Got your book and have been going through it. Makes a really great learning-tool as well as a checklist to 'sanity-check' your layout and design. Thanks for doing the 2-layer; it is now obvious how thin multi-layer boards can really push up the frequency. Great video & 73...
Those are some great design rules and it's great to see why we should use them. Interesting to realize that the distance of traces is incredibly important.
I have enjoyed these videos so far, thankyou! A tentative suggestion: I'd love to see you do some videos later about grounding in mixed-signal type systems: there seems to be a real shortage of resources/tutorials/worked examples in this area that are suitable for hobbyists/amateurs like myself.
Thanks a lot! For mixed signal the exact same rules apply. Just remember that when working with digital signals, the required bandwidth is at least 9x the maximum clock frequency, you need up to the ninth harmonic!
Hi Hans, Hey, I thought I was subscribed to your channel... I must have had a signal crosstalk in my brain... LOL. Anyway, I am subscribed now! I like that you're doing these videos and research too and then the best thing of all is that you are willing to share it with us. Thanks so much! Cheers.
Thanks a lot. I'm currently developing a paid course based on this document. I'm getting really excited, there is going to be so much more valuable details in there then I thought before I started developing it. Maybe it is something you can use (if your boss is willing to pay for it :-) ). Best regards, Hans
Thank you for the book and all info on PCB. I have used a excel sheet to remind me of everything that needs to be checked post layout but not pre layout. Great video.
I have a couple of questions: 1. What are your thoughts on curved traces, like bezier splines, and 2. I didn't see vias along the edges of the ground planes discussed. Is this important?
Curved traces work perfectly. Bezier splines (I had to look that up ;-) ) would also work fine. I'm covering vias on the edges of ground planes in my 4th video in great detail. And yes, it is important if you do RF ;-) Best regards, Hans
Your stuff is very good, but what got you my sub is how you treat people on comments. Thank you for everything you been doing. Best regards from Brazil.
There's a real derthe of ground stitching in the PCBs (in the first ~5 minutes). I suspect that may have some factor in the bump you see at 300 MHz on 2 layers. Most CAD programs have scripts to place vias at some interval (if there's no other signal there) which can be a good start. Usually I manually place vias at all prominences, and sometimes restrict ground flood at particular points if I can't place a via - unstitched (and sometimes even singly stitched) ground fingers can act as radiators. To paraphrase Devo: when a problem comes along, you must stitch it.... Stitch it good. Depending on your spacing, the coplanar ground is changing the circuit to a coplanar waveguide with ground (CPWG) instead of microstrip - impedances should be adjusted accordingly. For short circuits, the loss is a bit higher, but if your circuit is long enough (or has enough inductance) to radiate, CPWG will have lower radiation losses.
Wonderful job explaining what is going on as well as the practical test examples. I am really enjoying the series. I'm self taught in electronics and feel like I know so little. I usually just take simple circuit and combine them into the projects I need, usually micro-controller based. Please Send me the checklist.
Again great work. You asked for topics for future videos. I would like to see strategies for multi layer board routing analog and digital signals and power. Thank you.
Suggestion, you could design good and bad layouts for a PCB design for a simple switchmode DC buck regulator. The measure of success would be ripple. Id also like to see an explanation of how to design audio boards with single point grounding. Great series, thanks!
Thanks for the ideas. I was thinking about a dcdc regulator for a video already, but great that you show that there is a real interest there. About single point grounding: I'll make a video on that, its on my list for one of my next video's. I'm an audio design engineer myself for over 30 years, I know single point or star grounding is considered good practice but it really reduces audio quality. Planes are much better because they have a low impedance at high frequencies. Even though audio is low frequency, it will pick up high frequency interference that mixes down due to non linearity and get's into the audio band. I'll go deeper into it in the video which comes out in a few weeks. I have a vacation for a few weeks to come.
Linear Technology/Analog Devices has great application notes on switch mode converter PCB layout, AN136 and AN139 are well worth the read. I would say the measure of success is radiated and conducted EMI. Ripple is mainly defined by inductor size, switching frequency, switching mode (CCM, DCM, burst) and ESR of the output caps. I can tell from experience that laying out a DC/DC without any rules like it's just a 555 blinker can turn the circuit into a very powerful radio transmitter ☺
This is fantastic information. I've only designed and made a few PCB designs but each gets better thanks for excellent advice from true professionals like yourself. Keep up the good work!
Wow, you only posted three videos and already got almost three hundred subscribers (including me)! That's insane! Although there's almost nothing new for me, but I really appreciate your work for helping electronic engineers. It seems that many designs around us are full of dramatic flaws and situation needs to be changed.
Thanks Hans for another great video, One topic that would interest me a lot is the effect of the dielectric material used in the PCB. Let's say you compare the cheapest JLCPCB 2 layer PCB against a not-so-cheap EU 2 Layer PCB with the same basic parameters (i.e. PCB thickness, copper thickness). The only difference being the dielectric material of the PCB. I have heard of the loss factor, dielectric absorption etc. But what effect does the overall quality of the dielectric and weave pattern etc. have on high frequency applications? Is there any difference that can be measured? I really like this video series!
