Great Point! I got to check how every change affects the results from previous test. I filmed a timing test with the 11-degree gate rotation and the lower marble position but i edited it out of the video to stay on topic. Standard deviation was 0 ms still :)
@@Wintergatan I'd also suggest adding one more parameter: how fast you can drop marbles through your test setup. As speed goes up, so does the energy in the system. You should test for any effects marble throughput has on the performance of the gates and the resulting repeatability. Damping may be needed as the marble makes its last contact with the gate, and a softer material might help.
@@AndersJackson Well, if things are down to aiming a ball to drop in a specific area and within a tightly controlled time window and there is an expectation of high repeatability, I'd suggest doing that speed test early rather than finding out later that gate operation repetition rate mattered. The tests demonstrated in this video allowed a substantial time for the balls to stabilize in the queue whereas in the application there won't be several seconds between each release. Also, the queue depth will matter because there will be varying amounts of pressure on the system, which might require more of the channel to be modeled before arriving at an acceptable design. Even if one felt that gate design would not be affected by these dynamics, a feeling or belief is not engineering proof.
In software engineering it's so easy, you just create automated tests and you're done for any regression. In mechanical engineering it's soooo much worse and random and exhausting and this and that... I'm so glad to be a software engineer when I see Martin struggle with this huge list of constraints that are all super hard to put together 😂
The most surprising thing here was that Martin wasn't whining about it not being within a 2 nm tolerance. Hearing the words "good enough" was literal music to my ears.
He need to be designing each part to perform better then good enough, because he has to combine several parts, and all the imperfections will add or multiply up. The imperfection in timing will add up. Like the programming wheel and pick up etc. So have 0 delay and variation is good news. That will be needed later. And probabilities are multiplied. So if you have two parts that have probability of 0.9 to drop a marble, you have a system that have the probability of dropping a marble that is 0.9 * 0.9 = 0.81, which is way worse then 0.9. So you see, he HAS too design for as good as he can, to get a system that performs good enough.
@@AndersJackson I completely agree, but trying to make something absolutely perfect can be quite draining on the person. sometimes "good enough" is good enough :3
You say that, but he's gone about this backwards. He needs system requirements. Before doing any of these tests he needed to form his non-functional requirements. Why isn't 40mm good enough? Does he actually need more accuracy than that? He might, he might not, the point is that is is still arbitrarily chasing the highest precision he can, even if there is nothing to gain. He should have an exact number before going into the design. No matter how many times he says the MMX could never work, the problem is that his is striving for a precision and accuracy that is completely unnecessary and arbitrary. If he wants to be precise that's fine, but he needs to define how precise that is. I think he has gone way too far. With the timing he went far beyond the limit of human perception. With this I'm assuming the main issue he is trying to solve is bounce reliability and therefore making sure marbles don't end up on the floor. But marbles falling on the floor was never a critical failure (assuming in small volumes) yet he made it out to be the end of the world. Not every failure is a critical failure, not every failure needs to be 100% removed.
@@EricTrimbur the amount of marbles lost in the MMX was way too high and for this it's great to make sure every part can work better than they need to, as that means that any inaccuracy added later on is (effectively) already compensated for
@Wintergatan should really read this comment by @@EricTrimbur. I think we are seeing this issue because starting with general goals creates a lot of unknowns. Martin should go back to music, analyse the music he wants to make, then create a machine that can play that music. I was a fan of Wintergatan before this project. If we see more Wintergatan music, that is no loss. Martin should: create music, analyse that music, create instruments that can play that music, experiment with the instrument(s). This is the old recipe, he's done this before.
Martin, at first I was worried when I heard you say, "when designing a marble gate, the goal is to drop the marbles as accurately as possible." But then you explained that your tests show much more precision than you'll ever need. That's the most reassuring sentence I've heard during all these tests. If you carry that through with the other challenges, and make sure things can be "good enough", I believe in the dream. Just try to change that first sentence; when designing a marble gate, the goal is to drop the marbles accurately enough, and reliably enough. Now you just have to define "enough".
He's not reaching for perfection. As stated, he's trying to be as accurate _as possible._ By probing the limits of possibility he can be very confident in achieving realistic tolerances. It's a very reliable approach to designing. It's a bit like stress testing a structural member. First you test the material to destruction to see how much load it can _possibly_ take. Then, when you want to build something to take a certain load with a certain safety factor you know exactly what dimensions to make it.
As good as possible does not imply an attempt for perfection. It is about finding the edge of the envelope… the beginning of diminishing returns…. If that is beyond the needed margin, the confidence improves.
Im so tired of people gatekeeping martin as if he is some mad scientist descending into the endless pursuit of perfectionism. No Martin knows better than any of us that when you settle for good enough these factors compound for years and create the brutal endgame of the 2nd machine. He is being strict with himself now to see if tis possible to complete this project. How people fail to see this is insane.
increasing the target aria will increase the life of the target. Just doubling the diameter of the target will increase the number of hits to fail by a factor of 4.
@@awogbob I agree completely, and think Martin is doing exactly the right thing for getting a functional, world tour ready Marble Machine, and he was right to abandon MMX for that purpose. However I do think Martin could've gotten MMX to a better endgame and gotten it to perform one song for all of the people who believed in him (and funded him!) all those years. Hell if he got the original Marble Machine to play a song, he definitely could've gotten the MMX to do it. I think that's the sting people feel when they bring up his "perfectionism". But I agree that Martin is taking the right approach here in ensuring it's even possible to address the compounding issues before designing number 3.
For future testing of drop accuracy, I strongly recommend letting the marbles land on a layer of carbon paper. Like your old-school typewriter, the marble will act like a hammer, transferring a spot of colour onto a blank sheet of paper underneath. Sheets of carbon paper used to be how you could make several copies of the same letter when typing on a mechanical typewriter. You can buy a sheet from an office supply store.
I was also think while watching this might not have being the best approch for what Martin goals are. These test is great for measureing accuracy but based on his discriptions and what he should be try to test at this part of design process is precision. To get everyon one the same page ," accuracy" is how close test are to a target while "presisions" is how close test are to other test. At this point the goal should be high level of precision from the dropper. accuracy can be figure out later though calbrations. At his point you just need to keep at the back of the mind you will need some method of being able to aline the drops at later date. These test where measure as much differences between avg drop point and the center of the target as it was how consitent the test was. For test like this its just as important to track the types of fails as it is that there was failer. If it alwasy failed to same side this may be cause by bad calbrations rather then inconsitancy. I am not how much work went it to the calbrations of set natural it looks like this was edit out of video. The use of the phone camra different gave better data around precision then test did. I do think James Dominguez typewriter idea would give much better picture of precision of the dropper.
That is actually a good idea. Just drop as many marbles as you can (with a rudimentary marble loop), and you can get 100-200 marbles through it and see in the color the deviation of the paths.
@@AndersJackson - I have a very vague memory of doing exactly this back in high school physics, launching glass marbles off a ramp and onto a sandwich of paper and carbon paper, giving us a visual record of the impact grouping.
@@jamesdominguez7685 Uh, you realize the green target spot he's showing is practically a non-antiquated, digitally scanned "carbon paper equivalent" contact/hit spot analysis..?
Good work Martin! But do not forget to backtrack your progress to previous steps! Check if your changes in the angle and position of the release mechanism changed the timing precision for example! Keep up the great work!
Also, I'm sure Martin has thought of this, but if he does build a third machine he's going to need to take into account just how level it is and any vibrations, because with the kind of precision he's aiming for even the slightest misalignment in how the machine is standing (or slightly moving) during use will throw things off.
He's going to immediately run into it. "Backstroke timing" is for rapid drops. And the gate he used here is double-action which will inherently have a time delay between drop 1 and drop 2. And rapid drops will cause vibrations in the gate. His tests are with the gate securely fastened to a solid surface. It's going to be floating, if his prior designs are any indication of where he is going to go.
Yes, he is learning, that is why he in the previous video targeted +/- 0 ms variation in the drops. Because he could with not too much work. When he have all data, he can adjust so the marble gate is good enough to have the marble machine to also perform good enough. That is why each part has to be performing better then the performance of the machine with all parts put together.
Yes Martin. This is very good science. I really liked the Shadow idea. Importantly you're also bring realistic about the smallest diameter, not arbitrarily going for the 21mm hole. Keep this stuff up, really good.
