerometer is mounted. Therefor the dampening values are set wrong over an average of printing speeds.
Hey, thanks for your thoughts. I totally agree, input shaper is just one of the 10e6 moving parts that determine the quality our printers can deliver. Your comment comes across as if you're saying input shaping for 3D printing is useless.
I'd argue that the accelerometers do a very decent job for our use case. While it would be ideal to mount an ADXL in the exact same point in space that the nozzle is at, we just make the assumption that the print head is rigid enough so that resonance at one point is similar to any other point. Also, I don't think that resonance depends on print speeds. Resonance is a result of the construction of your printer, how rigid it is, what materials it is composed of, the surface the printer is sitting on. From what I can tell, the same input shaper works across different speeds
@@SpencersDesk IS is important. I don't know why my comment came across that way. I meant exactly what I said. IS is just one part of a good print. However I would not use an accelerometer to dial in IS alone. The dampening values of IS can throw off linear advance with different nozzles. The amplitude of the resonance is very important at the nozzle and not the accelerometer. Therefor dampening values are to low or too high. Depending on speed printed. This can than be amplified by the pressure advance settings. You can see ringing artifacts produced on all auto calibrating printers - with accelerometers, unless tuned to the exact print speed and pressure behavior of the nozzle/filament. Not even considering the impact of acceleration and jerk settings on IS.
I see you are also a user of the coin pusher style desktop environment. I'm about to hit a jackpot.
Haha, I've never heard it described that way, but it's too accurate! Hope you don't lose your slicer!
I never heard of this before and it was very interesting. As a total neewbie, I didnt quite get, how this gets applied. If I understand correctly, this get measured once for the printer and then acts as a correction function. But as what point is it used. For example I use oftoprint with a raspberry pi on my artillery sidewinder.
Is this correction something that gets applied in the slicer or does it need a totaly different way of controlling the printer directly with the pi?
Hey, great question! Klipper itself dictates how the printer will move. The slicer says how fast, and what direction to go, but the actual motion is determined by Klipper. So, Klipper takes the gcode you give it, and determines how the printer will move. If you’re not on Klipper, then other firmwares can also do this. For example, prusa machines have input shaper built into their firmware (depending on which version you have) Hope that answers your question!
What a well organized video!
If possible, can you add a video chapters?
Thanks! It's my first longer video. Looking back at it, there are plenty of improvements I could have made, but that's okay! Chapters are added now.
@@SpencersDesk Thanks 😉
Your explanation of IS was really intuitive and informative. Can't imagine how much time would have put into this video
Very good quality!!! Congrats!😊
great video. the theory is similar to active noise cancellation
Hey, yes sort of. The way I described it was a little wrong. We aren’t “canceling” the oscillations but instead “preventing” them. Noise canceling is definitely the canceling version!
Pico & pack-o-sensors ordered!!! Now I just have to get back to trying to make Klipper work! :)
I'm working on a long form Klipper installation video! Hopefully you get it figured out before that comes out though!! Best of luck
@@SpencersDesk - hope so! My PiOS install is pretty gimpy, so I might flatten it and just start over. Not that hard to put in KIAUH and MC, and all that now that I've done it once. Pi won't even see my machine now, so maybe I gimped the USB ports, or... who knows!! :D
Hello, do you have tutorials on how to install Klipper on the printer step by step?
I have seen many but they don't do it step by step, please
I don’t have one yet… but I am planning on making one!
Quick questio, but kinda complex. Is there any way to combine input shaping results from the accelerometer on both the head and bed for one axis and input shape based on both resonance values combined? I know the values are far lower for X on bed for example, but would there be value in compensating for that as well? Just a thought.
I’m not 100% sure I understand your question. Do you mean, take measurements on both the head and the bed separately then combine them together? Or are you referring to the head having resonance in the x, y, and z and compensating for each of those?