I may have answered that one already in video 1. Somewhere in that video I calculate the loss of the 10cm transmission line and compare it with the real measurements. They're spot on: AND I MEAN SPOT ON!! :-) Within 0.02dB, which is nothing. This was a board by PCBWAY. So I think those material properties are quite easy to control with modern equipment. If you really want high performance RF, you need to use rogers board material, however, a board is very very expensive when you use that.
In machining apart some junk 4 layer boards on hand I have found that they seem to have the layers evenly spaced rather than the way you show in the pictures. Do you have to specify the way the layers are spaced when buying the boards? I have never seen this mentioned.
Every supplier has a different standard stackup. Eurocircuits uses 0.38 / 0.8 / 0.38 for their FR4 cores, pcbway does 0.19 / 1.2 / 0.19. However, all kinds of thicknesses are possible. There are also 75um dielectrics. If you deviate from the standard a supplier has, it does tend to get really expensive really fast :-)
Usually, the PCB stack-up for a 4+ layer board is made up of "core" and "prepreg" layers which are laminated together in a press. If you have a 6 layer board, then there will be copper foil layers added also. Generally, the core material is quite thick compared to the prepreg material. Not saying it wouldn't ever happen, but I would be surprised to see a 4 layer board that truly had evenly spaced layers, i.e. layer 2 the same distance from both layer 1 (top) and layer 3.
that would work, it would still have some inductance though. The ground plane is still quite far away, so I expect adding multiple vias will still help in that case
@@HansRosenberg74 Thanks. I was opperating under a false assumption that I really should have known. Size of the conductor has some influence on inductance,,, multiple inductors in parallel are more efffective. I get it now. If I had really payed attention in high school I likely would have had this engraved in my brain,,, but ,,,,. That was 50 years ago.
Very good series of videos. Looking forward what will come in the future. one question, maybe a silly one. Do the connections between the filter components not also act as a transmission line and that has to be taken into account?
Just read your checklist. Good read and helpful. You ask in your book what are we struggling with, so here goes. I am new to all of this, I had my first board made and it was a success. It was a simple LND150 Mosfet gain stage for use in a tube amp. I want to do more and I don’t understand layering and trace sizes. I do not understand vias and how to size them. I am used to building my board as turret boards or eyelet board. I need to understand how to select the trace size. For high current I use 18awg-20awg wire and for my small signals I use 22awg wire. How does this equate to trace size? I will still need large wire connections is this just a large via? I am really enjoying your videos. Thank you. Chris.
Hi Chris, thanks a lot for responding to my question for ideas people are stuggeling with! These are some really good questions that I wouldn't have come up with because I've been doing this for soo long :-) I'll need to think about these to come up with a good answer. I want to come up with a simple set of rules to pick the right sizes. About layering: Standard is 1.6mm thick pcb material with copper on both sides. This is what every supplier produces the most. You can make almost anything with this unless you start messing around with microcontrollers, ram, fine-pitch components etc. At pcbway, you can get a 10x10cm board for a little as 5 euros (you get 5 boards) plus shipping so using turret or eyelet boards is not economical anymore if you know how to design your own boards. Best regards, Hans
Great videos, thank you for sharing your experience! Here is my idea for a test: I would like to know whether a GND lane between the two signal lanes will improve the situation, maybe in comparison to the actual shield.
Some ideas for improving 2-layer board: 1mm thick board; lower via aspect ratio and multiple vias; via walls and multiple randomly spaced stitching vias between top and bottom grounds.
Nice, you just named the whole list I wrote down this morning for the next pcb. I want to see what I can do with 1.6mm since that is mechanically stronger.
Link is in the description. I'm currently developing a paid course based on this document. I'm getting really excited, there is going to be so much more valuable details in there then I thought before I started developing it (I discover there are a lot of things I do without thinking about it, now that I'm paying attention to all the things I think about it's much more than I thought. I'm probably confusing now :-) ). Maybe it is something you can use (if your boss is willing to pay for it :-) ). Best regards, Hans
Your videos are top gold, I hope for a lot more to come! How important are these design principles, when dealing with digital circuits of relatively low frequencies? Like ESP boards and their usual peripherals like I2C, SPI or USB? So far I only did 2-layer boards to save on cost and never bothered about the ground plane and messed it up by routing signals lines through it and thus cutting it up in smaller connected areas. I guess in a multi Ghz board, this would have disastrous consequences, but so far my boards do what they supposed to do.
I'll make a video on this maybe. Remember, a 10MHz clock needs a 90Mhz bandwith, you will need the 9th harmonic to make it look a bit like a square. This means you're already 'jamming' FM radio without a groundplane :-)
@@HansRosenberg74 I'm going to test this. I should be able to pick this up using the HackRF and if not, it shouldn't be too bad. Most of my designs are battery powered and only on for a few seconds per day. The other is encased in a closed, grounded box. Nevertheless, I would appreciate a video about EMF and how to check designs for that during design phase and after production with tools an amateur can afford.