Martin, you actually have SIX design roadblocks. Remember for the marble drops there are at least two other factors that the gate cannot control: 1) ambient air movement between gate and the chime bar which can affect the marble in uncontrolled flight; 2) the angle of the chime bar, needed to direct the bounce, means that there will be a fractional difference in the drop length/time depending on were the marble hits. Those two factors complicating perfectly tight timing suggests that the sixth design roadblock is perfectionism. It's not a medical device or a submarine or a rocket. It's ok to be good enough.
He is not aiming for perfection. He is aiming for a good enough system. But each part has to be as good as he can get, to be able to get the total system perform good enough. The final product is not this marble gate. It is the total system. Each parts probability of not failing is multiplied: If you combine two part that have probability of working which is 0.9 (1 failure in 10 drops), you will get the probability of 0.9 * 0.9 = 0.81 probability of working (that is probability of 0.19 to fail). Each parts timing variation is added: Have you two parts that have +/- 10 ms variation, then you get a total of +/- 20 ms. The probability of failure have several parts in the marble loop: the marble lifter, marble delivery, marble gate, marble catcher, marble return to lifter. The delay have other parts, like the programming drum, pick up, mechanism to control marble gate, marble gate, distance to instrument impact. And the system performance is each of those combined. So have great performance on one part give him some slack on other parts, which probably doesn't perform so great. And even so, something that is great on one attribute on for instance the gate could give worse performance on one other attribute in the marble gate. When all this is done, he can decide what he can adjust to get good enough performance. So yes, this isn't the road to perfection, this is the road to good enough.
@@cerrudmanuel Manuel, you're correct. I was thinking more about wind across an outdoor stage. Even then the short distance might not matter much unless the target is small and the weather is windy.
He has not changed anything from previous video. He is still aiming for as good performance as he can get. In the previous video he could reach perfection with quite small work, so he did. It will also give him some margin on other parts in the timing path where that isn't possible (programming wheel, pickups, mechanism between pick up and marble gate, marble gate and distance to instrument). Each variation in delay in that path will add up, ruin the goal of playing good enough. Same will happen with probability of the machine perform without failure. There it is even worse, as the parts are in serie and those probabilities are multiplied. Probability is between 1.0, where it never fail, and 0.0 where it always fail. If two machine parts have probability to work 0.9 (which means it's probability to fail is 0.1, 1 in 10), put in serie and failure is when at least one fail, the probability to work is 0.9 * 0.9 = 0.81, which gives probability to fail being 0.19. So nearly combined probability of failing 0.2, instead of each parts 0.1. So he needs margins in each part. And when he have done all tests on the marble gates, he can combine the different attributes, as some attributes could working against each other. But then he have some data to choose the final(?) design where he can put into MMXX, if he can design all parts to be good enough for the total system. So yes, that is great to reach perfection in tests, but not critical. It is the best combination that will be the final design.
Hey, I'm probably not your target audience, and I'm not sure you'll ever see this, but I wanted to say that I love that you're still doing the marble machines. Every, maybe, 6 months or so, I'll check back in, and seeing new videos, all with marble machine based content just makes me so happy. So, thank you for your passion, and your persistence. You're really an inspiration.
Love that you’re back at it on your channel. I am thoroughly enjoying your journey. And I love the music Wintergatan has made. Keep up the awesome work.
You should re-check the timing in this new orientation for the marble gate, I think last time we determined that touching the marble like this degrades the timing accuracy.
Good Enough. You're about to go take a Six Sigma course, and you don't need it. The original machine was glorious, and so was the second! We love the art.
I've been following the MMX since its announcement and its kind of amazing how Martin has slowly but surely learned hard engineering lessons and taken them to heart. Just remember that all those fantastical mechanical instruments from centuries ago took many years to make as well. Can't wait to see your next set of engineering experiments. What an amazing journey!
Everybody else has good suggestions. I would also just say that the precision of your stacking of the hoops perfectly centered on each other becomes a concern as you get down to clearances of just a couple millimeters. I'd suggest putting some self-centering notches and protrusions on each ring so that they are guaranteed to be perfectly concentric as they stack on top of each other. And also, some sort of locked-in, unmovable, repeatable mount for the arms of the rings, so the center doesn't shift over time as you fiddle with them adding and removing mics and such.
it doesn't really matter because he can tune the position before doing the 30 marble drop test. and because he is filming from above he knows where the marbles are hitting and by how much, so if every marble hits at the same place, he knows that he fd up the setting of the ring and that it isn't a precision problem of the marble gate.
I just want to say, I'm very happy to see you're uploading once again and I hope you're able to find it in you to take another shot at the marble machine. I completely understand your perfectionist attitude but I believe I can speak for many of your viewers that we're just marbles happy to be here for the ride.
Don't forget to consider the impact of vibrations - if you mount this on a marble machine and it doesn't have adequate stiffness, the vibration of the machine will impose extra lateral movement on the dropping marbles.
Martin the musical engineer. Let me give you some advice. Practice doesn’t make perfect, it makes improvement. Iteration is the same. Just as you had to practice to improve your skill at music, so to must you iterate to improve your skills at engineering. And these videos are documenting a wealth of valuable information that future engineers will learn valuable lessons from. Please never give up. We believe in you, and believe that it is possible to build a marble machine that will surpass your expectations. It just takes patience and iteration.
I am SOO with you on this. Thanks. And I am soo sad to see those that post negative posts. Because they obviously are not engineers that know that probabilities when combine in serie multiply (and deviation in time adds up). So design with margin will get you in the specs of what he need. Just aim for parts that are within the end specs WILL end up with a machine the are not within specs.
@@AndersJackson Thank you both! It's so frustrating to see all the negativity. It's like people are trying to find something to be mad about and aren't at all focusing on what is gained throughout the process, whether it's functional parts or better learning for the next attempt. Either way, it gets Martin closer to a working machine. Even if this one were to not work out again, I've enjoyed these videos SO much over the years!
You've become a problem solving machine. I can tell by the low amount of iteration you took to improve the precision, without surrender to over engineering. I'm pretty confident that your work from now on will be clean and lean. It's notable your personal development and I'm proud that you didn't gave up on this exhaustive journey of continuous improvement. :)
Martin, I am so greatful that you're still going forward with this project. You inspire me to be a better (software) engineer in my day to day. Your new process feels a lot like test-driven-development in software. This video definitely gave me unit testong vibes
This is interesting! I think the reality is that outside of the lab, you're going to be dealing with unavoidable forces like the tilt of the ground at a site you play at or the vibrations of the entire machine as things are moving (hell, maybe even the temperature and humidity lol), but it is a really good illustration of how even REALLY tiny tiny adjustments can make really noticeable differences. Of course, figuring these out now in the lab could end up being good for having the tolerances to deal with less than perfect conditions in real life. Either way, it's good content and I hope you're finding this rewarding in its own way!
The tilt shouldn't be a big issue as he always planned to have the machine leveled wherever it is before playing. He has also been trending away from materials that are susceptiple to temp and humidity changes towards more inelastic ones.
@@michaelkrelwitz6203 For sure, but either way, I think this is an illustration of how very small (and often uncontrollable) factors can still have an impact. Which is why in a real-world situation, things aren't going to match this level of precision (which he understands, of course). But I am enjoying the process regardless, which is the point of all this, so I think this is time well spent. I don't think all this will eliminate all the roadblocks if and when he does make MMX3, but this is an interesting new approach that I'm glad is being documented for us.
I think if "REALLY tiny tiny adjustments can make really notifiable differences" then this is not a good design solution for the machine. I would say the goal should be to have a design so robust that it shouldn't rely on pretty fine tuning to make it wprk properly.
Seems to me he's trying to reduce randomness and learn what exactly goes into tuning each component. When everything is put together all the randomness of every component will be summed up too. Tho wouldn't surprise me if the machine's vibrations or some other thing would invalidate some of the tests and make it perform worse.
I was worried about perfection over progress in previous episodes, but Martin has assuaged those fears today. This really does seem like the kind of thoroughness and "from first principles" design that I would expect from someone aiming to reliably succeed. Instead of building a machine, then endless failure mitigation, prevent the failures from coming up in the first place. The design will follow. Keep up the good work, Martin.
I personally think the sway of your table is adding quite a bit of variance to the path the marble takes, which is most noticeable at 9:00 or so. This isn't huge, but any variance in the sway becomes multiplied by the distance it travels. Still, in a machine like the one you'll likely build, there will be a far higher sway and vibrations to kick the ball around. If you can achieve the 4 mm tolerance in a machine like that you should be handed a prize, It sounds pretty much impossible. Good shit my man!