@@SpencersDesk Yes taking readings on both and combining. As in, combine the X reading for both bed and head measurements and apply it to X movement and same for Y reading and the bed. It's probably just not significant enough to worry about.
Great Video. Thanks. What about the cross correlation with Pressure Advance? When I did the input shaping I had to repeat my Pressure advance calibration. And I found pressure advance to be very sensitive to the acceleration I am using. Did see any impact to your pressure advance setring when you have changed the acceletions?
After input shaping I was able to run with 3500 mm/s2, but only for the generic calibration test. When printing real parts the print quality was6not good, and I think because of the cross correlation between acceleration and pressure advance.
Hey, great thing to point out! I did recalibrate pressure advance after all of my benchy’s. I should’ve pointed that out. I wouldn’t call it a cross correlation as it’s kind of a one way relatiknship, but yes, changing the accelerations requires you to recalibrate PA. Thanks!
Very helpful vid. I was thinking i'm lazy and it would be great if I can just leave the pi pico on the machine, but my raspberry pi3a runs hot and I'm concerned about the power draw. So I was wondering if I could connect my pico through a U2C [usb to can already takes power from the power supply so shouldn't strain raspi] converter with canboot and plug pico into that. I'm assuming the pin callout would look different, but also like would it be possible.
Truthfully, I'm not sure! Is the Pi3 running hot because of the power running through it? Or just from computing?
increase your accel per hz and you'll get cleaner graphs. the power spectral density is only scaled up to 1e3, which is usually an indicator that you're measuring a lot of non-data
Is that a parameter in the resonance testing sections? I’ll look into that. From what I’ve seen the 1e3 is pretty typical. What do you mean by non-data? If you mean noise, I don’t think the smaller peaks I’m seeing are noise
@@SpencersDesk yeah, it's in the tuning parameters, though it's used less. it's real data but the actual impact on printing is next to nothing, so while not noise, it's not as useful as the major peaks that will show up better with higher accel per hz. I guess you could call that noise if you wanted, but it's not reaaally noise in my estimation. it's still useful data, especially since you found loose bolts, but not what I would base my input shaping on. most of the machines i see have at least 1e4, but that could just be the types of machines I'm used to seeing
I see what you mean. The small peaks are real data, and have some cause, but you’re not taking them into consideration when determining an input shaper. But yeah, I think e3, e4 is normal range
Have you tested how much vibration is acceptable? If you use the less conservative of the shapers, do you notice a reduction in print quality?
I have not yet. I am planning on investigating whether the recommended input shaper is worth listening to/if adjusting its recommendation is any good, so hopefully I’ll have some answers for you!
@@SpencersDesk Thanks, I'm going to go a little above the recommended accels for more speed and see what happens.
there is or was a problem.. not sure if still present. the very first query to accelerometer after firmware restart on a pico will always fail. the second will succeed. dont give up to soon. (has todo with late spi initialization)
I'm not 100% sure but I don't remember this happening when I did it. I definitely should've mentioned that the first step is... try again. Thanks for the comment!
Great video!! Really needed this, much appreciated.
Thank you!! I'm so glad it was useful!
I would definitely love if you could dive a bit into the limited cartesian kinematics. From what I have seen it allows you to have a separate x and y axis acceleration which allows you to fully utilize the maximum input shaper speeds without being limited by a certain axis.
For example, at around 23:25 If I wanted to use the input shaper EI with an acceleration of 3000, I wouldn't be able to utilize input shaper 3HUMP_EI to its full 3300 acceleration.
I may be misunderstanding your comment, but for Cartesian printers, you are able to define different input shapers for the x and y axis. The same goes for Delta and CoreXY or CoreXZ printers. I don't believe there is any printer out there which behaves the same in both the x and y direction. So, you will need to assign different input shapers to each one. You definitely can have 5,000mm/s^2 on one axis and 1,000mm/s^2 on the other, it shouldn't matter
@SpencersDesk my apologies, haven't done a ton of research on it. It more on setting independent acceleration limits for x and y axes. I can't sent links in comments but if you Google "limited cartesian klipper" it's the first link!