Would be great to see a 4 layer tutorial. A couple of different power lines and a few different signal wires such as spi and i2c. What's the best way. I'll be waiting 😊
One Suggestion I would have for improving this kind of two layer PCB would be to make a slot in the PCB and soldering the shield to both sides of the PCB, and therefore eliminating any chance of field passing by.
haha, that would be really really drastic and effective! I'm thinking about a via-wall now. that should help a lot as well. It's also a little less rigorous :-)
@@HansRosenberg74 Via wall was the first thing that came to mind. Curious how via stitching would effect the filter circuit on both the two layer and four layer boards.
Hello, I don't really have long term experience in PCB design or electronics as I've been in embedded software development for most of my career. However, I am now retired and doing some electronic projects, including PCB design, mostly as a hobby. I also watched many videos from Robert Ferarec, I suppose you know him. I downloaded your Checklist Guide which it's great overall, but I found something I am not fully comfortable with based on what I learned previously. It's about your suggested 4-layer stackup. Some designers suggest (1) placing continuous ground planes on both Inner layers and signal/power traces only on top and bottom layer, (2) trying to minimise signal vias and (3) always placing at least one sticking GND via next to any signal transfer between the outer layers. With this stackup you make sure that one ground plane is always very near all the signal traces, which minimises crosstalk and maintains constant trace impedance regardless of trace layout. It also prevents all disruptions or cuts or signal return paths. On the other hand, your suggested layer assignment does not seem to guarantee it all, unless you are very careful with the actual layout, which is more prone to mistakes or omissions (?). What's your view on that?
Hi, what I generally do is put the signal and power traces on inner2 and put as much ground as I can on layer 4. Often this means layer 4 is a near perfect ground plane. So basically, compared to what your are saying, I'm swapping inner2 and the bottom layer. Electrically, the ground will have the same distance to the signal layer. However, by putting the signal/power traces on inner2, they are now completely shielded on both sides by ground, minimizing possible interfering transmissions. You are also offering return paths on both sides of the signals on inner 2, which is another improvement, though a minor one because of the distance to the plane on inner1. I think from an RF viewpoint, this is a better solution. A practical disadvantage could be that you cannot reach those traces for reworking or debugging. Adding ground vias next to signal crossings is a great idea, it guarantees the shortest path for the return current. Let me know if this clears things up. (I revised my first answer, maybe you already got this in the mail, I'm not sure, I answered to quickly but when I thought about this longer I found this was the better reply) Best regards, Hans
@@HansRosenberg74 Thanks, yes this makes sense to me now. As you pointed out it's basically like swapping my 3rd and 4th layers, with the added advantage of ground shielding on both sides of 3rd layer. Great! I have now another question: According to your suggested layout, in which you use inner2 to signal/power, does it make sense to fill that layer with copper attached to power?. I mean, a mostly continuous plane of power on that layer, instead of actual traces for power. I think this will create, to some extent, a continuous capacitance between power and the adjacent ground layer which should improve power distribution (?) Does this make sense, or may it introduce other problems? Thanks again
Sounds like an excellent idea. Especially if you make sure you have decoupling caps at regular distance (which you probably already have on a lot of locations) the power plane will also perfectly function as a ground return current path. You could even place np0 / cog caps near the start and end of a signal trace between gnd and power to offer the return current a very short path along the power plane. But this would only be necessary for extreme cases where you have a very high speed line and you really can't afford a ground plane because you need to conduct a large current in the power plane so traces really won't work there.
Very interesting videos. But in practice 4 layer boards are way more expensive for a small scale project. Do you think working on a 0.8mm 2 layer board is a good compromise?
Yes, it is, but I always worry about 0.8mm boards with smd components. They bend easily so I worry about cracking my smd components. Bye the way, you have a 5 pieces 10x10cm 4 layer pcbs for around 50 USD these days including shipping. So it's not that bad!
Another amazing video! These rare nuggets of 'integrety' or whatever you want to call it are what are still showing how good the internet can be despite it being drowned in an almost endless ocean of shit. On the topic of PCB-design or maybe electricity in general, I've heard it being said that 'the energy is in the fields/the dielectric etc etc'. Which kind of sounds profound but I can't help but feel it kinda sounds a bit like saying 'the energy in a waterpump and pipe system is in the pipes' because there is only pressure and direction of flow because the pipes prevent the water from going everywhere. I was wondering if you had any thoughts on this? I have little to no experience in RF electronics so maybe im just talking nonsense, but yeah.
That is a big compliment, thanks! I haven't thought of the 'water equivalent' for RF yet. I will be making another video that explains RF a little bit further again, hopefully I can post that next week.
Thanks for the idea's, I wanted to try rogers, but man: That stuff is really expensive. Once my channel is monetized maybe I can invest a bit more. I did use rogers in the past when I was evaluating a 4-5GHz radar chip, it works wonders!