@@michaelkrelwitz6203 he doesn't need to do that, as the construction of the holders etc will decide how much the machine will sway. There are ways of compensate for that too. How much that will add to the precision in landing area, which this actually are, can be calculated. So no need to add that variable to this test. He is testing one thing at the time, like a real engineer or scientist do when testing. Each variable removed when testing is one less complication in interpretation of the result to care about. For fun? Yes he could. To test the performance of the gate alone, not so much.
One of my favorite quotes about design is one that Destin (Smarter Every Day) said on his podcast: "This is acceptable because I have to move forward." We would love to iterate and tweak and change some things forever, but eventually we have to accept "good enough" because we *have to move forward.* I like this approach better! I love how you can optimize a bit until it's robust and good enough, then acknowledge that it's "all the precision you need." I'm glad to see you posting so much again! I still believe in the MMX!
@@YogiTheBearMan and @creageous , I do care! So to follow your own teaching. Write down your suggestion on a paper, put it in an envelope, stamp it and send to someone who cares! 🙂 🙃😎
*@Wintergatan* 1:55 Look up "Binary search", it will help you find your answer faster. Basically start in the middle, if the marble touches = increase the hole-size, otherwise decrease the size. But you do not go +1 size, but you jump to the next that would be the "middle" size. In this case (should have started at 30, but anyway), because 40 did not touch, jump to 30 (because 30 is close to halfway between 21 & 40) Formula: *40-21 = 19* _(the size difference between max & min)_ *19 / 2 = 9.5* _(divide that difference in half)_ *9.5 + 21 = 30.5* _(add that half-difference to the smallest size, just to be inside the valid range of sizes)_ Total formula in one go: ((40 -21)/2)+21 ((max-min)/half)+min
By the way, in the second-to-last test you can clearly see the entire structure jiggling horizontally… That may be messing with precision. Obviously, the MM3 would have a metal framework and be more rigid, but it’s still something to look into.
The marble machine is an instrument. When Cristofori invented the Piano, he focused primarily on its function. It wasn't till later that Steinway would go on to perfect the aesthetic and function of the piano. The Marble Machine X was your attempt to build a Steinway, when you should have focused on building the Cristofori Fortepiano first. Either way, you have me intrigued on where your research and development will take you next.
I just want to say thank you so much for letting us go on this journey with you, it's a learning experience for all of us, and it is impossible for anyone to look down on you or to think less of you for all of the work you're doing, this has been an incredible journey all the way from the first marble machine and I know when the third one comes, it will meet your expectations and exceed them, I can't wait to see you on the world tour.
Yep, I even stopped the emergency heart surgery I was in the middle of undertaking just to watch, the patient was the one who suggested it, they knew how important of a watch it was!
Really enjoying to watch your process! As a hardware designer for 30+ years, it’s fascinating to watch you teach yourself. I’m sure you’ve considered it, but a topic you haven’t mentioned much is transport. If you plan to take this on tour, minimal assembly/ disassembly will be critical, along with packing, and mitigating the impact of shocks/ movement on cantilevered components. I think you’re right to focus on accuracy and repeatability first, but your next embodiment has plenty of opportunity to consider the world tour as well!
Excellent job - just remember to make sure the change to the gate here didn't throw the timing off. (It *shouldn't* have, but it is possible the extended guidance may throw off the timing slightly.)
Nice! This science seems useful since small deviations are magnified when a marble bounces off a key. It also seems to show what variables could effect the precision of other mechanisms. It looks like bounces are agents of chaos.
Have you thought about the pressure from the waiting marbles on the marble gate? In a real case scenario, more than 7 marbles will wait (maybe about 30-50?) it may have an impact on the timing. I mean more waiting marbles means more pressure on the marble to drop, hence a shorter drop time.
The entire purpose of the clockwork escapement gate was to remove the back pressure of awaiting marbles and it was a major design hurdle that was addressed during the later design phases of the MMX. With this escapement gate we are seeing here, there effectively is no weight being placed on the marble about to be dropped because the clockwork arm is lifting all the other marbles back when the next marble is loaded.
the guiding rail actually wasn't a bad idea, but it did not adhere to an important design principle and you can see that the marble initially looses contact to the guide causing a bounce later. It's a bit like road construction. A roud doesn't just go straight and then immediately into a turn, there is a bit inbetween which is a clothoid (eg constant changing radius)
Martin, I'm just so happy to watch you again. No pressure to finish, but being along on the journey is so enjoyable. You've brought me joy for years, thank you
Eliminating as much error as possible before integration seems like the best approach. I really hope that the machine survives "Integration Hell". Achieving precision on individual systems is methodical but achievable work. Integrating many systems into one platform is orders of magnitude more challenging (as we've already seen in MMX) when you try to maintain the same precision on the final platform. Imagine all of the marble channels clicking and wiggling next to each other, that's a lot of vibration. Some level of error must be tolerated at that point, because even a 1000kg surface plate as the machine's foundation can't prevent every frequency transmitted throughout the machine's frame.
Martin, I know in the past that these videos take time and energy from other areas where it makes sense to apply that energy. I also understand you took a break from this. I'm so glad you're back. Thank you for these videos. I appreciate your attention to detail.
I know it can be hard to take pride in your own work, but both if the machines you’ve made so far are amazing!!! They definitely don’t suck! Keep up the excellent work, whatever you choose to do :)
I think at this point, the channel has turned from a music + engineering + art channel into just an engineering channel. Just makes me feel a bit sad that the original goal was lost in the pursuit of perfection, but if this is what you enjoy then so be it
This is a process. An artist or musician can only be as good as their tools. He is basically creating his own instrument and his own artistic medium. I think if any musician had to design his own instrument from scratch, a quest for perfection of expression would consume them as well. Rarely do you get to watch the process (even more rare that it is attempted), but I think it is fascinating. Imagine watching a great painter have to create his own paint. Watching the obsession of getting exactly what they invision would make for a deeper appreciation of the result.
@@GetaVe How many musicians are ok with instruments out of tune? How many painters paint on a surface that doesn't hold the paint where they want it? Artists want their chosen device to perfectly express what they want.
@@MorganKreig if you can’t work with what you’ve got then you’re not much of an artist. There comes a point when the pursuit of perfection has completely undermined the original intent of the artist. I’m not saying there’s no reason for a finely tuned instrument, in any craft, but when you start excusing lack of productivity based on the tools you’re working with, time and time again, then at that point you’re just wasting time and have lost your way.
I know last video a lot of people were saying that the precision you were going for was far too tight, but I do like seeing how tight everything's getting in the prototyping stage. The tighter it is now, the more wiggle room you'll have when it's all attached to a machine that has compounding tolerances and vibrations. I'm sure if/when MMmk3 gets built it'll sound great!
I only have one request. Always include some music (like you did here). I am constantly amazed at your different styles of improvisation and original ideas even in short pieces or ideas. Thank you for your art as well as your engineering! I tune in for both.
Like the saying "When you have a hammer, everything looks like a nail." When Wintergatan has a contact microphone..... Really smart use of your familiar tools to new problems.
I sorta understand your pursuit of precision, but it feels like you're trying to make the machine significantly more precise (in both time and space) than a person could ever be
@@C4pt41nN3m0 Synthesizers aren't programmed with physical parts. If the parts are all minutely imprecise individually, all of those imprecisions will stack up to become one giant standard deviation.
Promising test! I like the way you've reduced the design hurdles to specific issues, and you're tackling them one by one. I think you will end up with a viable and functional machine if you thoroughly design and iterate each problem component in this way.
I love these 10 min videos! I dropped out of your content flow during the streaming period, but now i am consistently following along again, thank you!
Are you sure that the modification you are doing (and will do in the future) to the marble gate will not affect previous progress? Like maybe the tilt of the gate could worsen the timing?
i love that you haven't wavered from your goal of perfection. you kept wanting to make a perfect machine. the question has gone from "how it can be perfect" to a much more important "if it can be perfect" and then working from there. I hope to one day see a machine that can not only sate the audience but satisfy your desire for a sufficiently optimized music machine.
I followed your videos for at least 3 to 4 years but I joined the military and spent most of my time with no internet or access to youtube. You said you already build 2 marble machines, does that mean Marble machine X was completed? And if so did you do the world tour like you planned? I know with the whole pandemic everything social slowed down but I'm curious if it happened or if it'll ever happen then. Still love your videos.
The Marble Machine X was never fully completed, and the world tour hasn't happened yet. Martin is currently in the designing and planning stages of his next marble machine project...