Hey, I checked out some links and it seems to say that Klipper, while letting you define separate values for the 2 axes, still only uses the lower value. I saw that someone developed a method to change this? But it isn't officially supported yet.
@@SpencersDesk Gotcha! Yeah someone made their own python script for the kinematics. Would you see any benefit of doing this or should I do more research before implementing?
I can't speak to it honestly. If you're not satisfied with your printer's current state then I'd say go for it. But, I don't know how much benefit it brings. Maybe others can speak to the difference it makes?
cannot belive there is a signgle thing thats this complicated that people are going to go through if its not their job and if they are not qualified. 3d print community is sth wild really. Like everyone are programmers.
I appreciate your work on the topic.
Unfortunately, this only applies to one printer model.
For me - a Delta owner - this information is only a theory.
The theory applies to whatever printer model you have. It even extends beyond printers, so it should still be useful on a delta too!
Great video, thx! Really informative and helpful.
This is a lot of work. Are there printers that come standard with this?
Yes! Prusa recently added input shaper to their printers. Bambu labs does it a bit better as I think they can measure the resonance when you set it up whereas Prusa does it at the factory. Maybe that makes it account for the work surface better? Or maybe the Prusa is hand tuned and better that way?
Only 7 minutes in but the entire time I was wondering how do you input shape on a bed slinger printer. With the bed moving in the y axis and the print head moving in the x axis where do you put the senors in order to capture both oscillations. Honestly I felt like input shaping was only for CoreXY printers where the print head moved in both the x and y direction. I'll continue to watch.
in classic fashion less than 2 minutes after commenting, you can just use two input shapers.
Well enjoy the free engagement
Haha, thank you for the free engagement!
Great video, subscribed and can't wait for more 😁
Thank you so much!
Hi Spencer are you happy with your extruder? is it lightweight?
Hey! If you’re looking for a direct drive setup, I think this extruder is the one to go for! Made With Layers or CNC Kitchen did a video comparing a bunch of extruder motors which might be more helpful to you! I think in the future I’m going to opt for a Bowden setup and see how that treats me!
@@SpencersDesk Thank your for answer, Nice Content! I. My Bondtech clone is just fine with simple screw and nuts mod (without spring) but I need a more versatile Extruder, I checked The Orbiter 2.0 but I didn't like plastic gear idea in a filament Extruder... After lots of research I decided to buy MicroSwiss NG Linear Edition. Thank you.
Could this be what saves delta printers after all??
I liked the video and the info; I thought it was well done. However, I would have liked to see more of the actual print results.
Future videos will have this! Sorry that this one did not :/
Great work! Thanks for sharing👍
I don't have a printer but how about this technique. Draw/print the line then stop the printer at the corner then proceed downward. Would that make for sharper corners?
If your printer was moving slowly yes. At higher speeds, the corners aren’t sharp anymore due to the printer not being perfectly rigid. The advantage of input shaper is that it can create these sharp corners at higher speeds!
Thanks for the explanation!
Hi, what music do you play in the background?
would putting fuses between the adxl and the pi board prevent shorting?
The shorting is mostly a risk when plugging things in. So, it shouldn’t be necessary as long as you’re careful!
@@SpencersDesk "as long as you’re careful"
i'll buy a pico
As the print grows higher, the mechanical oscillator for the X axis changes. Can the filtering be done for the whole range?
Hey! I really like that question. I’d argue that oscillation along all 3 axes change. I didn’t think of it until you mentioned it. I would say that the oscillations don’t change an insane amount though. As they mainly come from the resonance frequencies of rigid bodies, I don’t think the change should be too great. So my short answer is yes, input shaper will be effective at all heights. In the future I plan to add supports to the zframe and I’ll be sure to test the resonances at different heights to get a more concrete answer to this!
The "after" graphs are an average of three measurements (see headlines). Where did you probe for the final measurements?