Hi Hans, some top ground planes are already connected to the bottom groundplanes via the 'vias' next to the connector or am I mistaken? Great videos and thanks for sharing. Best regards, Ron
I have no idea, I only know the NanoVNA (and the very expensive 'real' stuff from hp/agilent/keysight... I wonder when they're going to change their name again :-D )
Nice videos !! Please send me the checklist! I have some coupling troubles with my homemade domotic: when I switch some lamps on or off, it also switches another one that has a secondary switch in the same area 😅 and it quite bothers my wife 😅😅
I'm pretty sure good grounding will fix this, sounds like a typical grounding problem. Also make sure you have good supply decoupling! You can get it here: www.hans-rosenberg.com/checklist Best regards, Hans
Very usefull information. I would also prefer some practical exemple. Take a design and design it with 2 layer and 4 layer. Anyway Interesting theoretical staff. How about the book, some exemple înșine?
Please check part 1 and 2, that covers differences between 2 and 4 layer designs with practical examples. The book does not have these examples, that would be something for if I turn it into a real book :-) You can get it here: www.hans-rosenberg.com/checklist Best regards, Hans
Hey Very informative, and well explained. If the book is half as detailed, then I'm going to love it! Please send me a copy, let me know how to get in touch.
Thanks a lot! This book is about the steps in a full pcb design process and common errors that can be avoided. You can get it here: www.hans-rosenberg.com/checklist Best regards, Hans
This is super interesting! It has so much of the opposite problems I can face when dealing with transimpedance amplifiers. For me currents coupled capacitively make problems and having more distance between planes can help. Frequencies are a lot lower though. Any thoughts on this? Digital signals coupling to analog amplifier inputs where slight currents become amplified.
Digital 'aggressors' attacking analog 'victims' is very common, since digital signals have a very high amplitude and lots of harmonics. If you really have a crosstalk problem, the best way to deal with that is expensive: Shield cans across your analog or digital circuitry. The absolute minimum you need is a very good ground plane, without that, nothing will work.
@@HansRosenberg74 Optical sensor -> transimpedance amplifier 22bit ADC gives 5pA sensitivity😝 Leave it on a dark table facing down and it still picks up vibrations from you walking on a concrete floor... Needless to say, any signals ON the same board can be picked up too. Things are differential to help, but I am sure more can be done on the layout side. Reading out the data from FIFO is also a worry as reading fast creates higher speed flanks/frequencies but allows more sleep between reads for battery life. Slow gives less noise from readout but consumes battery; and I am sure I can spot noise from the system "running" in the data as well (albeit faint). I think this is a "what is good enough" situation as you can always find things with high sensitivity, ultra low noise sensors.
Nice series - even if I'm not really into RF (except for an occasional NFC antenna) I still enjoyed it. But I'm more into doing stupid things with a handful, or two handfuls, of 74-series logic ICs. And being a bit on the cheap side I usually go for two-layer PCBs just like they did in the 1980s. Do you have any tips for power distribution (especially when using faster families like 74AC). And also does crosstalk matter in a digital bus, like if you do the but at 5/5mil or space the bus out to say 8/16mil?
Crosstalk is not really a problem for a hobby project. However, if you have a real product, and you don't have a ground plane and you 'forget' decoupling caps close to the power pins of your IC and don't put a ground via very close to the other side of the decoupling cap: You've build an excellent interference transmitter :-) This will get you into trouble with certification tests.
Haven't watched all of your videos, so I'm sorry if you've tested this already: What would the result of having a ground pour on both outer layers of the board (surrounding the signal traces), as opposed to having the ground plane be on the opposite side of the board, look like? You could essentially make two ground planes with some via stitching. Just curious since I didn't see you mention pours in this video. Thanks!
What would happen if you had a slot in the board in between the two connections and attached the shield directly to the bottom ground plane on the two-layer board? My apologies if this has already been suggested or thought of. One other thing with separating the ground plains, I know you're against that, but wouldn't the ground be dirty on the opposing side, and would it make sense to double the shield with an insulator in between and split the ground just at that point?
@@HansRosenberg74 yes I did that's how I know it you are against it. Well you don't have to have a split ground to have a small slot where the shield can be soldered to the bottom plane. Your thoughts on that.
@@HansRosenberg74 I wonder what would be cheaper vias + shield or a slot + shield. You're going to use a shield anyway then a slot would hold it real nice.
An Idea is to add a premade shielding can over the problematic parts of the board and connect it to the surface gnd. Parts that are influenced by each other woudl then be separated by a "grounded wall". singal traces are then routed on the other side to pass under the shielding can for an uninterrupted conenction of the can to gnd.
That would definitely work, however, 4 shielding cans on a filter would get rather expensive 🙂 In my next testboard I'm going to try some other measures to deal with it and improve it. Curious to see how this is going to work, I still have to make that board.
@@HansRosenberg74 What about using a nonconductive epoxy encapsulant like Stycast? In a 2-layer board where the shortest path to GND isn't necessarily between the signal trace and plane, would exchanging air for a higher Dk material help? Not saying this is practical from a cost of manufacturing standpoint, of course.