At the beginning I've watched because you were a funny marble machine enthusiast. Now I watch because you're a methodical and knowledgeable marble machine expert. I hope you keep it up :) Thanks!
as someone who's been following you since the first upload of the MMX, I have to say that your obsession with precision is becoming not only tedious to observe, but also seems to have become more of goal in itself rather than a feature. I mean, is this about making a cool machine that plays awesome music, or about inventing and optimizing processes and systems just for the fun of it?
I think (sadly) the marble machine has lost its soul as a beautiful piece of art that can play music, and will wind up as a marble robot controlled by MIDI. Hurray for an extra millisecond of precision, I guess?
The original machine made beautiful music despite loose timing and wobbly construction. The second machine was beautiful and was already producing beautiful music. It makes me sad to think about all the artists and craftsmen who contributed so much time and effort to producing parts for the MMX that seems to have just been abandoned now.
The thing is, I'm sure he's not having much fun achieving the insanely strict goals. But it needs to be done, as he chases after perfection, the only that will satisfy him and allow him to build the Marble Machine that he envisions, the one that reaches his almost unreachable standards of never failing and playing perfectly. Plenty of people doubt he'll ever be able to finish the machine, but I'm sure that in 50 years or so he'll finally finish the prototype :P
I started binge watching all your content and I wanted to say that I find your work quite impressive! I also want to add something that might be worth noticing for the third version of the machine. Part of what made the first marble machine so mesmerizing to watch was it's beautiful, intuitive yet complex architecture. The resealse mechanism for the marbles was made of a long stell cable that looks bended by hand, and it does large movements to drop the marbles. That mechanism was probably less precise than the one you are building right now, but I think that it was more intuitive for the viewers and also more beautiful. I feel like I'm nit picking here, but I think that the visual aspect of it is also worth considering.
I came to the video to watch some relief and got emotional by the end when you say "Good luck with everything you are doing". Man im needing some good luck these days, things are going pretty bad for me and your videos are a safe point where i can relax a bit and enjoy some good content.
I understand the need for perfection. I really do. But at this point I think people are going to lose interest. I was watching every single episode a few years ago. But now I just go straight to the comment section and see that my opinion is pretty much equal to everyone else. Love your work... But I think you are slowly, but perfectly, falling into a wormhole... Kind regards Martin
Ever heard of spin stabilization? As in rifled barrels. I wonder if that would help marbles fall more accurately. Also, for more predictable trajectories you need to insulate everything from air flow and material vibrations
Greater work ! To increase precision I think you have to ensure the stability of your backboard. It is not necessary I think since you have great results ! Bravo !
Yeah, in some shots it was quite obvious that the whole rig was shaking. I found myself yelling at my screen, but anyway what he has achieved is great already.
So happy to see these videos coming out again! I love how you’re seeking perfection because there’s so much to learn from the little details - as long as you’re having fun…
Martin, you are so concerned about precision, but what happens if you really go on tour with the MM3 and its raining or windy? Those factors would really destroy any precision you tried to achieve through your tests
Those other inaccuracies are the reason why Martin is going for a very precise marble gate. Every millisecond the gate doesnt take out of the inaccuracy budget is avalible for other parts of the system. A precise marble gate gives him more room with other parts like the programming wheel or external circumstaces like the weather.
Martin, I wish you would stop saying your first two marble machines sucked. They're not flawless, but they are amazing, beautiful machines that inspired tens of thousands if not millions of people. Keep up the great work!
@@AndersJackson I feel sad for you if you actually believe that, Martin clearly has lost track of what he is actually making. An instrument is never perfect, nor shal the Marble machine. I get huge problems, but as in the last video, 0.00001 Ms delay in marbles playing right, cant be heard by ear..
@@martinottesen7282 I am sad that you lost the belief in other people and in Martin. That is truly sad. What is also sad is that you post things like this, when you as easy just can unfollow Wintergatan and not be concerned about this any more. Probably better for you, so you don't need to be put down by what Martin do, that he doesn't follow your path. And also better for Martin and us that believe that he still can build a Marble Machine, if he do it systematic and from the ground up. And please don't be worried about me, I can cope. Be worried about you instead, where there actually are some one that care hearing about your negative posts about Martins choices. Because no one, except you, actually cares what you think. No, I don't, I just care that someone, that happen to be you, can't understand what it takes to make an instrument. And again, there is no way he could ever detect a difference in 1nS in his setup. And yes, that can't be detected by our ears. So, my suggestion? Come back each 6 month and have a look. You obviously don't have what it takes to follow development of an instrument and what it really take to design one that can be repeatable built and will survive performances without failure built into it (yes, of course there are always Murphy, but you don't need to build that into the design). Or, why not build the second Marble machine yourself, from what Martin has done so far? (And if you read it here. Yes, I am tired of gloomy followers that can't understand how engineering actually works. So sorry that you had to be the target, really.)
Thank you for keeping at it with attempting a properly working marble machine! I have truly enjoyed seeing all the process work over the years, and I have hopes for there one day being a fully functioning one. In the meantime, I love these videos so much!
Love that you're not either just giving up on the Marble Machine or charging ahead into it without any foresight. By taking some time to test you'll be better prepared to make the Marble Machine 3 and also you're still providing great insight into the musical and engineering sides of the project. I can't wait to see more!
I just love these. Your personal growth through the development process of these machines, lessons learned and how you come about deciding on how to move on is very interesting. So rare to see someone over time develop their skills over years like we've had the opportunity to do with your videos. Very interesting, useful and insightful all the way through. I hope some of the lessons you learn along the way rub off on me while I watch these videos :)
I'm sorry to say Martin, but should you not now confirm that timing is still good? Since you changed the gate?
Great Point! I got to check how every change affects the results from previous test. I filmed a timing test with the 11-degree gate rotation and the lower marble position but i edited it out of the video to stay on topic. Standard deviation was 0 ms still :)
@@Wintergatan that is great Martin from 3 years ago would have very confused with Martin from today 😁 it has been such an amazing journey to follow
@@Wintergatan I'd also suggest adding one more parameter: how fast you can drop marbles through your test setup. As speed goes up, so does the energy in the system. You should test for any effects marble throughput has on the performance of the gates and the resulting repeatability. Damping may be needed as the marble makes its last contact with the gate, and a softer material might help.
@@DLCasteror that could be done in the end. After all tests on the gate.
@@AndersJackson Well, if things are down to aiming a ball to drop in a specific area and within a tightly controlled time window and there is an expectation of high repeatability, I'd suggest doing that speed test early rather than finding out later that gate operation repetition rate mattered. The tests demonstrated in this video allowed a substantial time for the balls to stabilize in the queue whereas in the application there won't be several seconds between each release. Also, the queue depth will matter because there will be varying amounts of pressure on the system, which might require more of the channel to be modeled before arriving at an acceptable design. Even if one felt that gate design would not be affected by these dynamics, a feeling or belief is not engineering proof.
My only suggestion would be to check every so often that the adjustments you make for other accuracies don't mess up your timing accuracy
This is valid for every change,
Every change should trigger a test that it has not messed up the previous metrics.
In software engineering it's so easy, you just create automated tests and you're done for any regression.
In mechanical engineering it's soooo much worse and random and exhausting and this and that... I'm so glad to be a software engineer when I see Martin struggle with this huge list of constraints that are all super hard to put together 😂
Also known as regression testing
I was going to suggest the same thing..
Regression Testing.
The most surprising thing here was that Martin wasn't whining about it not being within a 2 nm tolerance. Hearing the words "good enough" was literal music to my ears.
He need to be designing each part to perform better then good enough, because he has to combine several parts, and all the imperfections will add or multiply up. The imperfection in timing will add up. Like the programming wheel and pick up etc. So have 0 delay and variation is good news. That will be needed later. And probabilities are multiplied. So if you have two parts that have probability of 0.9 to drop a marble, you have a system that have the probability of dropping a marble that is 0.9 * 0.9 = 0.81, which is way worse then 0.9. So you see, he HAS too design for as good as he can, to get a system that performs good enough.
@@AndersJackson I completely agree, but trying to make something absolutely perfect can be quite draining on the person. sometimes "good enough" is good enough :3
2 nanometers tolerance? Is Martin making semiconductor? If yes, he will need ASML's extreme ultraviolet (EUV) lithograph technology. 🤭
@@cronostvg IKR
@@cronostvg probably meant 2mm lol
"This is way more precision than I will ever need", it's great to hear Martin talking some bloody sense again!
That doesn't mean he's not going to attempt it. Again.