Hey, nice catch! Unfortunately, I didn't think it mattered at the time and rolled with it. Others have informed me that's not the case. The probe point did not change between tests.
So - no rigid mounting to shelf?
How much of compliance? Rubber feet or floating platform?
Nope! Nothing fancy here (yet). Right now it's just the stock rubber feet. In the future, I'd like to test out different options
Quite often you had way too much smoothing: 0.3 is crazy. Even 0.10 is quite a bit: you gotta go slower or move to linear rails.
I just moved to linear rails (video posted today) ;) I definitely still have a lot of learning to do on managing resonnace. I definitely think it's silly to expect great print quality at high speeds, so I printed at high speeds knowing there was that trade off, as I was trying to accentuate the errors. Thanks for your comment!
at around 4:15 you said the 2 videos are linked in the description? I cant see them anywhere?
look at the bottom of the description, or you could click the little "i" circle at the top-right of the video player. happy printing!
Hey, thanks for pointing that out. I forgot to add it! It's now in the description.
Great video and explanation of the graphs. Did you make the power cables for your printer or did they come with it?
Hey! Thanks!! If you’re referring to the orange cabling, I did that in my last video! If you search up PET wire loom on Amazon or something you can find it! Orange is my personal favorite color so that’s what I went with but black is the most common! I have a link to the orange loom I used in the last videos description!
@@SpencersDesk must have missed that part in your last video, I’ll take a look. Did you make your own cabling for the printer? Looks like the cabling fit it is super long.
Only asking because I have a ender 3 pro and I don’t like how short the cabling is and I want to rewire the machine
I sure did! I rewired every single electrical component. Steppers, limit switches, print head, PSU... I made them each super long because I plan to one day print high temperature plastics with this printer. Moving the electronics out of the printer enclosure is a must so I went ahead and installed the extended wiring I will need.
I will say, the steppers were a bit frustrating. On the Vyper, all the wires were black. This made figuring out how to wire them to the controller board a bit of a guess and check process. It also requires some extra tool for crimping the connectors. Let me know if you have any other questions about it!
Bigtreetech made a couple iterations of rpi2040 adxl so it’s a standalone solution
I learned of these after this video! Not the BTT version, but there is someone that makes a super cool looking one!
thanks a lot - subscribed
Not sure if that would improve bambulab printers too
Hey, I’m pretty sure that Bambu has this in their printers already! I think it’s tuned when you first setup the printer
Can i use this with 8bit board?
Hey, I’m not sure. I’m assuming you’re using a Pi and Klipper, but the 8bit board may just be too slow. This is mainly useful at high speeds, which you may not be able to reach with the 8bit? Disclaimer, I’ve not used an 8bit and can’t be sure. I would recommend a new board! They’re not too expensive for a reasonable one! I’ve got a video where I replaced my board if you need some inspiration!
My ender 3 v3 se without input shaping can do fine at 300mm/s but i usually print at 150mm/s
I pray I may obtain this power some day
How do I get input shaping on my ender 3 ke, I spend 20 bucks and plug in the sensor that should have came with the printer in the first place.
Hi, I’m not familiar with the E3 V3 KE but I just looked it up. It mentioned input shaper but I have no clue if it has an accelerometer or not. It may be like the Prusa machines where input shaping values were found at the factory and are built into the firmware already on the printer. You may already have it activated by nature of having the printer. Hope that helps!
@@SpencersDesk yes and no, I have an option to adjust values on the adjust page , however when I tap it, I get a prompt telling me that I don't have the available for mentioned "sensor", vibration sensor. Some call it a G sensor.
Odd. Could also be something like a firmware issue? I'd at the very least reach out to customer support for some help!
@@SpencersDesk no no no I've got the input shaping finished, I literally had to buy the sensor, a sensor that comes factory on some models.