Interesting! What if you milled a slot for the shield plate, and soldered it continously straight to the ground plane? That should close off the x-axis leakage path completely (and also probably be way more expensive to manufacture, just a thought experiment). Are the peaks in the 4 layer response with shield perhaps somehow related to the small leakage path that is still under the shield plate? (I guess some minor signal can still go the long way around the ends of the plate, unless you go full monty and completely faraday cage the connections.) PS. I'm a complete RF noob, still mostly black magic to me.
Milling a slot and putting in a shield will definately solve it. Those peaks are very common in RF. Stuff very easily starts resonating, only very clean controlled structures will not do that. This is not very 'sophisticated' so it will show this kind of behavior. Best regards, Hans
Thicker copper traces/layers (for instance 70um instead of 35um) does not help due to the skin effect. It will help for DC (less resistance), but not for RF. Best regards, Hans
Got your book and have been going through it.
Makes a really great learning-tool as well as a checklist to 'sanity-check' your layout and design.
Thanks for doing the 2-layer; it is now obvious how thin multi-layer boards can really push up the frequency.
Great video & 73...
Thanks!
Those are some great design rules and it's great to see why we should use them. Interesting to realize that the distance of traces is incredibly important.
Thanks!
Great job, thank you!
Request: effectiveness of guard rings.
Thanks. I'll look into that.
I have enjoyed these videos so far, thankyou! A tentative suggestion: I'd love to see you do some videos later about grounding in mixed-signal type systems: there seems to be a real shortage of resources/tutorials/worked examples in this area that are suitable for hobbyists/amateurs like myself.
Thanks a lot! For mixed signal the exact same rules apply. Just remember that when working with digital signals, the required bandwidth is at least 9x the maximum clock frequency, you need up to the ninth harmonic!
Hi Hans, Hey, I thought I was subscribed to your channel... I must have had a signal crosstalk in my brain... LOL. Anyway, I am subscribed now! I like that you're doing these videos and research too and then the best thing of all is that you are willing to share it with us. Thanks so much! Cheers.
you're welcome
Send me the checklist please. Great videos with really informative demonstrations, thanks for your great work
You're welcome. Link is in the description. Best Regards, Hans
VERY informative, educational and well presented. Thanks for that and the free checklist as well!
Thanks a lot. I'm currently developing a paid course based on this document. I'm getting really excited, there is going to be so much more valuable details in there then I thought before I started developing it. Maybe it is something you can use (if your boss is willing to pay for it :-) ). Best regards, Hans
Nice to see a growing channel!
Watching this video while I took a break from writing code for my own SPICE software. Already like your content!
Thanks a lot!
I am going to watch your videos in first few hours of posting :)
Thanks a lot :-)
Thank you for the book and all info on PCB. I have used a excel sheet to remind me of everything that needs to be checked post layout but not pre layout. Great video.
You're welcome, I used to have an excell sheet as well, until I decided to start a RUclips channel :-D
I have a couple of questions: 1. What are your thoughts on curved traces, like bezier splines, and 2. I didn't see vias along the edges of the ground planes discussed. Is this important?
Curved traces work perfectly. Bezier splines (I had to look that up ;-) ) would also work fine. I'm covering vias on the edges of ground planes in my 4th video in great detail. And yes, it is important if you do RF ;-) Best regards, Hans
Love your videos so far! I hope they keep coming - you got me subscribed.
Thanks!
Your stuff is very good, but what got you my sub is how you treat people on comments. Thank you for everything you been doing. Best regards from Brazil.
Thanks a lot!
That is a great compliment :-)
Very interesting content, very well presented, concise and up to the point. Keep up the good work!
Thanks, I plan to do this for 1 year and see what it brings.
Very informative video , keep it up...
Thanks for the encouragement! I plan to do this for 1 year, and then see if this is a viable way to make a living. Living on the edge, lol :-)
Thank you so much for your amazing content! I am learning a lot from these videos.
Thanks a lot. You should get my checklist, link is in the description. It's free and really has great info.
Thank you so much, for grate RF video series!
You're welcome!
There's a real derthe of ground stitching in the PCBs (in the first ~5 minutes). I suspect that may have some factor in the bump you see at 300 MHz on 2 layers.
Most CAD programs have scripts to place vias at some interval (if there's no other signal there) which can be a good start. Usually I manually place vias at all prominences, and sometimes restrict ground flood at particular points if I can't place a via - unstitched (and sometimes even singly stitched) ground fingers can act as radiators.
To paraphrase Devo: when a problem comes along, you must stitch it.... Stitch it good.
Depending on your spacing, the coplanar ground is changing the circuit to a coplanar waveguide with ground (CPWG) instead of microstrip - impedances should be adjusted accordingly. For short circuits, the loss is a bit higher, but if your circuit is long enough (or has enough inductance) to radiate, CPWG will have lower radiation losses.
I have a whole video on the subject, my part 4 video. Link in the description :-)
Wonderful job explaining what is going on as well as the practical test examples. I am really enjoying the series. I'm self taught in electronics and feel like I know so little. I usually just take simple circuit and combine them into the projects I need, usually micro-controller based. Please Send me the checklist.
thanks,
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Again great work. You asked for topics for future videos. I would like to see strategies for multi layer board routing analog and digital signals and power. Thank you.