You say that, but he's gone about this backwards. He needs system requirements. Before doing any of these tests he needed to form his non-functional requirements. Why isn't 40mm good enough? Does he actually need more accuracy than that? He might, he might not, the point is that is is still arbitrarily chasing the highest precision he can, even if there is nothing to gain. He should have an exact number before going into the design.
No matter how many times he says the MMX could never work, the problem is that his is striving for a precision and accuracy that is completely unnecessary and arbitrary. If he wants to be precise that's fine, but he needs to define how precise that is. I think he has gone way too far. With the timing he went far beyond the limit of human perception. With this I'm assuming the main issue he is trying to solve is bounce reliability and therefore making sure marbles don't end up on the floor. But marbles falling on the floor was never a critical failure (assuming in small volumes) yet he made it out to be the end of the world. Not every failure is a critical failure, not every failure needs to be 100% removed.
finally hes satisfied with something lawl
@@EricTrimbur the amount of marbles lost in the MMX was way too high
and for this it's great to make sure every part can work better than they need to, as that means that any inaccuracy added later on is (effectively) already compensated for
@Wintergatan should really read this comment by @@EricTrimbur. I think we are seeing this issue because starting with general goals creates a lot of unknowns.
Martin should go back to music, analyse the music he wants to make, then create a machine that can play that music.
I was a fan of Wintergatan before this project. If we see more Wintergatan music, that is no loss.
Martin should: create music, analyse that music, create instruments that can play that music, experiment with the instrument(s). This is the old recipe, he's done this before.
Martin, at first I was worried when I heard you say, "when designing a marble gate, the goal is to drop the marbles as accurately as possible."
But then you explained that your tests show much more precision than you'll ever need. That's the most reassuring sentence I've heard during all these tests.
If you carry that through with the other challenges, and make sure things can be "good enough", I believe in the dream.
Just try to change that first sentence; when designing a marble gate, the goal is to drop the marbles accurately enough, and reliably enough. Now you just have to define "enough".
He's not reaching for perfection. As stated, he's trying to be as accurate _as possible._ By probing the limits of possibility he can be very confident in achieving realistic tolerances. It's a very reliable approach to designing.
It's a bit like stress testing a structural member. First you test the material to destruction to see how much load it can _possibly_ take. Then, when you want to build something to take a certain load with a certain safety factor you know exactly what dimensions to make it.
As good as possible does not imply an attempt for perfection. It is about finding the edge of the envelope… the beginning of diminishing returns…. If that is beyond the needed margin, the confidence improves.
Im so tired of people gatekeeping martin as if he is some mad scientist descending into the endless pursuit of perfectionism. No Martin knows better than any of us that when you settle for good enough these factors compound for years and create the brutal endgame of the 2nd machine. He is being strict with himself now to see if tis possible to complete this project. How people fail to see this is insane.
increasing the target aria will increase the life of the target. Just doubling the diameter of the target will increase the number of hits to fail by a factor of 4.
@@awogbob I agree completely, and think Martin is doing exactly the right thing for getting a functional, world tour ready Marble Machine, and he was right to abandon MMX for that purpose. However I do think Martin could've gotten MMX to a better endgame and gotten it to perform one song for all of the people who believed in him (and funded him!) all those years. Hell if he got the original Marble Machine to play a song, he definitely could've gotten the MMX to do it. I think that's the sting people feel when they bring up his "perfectionism". But I agree that Martin is taking the right approach here in ensuring it's even possible to address the compounding issues before designing number 3.
For future testing of drop accuracy, I strongly recommend letting the marbles land on a layer of carbon paper. Like your old-school typewriter, the marble will act like a hammer, transferring a spot of colour onto a blank sheet of paper underneath. Sheets of carbon paper used to be how you could make several copies of the same letter when typing on a mechanical typewriter. You can buy a sheet from an office supply store.
I was also think while watching this might not have being the best approch for what Martin goals are. These test is great for measureing accuracy but based on his discriptions and what he should be try to test at this part of design process is precision. To get everyon one the same page ," accuracy" is how close test are to a target while "presisions" is how close test are to other test. At this point the goal should be high level of precision from the dropper. accuracy can be figure out later though calbrations. At his point you just need to keep at the back of the mind you will need some method of being able to aline the drops at later date.
These test where measure as much differences between avg drop point and the center of the target as it was how consitent the test was. For test like this its just as important to track the types of fails as it is that there was failer. If it alwasy failed to same side this may be cause by bad calbrations rather then inconsitancy. I am not how much work went it to the calbrations of set natural it looks like this was edit out of video. The use of the phone camra different gave better data around precision then test did. I do think James Dominguez typewriter idea would give much better picture of precision of the dropper.
That is actually a good idea.
Just drop as many marbles as you can (with a rudimentary marble loop), and you can get 100-200 marbles through it and see in the color the deviation of the paths.
@@AndersJackson - I have a very vague memory of doing exactly this back in high school physics, launching glass marbles off a ramp and onto a sandwich of paper and carbon paper, giving us a visual record of the impact grouping.
@@jamesdominguez7685 Uh, you realize the green target spot he's showing is practically a non-antiquated, digitally scanned "carbon paper equivalent" contact/hit spot analysis..?
@@DatsWhatHeSaid - Did you see any visual data represented anywhere? I didn't.
Good work Martin!
But do not forget to backtrack your progress to previous steps! Check if your changes in the angle and position of the release mechanism changed the timing precision for example!
Keep up the great work!
Great suggestion!
Don't worry, the repeatability and reliability tests will almost certainly include all three of the the first roadblocks in one test
Also, I'm sure Martin has thought of this, but if he does build a third machine he's going to need to take into account just how level it is and any vibrations, because with the kind of precision he's aiming for even the slightest misalignment in how the machine is standing (or slightly moving) during use will throw things off.
@@corb805 still good to know as early as possible if current adjustments negatively impacted other requirements. Why would you continue otherwise?
He's going to immediately run into it. "Backstroke timing" is for rapid drops. And the gate he used here is double-action which will inherently have a time delay between drop 1 and drop 2.
And rapid drops will cause vibrations in the gate. His tests are with the gate securely fastened to a solid surface. It's going to be floating, if his prior designs are any indication of where he is going to go.
"That is all the precision I need" put such a smile on my face. Glad to see he's learning
Yes, he is learning, that is why he in the previous video targeted +/- 0 ms variation in the drops. Because he could with not too much work. When he have all data, he can adjust so the marble gate is good enough to have the marble machine to also perform good enough. That is why each part has to be performing better then the performance of the machine with all parts put together.
Yes Martin. This is very good science. I really liked the Shadow idea. Importantly you're also bring realistic about the smallest diameter, not arbitrarily going for the 21mm hole. Keep this stuff up, really good.
As long as the shadow was from a studio light and not the moving sun. 😄🤣
the slow-mo of the marble releasing form the gate directly over the cell phone camera looks AMAZING.
Happy Wintergatan Wednesday!
Sweetness Defined?
Let me finish the video first haha
WW is back on the menu, boys!
I foresee another merch design.
Indeed
Fact
Martin, you actually have SIX design roadblocks. Remember for the marble drops there are at least two other factors that the gate cannot control: 1) ambient air movement between gate and the chime bar which can affect the marble in uncontrolled flight; 2) the angle of the chime bar, needed to direct the bounce, means that there will be a fractional difference in the drop length/time depending on were the marble hits. Those two factors complicating perfectly tight timing suggests that the sixth design roadblock is perfectionism. It's not a medical device or a submarine or a rocket. It's ok to be good enough.
He is not aiming for perfection. He is aiming for a good enough system. But each part has to be as good as he can get, to be able to get the total system perform good enough. The final product is not this marble gate. It is the total system.
Each parts probability of not failing is multiplied: If you combine two part that have probability of working which is 0.9 (1 failure in 10 drops), you will get the probability of 0.9 * 0.9 = 0.81 probability of working (that is probability of 0.19 to fail).
Each parts timing variation is added: Have you two parts that have +/- 10 ms variation, then you get a total of +/- 20 ms.
The probability of failure have several parts in the marble loop: the marble lifter, marble delivery, marble gate, marble catcher, marble return to lifter.
The delay have other parts, like the programming drum, pick up, mechanism to control marble gate, marble gate, distance to instrument impact.
And the system performance is each of those combined. So have great performance on one part give him some slack on other parts, which probably doesn't perform so great. And even so, something that is great on one attribute on for instance the gate could give worse performance on one other attribute in the marble gate. When all this is done, he can decide what he can adjust to get good enough performance.
So yes, this isn't the road to perfection, this is the road to good enough.