Does it really tell the printer to go BACK? I thought that it "cuts the corner", since frequencies are REMOVED and not added: going backwards involves adding high frequency movements. But in general the results are the ones you mention
Imagine using a Yo-Yo. When it gets to the bottom of the string, you do a jerking motion to get it to come back. It's similar to that, where it pulls a little extra back in the other direction to cancel out the move. It's more of a controlled stop then just slamming on the brakes. It isn't just cutting the corner, as you wouldn't need to measure frequencies to round a corner. Think of the ripples in a pond when there are splashes in two separate locations. The ripples cancel out in places where they have opposite amplitudes. Input shaper guesses at what the opposite of the result would look like, and does that to try and cancel it out.
@@SpencersDesk I understand what you say and I also see that that's what they do in the video of the object hanging from strings, but I thought that klipper just band-blocks the resonating frequencies and that implies smoothing the motion, not adding a high frequency compensation (the "pulling back").
After all, excessive compensation causes smoothing of the corners, which is a confirmation of my interpretation of the correction applied.
Let me ask the devs on the klipper discourse. If you look for new threads in few hours you'll be able to participate to it.
I will say, by adding two identical waves that are phase offset by 180 will result in no oscillations. In other words, adding more, but opposite vibrations will lead to no vibrations. But yes, as others have confirmed in here, it is a changing of accelerations. The ZV smoother splits acceleration into 2 steps, while other shapers add more and more steps, becoming more robust against error and change, but cause more smoothing. I don’t thinks it’s a band pass of any sort. And sounds good! I can’t wait to see what comes of it!
@@SpencersDesk The main developer of the IS module answered on the Klipper Discourse, you find the thread in the section Features. You are right that there is no direct band-stop filtering, the filtering is the result of splitting the motion in 2/3/... moves which results in the same effect. However Klipper does not use negative coefficients (just like the videos you incorporated), so the carriage is never actually pulled back. Only the timing of the movements is adjusted.
Yes, through other's comments and some more reading I've figured that out. I guess in my head, by pull back, I meant a deceleration. But I definitely did not represent that in the video, haha. I will try again one day when I'm ready to create a more airtight video on the topic!
great stuff bro!
Nicely done .. ThankYou for sharing .. Cheers :)
Kann ein Input Shaper auch für Audio-Wiedergabe (20Hz bis 20kHz) genutzt werden? Kann ein Verstärker mit vorgeschaltetem Messmikrofon die Trägheiten eines Lautsprechers korrigieren, dämpfen oder sogar kompensieren ?
Tut mir leid, ich spreche kein Deutsch, aber das ist von Google Translate. Denken Sie daran, dass Input Shaper ein Steuerungssystem mit offenem Regelkreis ist. Also modellieren wir unser System mit Messungen und wenden dann die Korrekturen an. Wir messen und reagieren nicht. Es ist eine sehr coole Frage, auf die ich keine Antwort habe. Ich stelle mir vor, dass der größte Teil des Bewegungsmanagements durch die Konstruktion des Lautsprechers erfolgt. Aber ich bin mir sicher, dass es proprietäre Möglichkeiten gibt, Lautsprecher anzusteuern?
Prinzipiell ja, aber ich denke damit geht auch das Spektrum des Lautsprechers verloren.
Very interesting, Thanks!
Please do a video for Marlin firmware and getting data from the adxl with arduino board
I'm unlikely to do this. I'm not familiar with Marlin firmware. I'll be making a long video on how to get Klipper! If you make the switch, then you can come back here ;)
Hi,
small suggestion for improvement on video production: The background music during the introduction part was really distracting to me, so it was harder to follow you. Firstly it was a bit too loud and secondly - way more important - it had a lot of sounds in the same frequency range as your voice. So as a result your voice is competing with the background music. To resolve this you could either choose music that avoids these frequencies altogether or eq them out.
Other than that - great video! I'm going to re-assess my input shaping
Hey, thanks for that! Music is something I've struggled with. Completely removing it makes it feel like something is missing. I'll be taking this into account, thank you!