Thanks. I'll look into that. My design checklist does contain a standard layer assignment for 2 and 4 layers I believe, that could be a start!
Good job on the testing and presentation!!
Thanks!
I like this.
Thanks!
You're welcome!
Suggestion, you could design good and bad layouts for a PCB design for a simple switchmode DC buck regulator. The measure of success would be ripple.
Id also like to see an explanation of how to design audio boards with single point grounding. Great series, thanks!
Thanks for the ideas. I was thinking about a dcdc regulator for a video already, but great that you show that there is a real interest there. About single point grounding: I'll make a video on that, its on my list for one of my next video's. I'm an audio design engineer myself for over 30 years, I know single point or star grounding is considered good practice but it really reduces audio quality. Planes are much better because they have a low impedance at high frequencies. Even though audio is low frequency, it will pick up high frequency interference that mixes down due to non linearity and get's into the audio band. I'll go deeper into it in the video which comes out in a few weeks. I have a vacation for a few weeks to come.
Linear Technology/Analog Devices has great application notes on switch mode converter PCB layout, AN136 and AN139 are well worth the read. I would say the measure of success is radiated and conducted EMI. Ripple is mainly defined by inductor size, switching frequency, switching mode (CCM, DCM, burst) and ESR of the output caps. I can tell from experience that laying out a DC/DC without any rules like it's just a 555 blinker can turn the circuit into a very powerful radio transmitter ☺
This is fantastic information.
I've only designed and made a few PCB designs but each gets better thanks for excellent advice from true professionals like yourself.
Keep up the good work!
Thanks!
@@HansRosenberg74 all my PCBs so far have been two layer. It's time to go four layer!
Loved the video series! Please send me the checklist. 😊
thanks, Please really helps :-) Link is in the description, best regards, Hans
Keep up the good work!
Thanks a lot!
Excellent, thank you!
You're very welcome!
Wow, you only posted three videos and already got almost three hundred subscribers (including me)! That's insane! Although there's almost nothing new for me, but I really appreciate your work for helping electronic engineers. It seems that many designs around us are full of dramatic flaws and situation needs to be changed.
Thanks, i've seen may of these simple mistakes made in boards, that's why I'm posting this.
Great job Hans! You rock explaining!
Wow thanks :-)
Very informative videos especially showing the testing on actual PCBs! Please send me the checklist.
Thanks
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Thank you for your great videos :D
Thanks!
Nice video,looking forward to more of these
Working on it 🙂
Hello. I downloaded the checklist from your website. Thanks!
you're welcome
Thank you for your videos, we can feel your experience and passion for electronics. Very usefull contents and very well explained too! Thank you
Thank you very much!
Again. Very nice video.
Hi, the link is in the description. Sorry for the late reply, I was on the road for 2 days. Best regard, Hans
Yes, please, send me the checklist. Great videos! Thanks!
You're welcome. Link is in the description. Best regards, Hans
Excellent presentation. Easy to follow and understand. I'm just getting into PCB design so a copy of your checklist would be appreciated, thanks!
Thank you. The link is in the description. Best Regards, Hans
Thanks for this really good content
Glad you enjoy it!
Thanks Hans for another great video, One topic that would interest me a lot is the effect of the dielectric material used in the PCB. Let's say you compare the cheapest JLCPCB 2 layer PCB against a not-so-cheap EU 2 Layer PCB with the same basic parameters (i.e. PCB thickness, copper thickness). The only difference being the dielectric material of the PCB. I have heard of the loss factor, dielectric absorption etc. But what effect does the overall quality of the dielectric and weave pattern etc. have on high frequency applications? Is there any difference that can be measured?
I really like this video series!
I may have answered that one already in video 1. Somewhere in that video I calculate the loss of the 10cm transmission line and compare it with the real measurements. They're spot on: AND I MEAN SPOT ON!! :-) Within 0.02dB, which is nothing. This was a board by PCBWAY. So I think those material properties are quite easy to control with modern equipment. If you really want high performance RF, you need to use rogers board material, however, a board is very very expensive when you use that.
Very good series
thanks!
this is a really beneficial , ı learned and learning a lot of things. my poaint is not memorising, understanding
thanks!
In machining apart some junk 4 layer boards on hand I have found that they seem to have the layers evenly spaced rather than the way you show in the pictures. Do you have to specify the way the layers are spaced when buying the boards? I have never seen this mentioned.
Every supplier has a different standard stackup. Eurocircuits uses 0.38 / 0.8 / 0.38 for their FR4 cores, pcbway does 0.19 / 1.2 / 0.19. However, all kinds of thicknesses are possible. There are also 75um dielectrics. If you deviate from the standard a supplier has, it does tend to get really expensive really fast :-)
Usually, the PCB stack-up for a 4+ layer board is made up of "core" and "prepreg" layers which are laminated together in a press. If you have a 6 layer board, then there will be copper foil layers added also. Generally, the core material is quite thick compared to the prepreg material.