Seven. You forgot the kinetic finger roadblock.
I guess Coriolis effect is quite null at these short distances
@@cerrudmanuel Manuel, you're correct. I was thinking more about wind across an outdoor stage. Even then the short distance might not matter much unless the target is small and the weather is windy.
@@SimonRosen- nej, och nej, jag räknade inte upp alla hinder.
Loving the fact that you dtopped aiming for perfection and shifted more towards acceptable margins approach
He has not changed anything from previous video.
He is still aiming for as good performance as he can get. In the previous video he could reach perfection with quite small work, so he did.
It will also give him some margin on other parts in the timing path where that isn't possible (programming wheel, pickups, mechanism between pick up and marble gate, marble gate and distance to instrument). Each variation in delay in that path will add up, ruin the goal of playing good enough.
Same will happen with probability of the machine perform without failure. There it is even worse, as the parts are in serie and those probabilities are multiplied. Probability is between 1.0, where it never fail, and 0.0 where it always fail. If two machine parts have probability to work 0.9 (which means it's probability to fail is 0.1, 1 in 10), put in serie and failure is when at least one fail, the probability to work is 0.9 * 0.9 = 0.81, which gives probability to fail being 0.19. So nearly combined probability of failing 0.2, instead of each parts 0.1. So he needs margins in each part.
And when he have done all tests on the marble gates, he can combine the different attributes, as some attributes could working against each other. But then he have some data to choose the final(?) design where he can put into MMXX, if he can design all parts to be good enough for the total system.
So yes, that is great to reach perfection in tests, but not critical. It is the best combination that will be the final design.
Hey, I'm probably not your target audience, and I'm not sure you'll ever see this, but I wanted to say that I love that you're still doing the marble machines. Every, maybe, 6 months or so, I'll check back in, and seeing new videos, all with marble machine based content just makes me so happy. So, thank you for your passion, and your persistence. You're really an inspiration.
Love that you’re back at it on your channel. I am thoroughly enjoying your journey. And I love the music Wintergatan has made. Keep up the awesome work.
You should re-check the timing in this new orientation for the marble gate, I think last time we determined that touching the marble like this degrades the timing accuracy.
Good Enough. You're about to go take a Six Sigma course, and you don't need it. The original machine was glorious, and so was the second! We love the art.
I've been following the MMX since its announcement and its kind of amazing how Martin has slowly but surely learned hard engineering lessons and taken them to heart. Just remember that all those fantastical mechanical instruments from centuries ago took many years to make as well. Can't wait to see your next set of engineering experiments. What an amazing journey!
Both of your marble machines are fantastic and I've loved watching you build and improve them, so don't sell yourself short!
Everybody else has good suggestions. I would also just say that the precision of your stacking of the hoops perfectly centered on each other becomes a concern as you get down to clearances of just a couple millimeters. I'd suggest putting some self-centering notches and protrusions on each ring so that they are guaranteed to be perfectly concentric as they stack on top of each other. And also, some sort of locked-in, unmovable, repeatable mount for the arms of the rings, so the center doesn't shift over time as you fiddle with them adding and removing mics and such.
it doesn't really matter because he can tune the position before doing the 30 marble drop test. and because he is filming from above he knows where the marbles are hitting and by how much, so if every marble hits at the same place, he knows that he fd up the setting of the ring and that it isn't a precision problem of the marble gate.
Martin, you are the real pog. Glad to see you back around. Great work out there
Glad you're back to regular posting!
Glad to have you back Martin!
I just want to say, I'm very happy to see you're uploading once again and I hope you're able to find it in you to take another shot at the marble machine. I completely understand your perfectionist attitude but I believe I can speak for many of your viewers that we're just marbles happy to be here for the ride.
Don't forget to consider the impact of vibrations - if you mount this on a marble machine and it doesn't have adequate stiffness, the vibration of the machine will impose extra lateral movement on the dropping marbles.
Martin the musical engineer. Let me give you some advice. Practice doesn’t make perfect, it makes improvement. Iteration is the same. Just as you had to practice to improve your skill at music, so to must you iterate to improve your skills at engineering.
And these videos are documenting a wealth of valuable information that future engineers will learn valuable lessons from. Please never give up. We believe in you, and believe that it is possible to build a marble machine that will surpass your expectations. It just takes patience and iteration.
or engineering musician...
I am SOO with you on this. Thanks.
And I am soo sad to see those that post negative posts. Because they obviously are not engineers that know that probabilities when combine in serie multiply (and deviation in time adds up). So design with margin will get you in the specs of what he need. Just aim for parts that are within the end specs WILL end up with a machine the are not within specs.
@@AndersJackson Thank you both! It's so frustrating to see all the negativity. It's like people are trying to find something to be mad about and aren't at all focusing on what is gained throughout the process, whether it's functional parts or better learning for the next attempt. Either way, it gets Martin closer to a working machine. Even if this one were to not work out again, I've enjoyed these videos SO much over the years!
Good to see that you are still enjoying tinkering and trying to make something beautiful!
You've become a problem solving machine. I can tell by the low amount of iteration you took to improve the precision, without surrender to over engineering. I'm pretty confident that your work from now on will be clean and lean. It's notable your personal development and I'm proud that you didn't gave up on this exhaustive journey of continuous improvement. :)
Trip Hammers! Martin I don't think hardly anyone would be disappointed if you simplified and went with trip hammers.
Martin, I am so greatful that you're still going forward with this project. You inspire me to be a better (software) engineer in my day to day. Your new process feels a lot like test-driven-development in software. This video definitely gave me unit testong vibes
This is interesting! I think the reality is that outside of the lab, you're going to be dealing with unavoidable forces like the tilt of the ground at a site you play at or the vibrations of the entire machine as things are moving (hell, maybe even the temperature and humidity lol), but it is a really good illustration of how even REALLY tiny tiny adjustments can make really noticeable differences. Of course, figuring these out now in the lab could end up being good for having the tolerances to deal with less than perfect conditions in real life.
Either way, it's good content and I hope you're finding this rewarding in its own way!
The tilt shouldn't be a big issue as he always planned to have the machine leveled wherever it is before playing. He has also been trending away from materials that are susceptiple to temp and humidity changes towards more inelastic ones.
@@michaelkrelwitz6203 For sure, but either way, I think this is an illustration of how very small (and often uncontrollable) factors can still have an impact. Which is why in a real-world situation, things aren't going to match this level of precision (which he understands, of course). But I am enjoying the process regardless, which is the point of all this, so I think this is time well spent. I don't think all this will eliminate all the roadblocks if and when he does make MMX3, but this is an interesting new approach that I'm glad is being documented for us.
I think if "REALLY tiny tiny adjustments can make really notifiable differences" then this is not a good design solution for the machine. I would say the goal should be to have a design so robust that it shouldn't rely on pretty fine tuning to make it wprk properly.
Seems to me he's trying to reduce randomness and learn what exactly goes into tuning each component.
When everything is put together all the randomness of every component will be summed up too. Tho wouldn't surprise me if the machine's vibrations or some other thing would invalidate some of the tests and make it perform worse.
I was worried about perfection over progress in previous episodes, but Martin has assuaged those fears today.
This really does seem like the kind of thoroughness and "from first principles" design that I would expect from someone aiming to reliably succeed.
Instead of building a machine, then endless failure mitigation, prevent the failures from coming up in the first place. The design will follow. Keep up the good work, Martin.
Thanks
I personally think the sway of your table is adding quite a bit of variance to the path the marble takes, which is most noticeable at 9:00 or so. This isn't huge, but any variance in the sway becomes multiplied by the distance it travels. Still, in a machine like the one you'll likely build, there will be a far higher sway and vibrations to kick the ball around. If you can achieve the 4 mm tolerance in a machine like that you should be handed a prize, It sounds pretty much impossible. Good shit my man!
He should test the accuracy while vibrating the table a moderate amount, should be close to realistic
@@michaelkrelwitz6203 he doesn't need to do that, as the construction of the holders etc will decide how much the machine will sway. There are ways of compensate for that too. How much that will add to the precision in landing area, which this actually are, can be calculated. So no need to add that variable to this test. He is testing one thing at the time, like a real engineer or scientist do when testing. Each variable removed when testing is one less complication in interpretation of the result to care about.
For fun? Yes he could. To test the performance of the gate alone, not so much.
the world needs the Marblemachine! Never give up
Aim for perfection, settle for excellence.