Not saying it wouldn't ever happen, but I would be surprised to see a 4 layer board that truly had evenly spaced layers, i.e. layer 2 the same distance from both layer 1 (top) and layer 3.
Yes, never saw that either :-)
Great video. Very informative. I would also like a copy of your handbook. Thank you.😊
You're welcome.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Nice!!
Very interesting videos!
thanks
Rather than a via you could use a through hole compoinent hole and solder a lead in. These videos have been very helpful.
that would work, it would still have some inductance though. The ground plane is still quite far away, so I expect adding multiple vias will still help in that case
@@HansRosenberg74 Thanks. I was opperating under a false assumption that I really should have known. Size of the conductor has some influence on inductance,,, multiple inductors in parallel are more efffective. I get it now. If I had really payed attention in high school I likely would have had this engraved in my brain,,, but ,,,,. That was 50 years ago.
Very good series of videos. Looking forward what will come in the future. one question, maybe a silly one. Do the connections between the filter components not also act as a transmission line and that has to be taken into account?
Thanks, The wavelength is too long for the traces between to be a problem at these frequencies.
Nice series of videos you got here. Now as you said:
"Send me the check list" please
Thanks!
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Just read your checklist. Good read and helpful.
You ask in your book what are we struggling with, so here goes.
I am new to all of this, I had my first board made and it was a success. It was a simple LND150 Mosfet gain stage for use in a tube amp. I want to do more and I don’t understand layering and trace sizes. I do not understand vias and how to size them. I am used to building my board as turret boards or eyelet board. I need to understand how to select the trace size. For high current I use 18awg-20awg wire and for my small signals I use 22awg wire. How does this equate to trace size? I will still need large wire connections is this just a large via?
I am really enjoying your videos. Thank you. Chris.
Hi Chris, thanks a lot for responding to my question for ideas people are stuggeling with! These are some really good questions that I wouldn't have come up with because I've been doing this for soo long :-) I'll need to think about these to come up with a good answer. I want to come up with a simple set of rules to pick the right sizes.
About layering: Standard is 1.6mm thick pcb material with copper on both sides. This is what every supplier produces the most. You can make almost anything with this unless you start messing around with microcontrollers, ram, fine-pitch components etc. At pcbway, you can get a 10x10cm board for a little as 5 euros (you get 5 boards) plus shipping so using turret or eyelet boards is not economical anymore if you know how to design your own boards.
Best regards,
Hans
Thank you for the checklist, you earned a subscriber. Cheers. 🎉
thanks!
Excellent video! Did you try to make a shield inside the PCB, using some aligned vias connected to GND?
next video :-)
Of course I want the book also. Thank you în advance.
You're welcome.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Great videos, thank you for sharing your experience!
Here is my idea for a test: I would like to know whether a GND lane between the two signal lanes will improve the situation, maybe in comparison to the actual shield.
Thanks, that was on my list :-)
Send me the checklist
Really enjoying your videos so far!
Thanks. Link is in the description. Best regards, Hans
Some ideas for improving 2-layer board: 1mm thick board; lower via aspect ratio and multiple vias; via walls and multiple randomly spaced stitching vias between top and bottom grounds.
Nice, you just named the whole list I wrote down this morning for the next pcb. I want to see what I can do with 1.6mm since that is mechanically stronger.
I would like to get the checklist as well 😊
Link is in the description. BR, Hans
Tks 4 the checklist!
You're welcome!
Send me the checklist. Enjoyed all 3 videos. Thanks.
nice to hear! Link is in the description, best regards Hans
Great stuff! Thank you!!
Glad you liked it!
Send me the checklist. Excellent videos.
Thanks.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Very Interesting, thanks those simple rule overview, will a dense row of via's under the shielding plate not resolve crosstalk ?
You're welcome. Indeed, the 'via wall' is going to be on my next board :-)
Hey Hans, I would love to get your checklist
Link is in the description. I'm currently developing a paid course based on this document. I'm getting really excited, there is going to be so much more valuable details in there then I thought before I started developing it (I discover there are a lot of things I do without thinking about it, now that I'm paying attention to all the things I think about it's much more than I thought. I'm probably confusing now :-) ). Maybe it is something you can use (if your boss is willing to pay for it :-) ). Best regards, Hans
Thanks Hans for your excellent videos. Please send me the checklist.
You're welcome.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
What's the effect of the PCB thickness? I guess the thinner the better. How good can a very thin 2-layer-board be?
Should be really good, it will just get a bit flimsy mechanically.
Your videos are top gold, I hope for a lot more to come! How important are these design principles, when dealing with digital circuits of relatively low frequencies? Like ESP boards and their usual peripherals like I2C, SPI or USB? So far I only did 2-layer boards to save on cost and never bothered about the ground plane and messed it up by routing signals lines through it and thus cutting it up in smaller connected areas. I guess in a multi Ghz board, this would have disastrous consequences, but so far my boards do what they supposed to do.
I'll make a video on this maybe. Remember, a 10MHz clock needs a 90Mhz bandwith, you will need the 9th harmonic to make it look a bit like a square. This means you're already 'jamming' FM radio without a groundplane :-)
@@HansRosenberg74 I'm going to test this. I should be able to pick this up using the HackRF and if not, it shouldn't be too bad.