Thanks for letting us join you on your journey sir. You’re an inspiration to many, myself included :)
One of my favorite quotes about design is one that Destin (Smarter Every Day) said on his podcast: "This is acceptable because I have to move forward." We would love to iterate and tweak and change some things forever, but eventually we have to accept "good enough" because we *have to move forward.*
I like this approach better! I love how you can optimize a bit until it's robust and good enough, then acknowledge that it's "all the precision you need."
I'm glad to see you posting so much again! I still believe in the MMX!
I can't believe we didn't ask Wilson about his opinion on the results. How could we forget the most important member of the team?
You should write your suggestion down on a piece of paper, put in an envelope, stamp it and send it to someone who cares!
You should wrong your suggestion up, put it on an antelope, stomp on it and send it to someone whos a Care Bear.
@@YogiTheBearMan and @creageous , I do care!
So to follow your own teaching. Write down your suggestion on a paper, put it in an envelope, stamp it and send to someone who cares! 🙂
🙃😎
Not Wilson please
love the new approach!
I really like all the 3d printing you’ve been doing
I hope the final piece is made from steel. Plastic flexes too much.
@@billkeithchannel yes you are right. But it is much easier to prototype in plastic. 🙂
*@Wintergatan* 1:55 Look up "Binary search", it will help you find your answer faster.
Basically start in the middle, if the marble touches = increase the hole-size, otherwise decrease the size.
But you do not go +1 size, but you jump to the next that would be the "middle" size.
In this case (should have started at 30, but anyway), because 40 did not touch, jump to 30 (because 30 is close to halfway between 21 & 40)
Formula:
*40-21 = 19* _(the size difference between max & min)_
*19 / 2 = 9.5* _(divide that difference in half)_
*9.5 + 21 = 30.5* _(add that half-difference to the smallest size, just to be inside the valid range of sizes)_
Total formula in one go:
((40 -21)/2)+21
((max-min)/half)+min
Just wanted to say that i love yor work and content,keep it up and never give up,wish you all the best Martin.
Pain is temporary - Glory is forever.
By the way, in the second-to-last test you can clearly see the entire structure jiggling horizontally… That may be messing with precision. Obviously, the MM3 would have a metal framework and be more rigid, but it’s still something to look into.
I swear Wintergatan has the nicest community out of any youtube channel.
Thanks. I love that Martin ended with “see you next week”. Entertaining as always though.
The marble machine is an instrument. When Cristofori invented the Piano, he focused primarily on its function. It wasn't till later that Steinway would go on to perfect the aesthetic and function of the piano. The Marble Machine X was your attempt to build a Steinway, when you should have focused on building the Cristofori Fortepiano first. Either way, you have me intrigued on where your research and development will take you next.
Arent there 2 previous versions of the marble machine already?
So really, perfection is the only thing left.
I just want to say thank you so much for letting us go on this journey with you, it's a learning experience for all of us, and it is impossible for anyone to look down on you or to think less of you for all of the work you're doing, this has been an incredible journey all the way from the first marble machine and I know when the third one comes, it will meet your expectations and exceed them, I can't wait to see you on the world tour.
the only channel that can move me IMMEDIATELY to the video
Yes. I paused the video I was watching and went to this one as soon as the notification appeared.
Yep, I even stopped the emergency heart surgery I was in the middle of undertaking just to watch, the patient was the one who suggested it, they knew how important of a watch it was!
@@DaftFader Good to know your priorities are right!
was just about to reply the same :)
Really enjoying to watch your process! As a hardware designer for 30+ years, it’s fascinating to watch you teach yourself.
I’m sure you’ve considered it, but a topic you haven’t mentioned much is transport. If you plan to take this on tour, minimal assembly/ disassembly will be critical, along with packing, and mitigating the impact of shocks/ movement on cantilevered components.
I think you’re right to focus on accuracy and repeatability first, but your next embodiment has plenty of opportunity to consider the world tour as well!
Excellent job - just remember to make sure the change to the gate here didn't throw the timing off. (It *shouldn't* have, but it is possible the extended guidance may throw off the timing slightly.)
Nice! This science seems useful since small deviations are magnified when a marble bounces off a key. It also seems to show what variables could effect the precision of other mechanisms. It looks like bounces are agents of chaos.
Have you thought about the pressure from the waiting marbles on the marble gate? In a real case scenario, more than 7 marbles will wait (maybe about 30-50?) it may have an impact on the timing. I mean more waiting marbles means more pressure on the marble to drop, hence a shorter drop time.
The entire purpose of the clockwork escapement gate was to remove the back pressure of awaiting marbles and it was a major design hurdle that was addressed during the later design phases of the MMX. With this escapement gate we are seeing here, there effectively is no weight being placed on the marble about to be dropped because the clockwork arm is lifting all the other marbles back when the next marble is loaded.
It's always a pleasure to hear you play music in your videos!
the guiding rail actually wasn't a bad idea, but it did not adhere to an important design principle and you can see that the marble initially looses contact to the guide causing a bounce later. It's a bit like road construction. A roud doesn't just go straight and then immediately into a turn, there is a bit inbetween which is a clothoid (eg constant changing radius)
>clothoid
Or in layman's terms - a spiral easement.
@@douglascaskey7302 neither of those fucking terms are layman’s terms
Martin, I'm just so happy to watch you again. No pressure to finish, but being along on the journey is so enjoyable. You've brought me joy for years, thank you
10th video: can the marble travel in time?
"In general - no, but with seven contact microphones..."
Eliminating as much error as possible before integration seems like the best approach.
I really hope that the machine survives "Integration Hell". Achieving precision on individual systems is methodical but achievable work. Integrating many systems into one platform is orders of magnitude more challenging (as we've already seen in MMX) when you try to maintain the same precision on the final platform. Imagine all of the marble channels clicking and wiggling next to each other, that's a lot of vibration. Some level of error must be tolerated at that point, because even a 1000kg surface plate as the machine's foundation can't prevent every frequency transmitted throughout the machine's frame.
Since I discovered you on the internet I've always admired how passionate you are about what you do. Good job sir!
Martin, I know in the past that these videos take time and energy from other areas where it makes sense to apply that energy. I also understand you took a break from this.
I'm so glad you're back. Thank you for these videos. I appreciate your attention to detail.
I know it can be hard to take pride in your own work, but both if the machines you’ve made so far are amazing!!! They definitely don’t suck! Keep up the excellent work, whatever you choose to do :)
You’re so much more specific about the challenges, keep up the fantastic work! And thank you for sharing your process with us!
I think at this point, the channel has turned from a music + engineering + art channel into just an engineering channel. Just makes me feel a bit sad that the original goal was lost in the pursuit of perfection, but if this is what you enjoy then so be it
So true.
This is a process. An artist or musician can only be as good as their tools. He is basically creating his own instrument and his own artistic medium. I think if any musician had to design his own instrument from scratch, a quest for perfection of expression would consume them as well. Rarely do you get to watch the process (even more rare that it is attempted), but I think it is fascinating.
Imagine watching a great painter have to create his own paint. Watching the obsession of getting exactly what they invision would make for a deeper appreciation of the result.
@@MorganKreig an artist is good in spite of their tools not because of them.
@@GetaVe How many musicians are ok with instruments out of tune? How many painters paint on a surface that doesn't hold the paint where they want it?
Artists want their chosen device to perfectly express what they want.
@@MorganKreig if you can’t work with what you’ve got then you’re not much of an artist. There comes a point when the pursuit of perfection has completely undermined the original intent of the artist.
I’m not saying there’s no reason for a finely tuned instrument, in any craft, but when you start excusing lack of productivity based on the tools you’re working with, time and time again, then at that point you’re just wasting time and have lost your way.
I know last video a lot of people were saying that the precision you were going for was far too tight, but I do like seeing how tight everything's getting in the prototyping stage. The tighter it is now, the more wiggle room you'll have when it's all attached to a machine that has compounding tolerances and vibrations. I'm sure if/when MMmk3 gets built it'll sound great!
Time to watch the funny European inventor man play with balls again!!!
Awesome engineering channel with sweet sweet music. :)
I only have one request. Always include some music (like you did here). I am constantly amazed at your different styles of improvisation and original ideas even in short pieces or ideas. Thank you for your art as well as your engineering! I tune in for both.
You’ve officially stopped being a musician and have became a full time engineer
Martin: 1ms timing variation is too much.
Also Martin: 4mm position variation is way better than I'll ever need.
Like the saying "When you have a hammer, everything looks like a nail." When Wintergatan has a contact microphone..... Really smart use of your familiar tools to new problems.
I sorta understand your pursuit of precision, but it feels like you're trying to make the machine significantly more precise (in both time and space) than a person could ever be
Isn't that partly the point of a machine? Speedy and precise repetition.
Aiming for 99.99% effectiveness is a good goal to reach and he is on point to do that.
@@TheBlazingKnightsProduction He is already there. But he thinks he needs to go further.
@@C4pt41nN3m0 So long as he's having fun and pursuing the goal....
@@C4pt41nN3m0 Synthesizers aren't programmed with physical parts. If the parts are all minutely imprecise individually, all of those imprecisions will stack up to become one giant standard deviation.
Promising test! I like the way you've reduced the design hurdles to specific issues, and you're tackling them one by one. I think you will end up with a viable and functional machine if you thoroughly design and iterate each problem component in this way.
I like that all instrumentation is done with sound. Talk about using a hammer…
Actually, it is probably the easiest setup for Martin. So i do agree, I love that setup to.
I love these 10 min videos! I dropped out of your content flow during the streaming period, but now i am consistently following along again, thank you!
Are you sure that the modification you are doing (and will do in the future) to the marble gate will not affect previous progress? Like maybe the tilt of the gate could worsen the timing?
I have been searching for this comment, seems like a good point to me. Lots of testing coming up😅
During the end of the timing video didn't he find out that timing was pretty perfect all along, he just fucked up the way he counted timing?
i love that you haven't wavered from your goal of perfection. you kept wanting to make a perfect machine. the question has gone from "how it can be perfect" to a much more important "if it can be perfect" and then working from there. I hope to one day see a machine that can not only sate the audience but satisfy your desire for a sufficiently optimized music machine.
I followed your videos for at least 3 to 4 years but I joined the military and spent most of my time with no internet or access to youtube. You said you already build 2 marble machines, does that mean Marble machine X was completed? And if so did you do the world tour like you planned? I know with the whole pandemic everything social slowed down but I'm curious if it happened or if it'll ever happen then.
Still love your videos.
The Marble Machine X was never fully completed, and the world tour hasn't happened yet. Martin is currently in the designing and planning stages of his next marble machine project...
At the beginning I've watched because you were a funny marble machine enthusiast. Now I watch because you're a methodical and knowledgeable marble machine expert. I hope you keep it up :) Thanks!
as someone who's been following you since the first upload of the MMX, I have to say that your obsession with precision is becoming not only tedious to observe, but also seems to have become more of goal in itself rather than a feature. I mean, is this about making a cool machine that plays awesome music, or about inventing and optimizing processes and systems just for the fun of it?
yes, agree 100%.
I too grow weary! let's play some music! Isn't that the whole purpose of creating the machine?
I think (sadly) the marble machine has lost its soul as a beautiful piece of art that can play music, and will wind up as a marble robot controlled by MIDI. Hurray for an extra millisecond of precision, I guess?
The original machine made beautiful music despite loose timing and wobbly construction. The second machine was beautiful and was already producing beautiful music. It makes me sad to think about all the artists and craftsmen who contributed so much time and effort to producing parts for the MMX that seems to have just been abandoned now.
The thing is, I'm sure he's not having much fun achieving the insanely strict goals. But it needs to be done, as he chases after perfection, the only that will satisfy him and allow him to build the Marble Machine that he envisions, the one that reaches his almost unreachable standards of never failing and playing perfectly. Plenty of people doubt he'll ever be able to finish the machine, but I'm sure that in 50 years or so he'll finally finish the prototype :P
I started binge watching all your content and I wanted to say that I find your work quite impressive!
I also want to add something that might be worth noticing for the third version of the machine. Part of what made the first marble machine so mesmerizing to watch was it's beautiful, intuitive yet complex architecture. The resealse mechanism for the marbles was made of a long stell cable that looks bended by hand, and it does large movements to drop the marbles. That mechanism was probably less precise than the one you are building right now, but I think that it was more intuitive for the viewers and also more beautiful. I feel like I'm nit picking here, but I think that the visual aspect of it is also worth considering.
You're getting insanely good.
I came to the video to watch some relief and got emotional by the end when you say "Good luck with everything you are doing". Man im needing some good luck these days, things are going pretty bad for me and your videos are a safe point where i can relax a bit and enjoy some good content.
I understand the need for perfection. I really do. But at this point I think people are going to lose interest. I was watching every single episode a few years ago. But now I just go straight to the comment section and see that my opinion is pretty much equal to everyone else. Love your work... But I think you are slowly, but perfectly, falling into a wormhole... Kind regards Martin
Ever heard of spin stabilization? As in rifled barrels. I wonder if that would help marbles fall more accurately. Also, for more predictable trajectories you need to insulate everything from air flow and material vibrations
Greater work ! To increase precision I think you have to ensure the stability of your backboard. It is not necessary I think since you have great results !
Bravo !
Yeah, in some shots it was quite obvious that the whole rig was shaking. I found myself yelling at my screen, but anyway what he has achieved is great already.
So happy to see these videos coming out again! I love how you’re seeking perfection because there’s so much to learn from the little details - as long as you’re having fun…
I believe one efficient method you could try is having the marble rails spiral around the marble as it drops, like the rifling on a gun
Keep up the work I can’t wait to see the next machine
It's Wednesday my dudes.
Happy to see you back martin. I've been missing your voice and your art. Your an inspiration.
He's never going to finish this machine, isn't he?
So happy to be able to see your problem-solving ideas again. Great job Martin
Martin, you are so concerned about precision, but what happens if you really go on tour with the MM3 and its raining or windy? Those factors would really destroy any precision you tried to achieve through your tests
Those other inaccuracies are the reason why Martin is going for a very precise marble gate. Every millisecond the gate doesnt take out of the inaccuracy budget is avalible for other parts of the system. A precise marble gate gives him more room with other parts like the programming wheel or external circumstaces like the weather.
I noticed the marble gate flexes from the pressure of back marbles pushing in closeup footage minute 1
another useless hole Martin is diging himself into... Do you think a musician will hit his instruments with a 1mm accuracy?
Martin, I wish you would stop saying your first two marble machines sucked. They're not flawless, but they are amazing, beautiful machines that inspired tens of thousands if not millions of people. Keep up the great work!
I dont think the marble macbine will ever be finished at this rate.. if this is what you focus on...
he's on his third i think lmao
This time he do it the right way from beginning. So he is serious about the designing phase now.
It clearly isn't :D
@@AndersJackson I feel sad for you if you actually believe that, Martin clearly has lost track of what he is actually making. An instrument is never perfect, nor shal the Marble machine. I get huge problems, but as in the last video, 0.00001 Ms delay in marbles playing right, cant be heard by ear..
@@martinottesen7282 I am sad that you lost the belief in other people and in Martin. That is truly sad.
What is also sad is that you post things like this, when you as easy just can unfollow Wintergatan and not be concerned about this any more. Probably better for you, so you don't need to be put down by what Martin do, that he doesn't follow your path. And also better for Martin and us that believe that he still can build a Marble Machine, if he do it systematic and from the ground up.
And please don't be worried about me, I can cope. Be worried about you instead, where there actually are some one that care hearing about your negative posts about Martins choices. Because no one, except you, actually cares what you think. No, I don't, I just care that someone, that happen to be you, can't understand what it takes to make an instrument.
And again, there is no way he could ever detect a difference in 1nS in his setup. And yes, that can't be detected by our ears.
So, my suggestion? Come back each 6 month and have a look. You obviously don't have what it takes to follow development of an instrument and what it really take to design one that can be repeatable built and will survive performances without failure built into it (yes, of course there are always Murphy, but you don't need to build that into the design). Or, why not build the second Marble machine yourself, from what Martin has done so far?
(And if you read it here. Yes, I am tired of gloomy followers that can't understand how engineering actually works. So sorry that you had to be the target, really.)
Thank you for keeping at it with attempting a properly working marble machine! I have truly enjoyed seeing all the process work over the years, and I have hopes for there one day being a fully functioning one. In the meantime, I love these videos so much!
Love that you're not either just giving up on the Marble Machine or charging ahead into it without any foresight. By taking some time to test you'll be better prepared to make the Marble Machine 3 and also you're still providing great insight into the musical and engineering sides of the project. I can't wait to see more!
I just love these. Your personal growth through the development process of these machines, lessons learned and how you come about deciding on how to move on is very interesting. So rare to see someone over time develop their skills over years like we've had the opportunity to do with your videos. Very interesting, useful and insightful all the way through. I hope some of the lessons you learn along the way rub off on me while I watch these videos :)