Most of my designs are battery powered and only on for a few seconds per day. The other is encased in a closed, grounded box.
Nevertheless, I would appreciate a video about EMF and how to check designs for that during design phase and after production with tools an amateur can afford.
Thank you
you're welcome
Would be great to see a 4 layer tutorial. A couple of different power lines and a few different signal wires such as spi and i2c. What's the best way.
I'll be waiting 😊
Thanks for the ideas! I really need to think what the next step will be :-) I have so many ideas.
Thank you.
you're welcome!
One Suggestion I would have for improving this kind of two layer PCB would be to make a slot in the PCB and soldering the shield to both sides of the PCB, and therefore eliminating any chance of field passing by.
haha, that would be really really drastic and effective! I'm thinking about a via-wall now. that should help a lot as well. It's also a little less rigorous :-)
@@HansRosenberg74 Via wall was the first thing that came to mind.
Curious how via stitching would effect the filter circuit on both the two layer and four layer boards.
Hello, I don't really have long term experience in PCB design or electronics as I've been in embedded software development for most of my career. However, I am now retired and doing some electronic projects, including PCB design, mostly as a hobby. I also watched many videos from Robert Ferarec, I suppose you know him.
I downloaded your Checklist Guide which it's great overall, but I found something I am not fully comfortable with based on what I learned previously.
It's about your suggested 4-layer stackup. Some designers suggest (1) placing continuous ground planes on both Inner layers and signal/power traces only on top and bottom layer, (2) trying to minimise signal vias and (3) always placing at least one sticking GND via next to any signal transfer between the outer layers.
With this stackup you make sure that one ground plane is always very near all the signal traces, which minimises crosstalk and maintains constant trace impedance regardless of trace layout. It also prevents all disruptions or cuts or signal return paths.
On the other hand, your suggested layer assignment does not seem to guarantee it all, unless you are very careful with the actual layout, which is more prone to mistakes or omissions (?).
What's your view on that?
So, please will you answer or comment on this? Thanks
Hi, what I generally do is put the signal and power traces on inner2 and put as much ground as I can on layer 4. Often this means layer 4 is a near perfect ground plane. So basically, compared to what your are saying, I'm swapping inner2 and the bottom layer. Electrically, the ground will have the same distance to the signal layer. However, by putting the signal/power traces on inner2, they are now completely shielded on both sides by ground, minimizing possible interfering transmissions. You are also offering return paths on both sides of the signals on inner 2, which is another improvement, though a minor one because of the distance to the plane on inner1. I think from an RF viewpoint, this is a better solution. A practical disadvantage could be that you cannot reach those traces for reworking or debugging.
Adding ground vias next to signal crossings is a great idea, it guarantees the shortest path for the return current.
Let me know if this clears things up. (I revised my first answer, maybe you already got this in the mail, I'm not sure, I answered to quickly but when I thought about this longer I found this was the better reply)
Best regards,
Hans
@@HansRosenberg74 Thanks, yes this makes sense to me now. As you pointed out it's basically like swapping my 3rd and 4th layers, with the added advantage of ground shielding on both sides of 3rd layer. Great!
I have now another question: According to your suggested layout, in which you use inner2 to signal/power, does it make sense to fill that layer with copper attached to power?. I mean, a mostly continuous plane of power on that layer, instead of actual traces for power. I think this will create, to some extent, a continuous capacitance between power and the adjacent ground layer which should improve power distribution (?) Does this make sense, or may it introduce other problems? Thanks again
Sounds like an excellent idea. Especially if you make sure you have decoupling caps at regular distance (which you probably already have on a lot of locations) the power plane will also perfectly function as a ground return current path. You could even place np0 / cog caps near the start and end of a signal trace between gnd and power to offer the return current a very short path along the power plane. But this would only be necessary for extreme cases where you have a very high speed line and you really can't afford a ground plane because you need to conduct a large current in the power plane so traces really won't work there.
@@HansRosenberg74 Thanks !!
Great videos Hans. Very well presented.
Please send me the check list.
You're welcome.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Very interesting videos.
But in practice 4 layer boards are way more expensive for a small scale project. Do you think working on a 0.8mm 2 layer board is a good compromise?
Yes, it is, but I always worry about 0.8mm boards with smd components. They bend easily so I worry about cracking my smd components. Bye the way, you have a 5 pieces 10x10cm 4 layer pcbs for around 50 USD these days including shipping. So it's not that bad!
Please send me the check list, your channel is outstanding!
Wow, thanks a lot :-)
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Excellent no nonsense video. Reminds me of Douglas Self's style. Please send me the checklist. Thank you
Thanks, that is a great compliment. I know Douglas Self's work!
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Thank you again
Always welcome
Send me the checklist. Your videos are awesome.
Thanks a lot!
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Another amazing video! These rare nuggets of 'integrety' or whatever you want to call it are what are still showing how good the internet can be despite it being drowned in an almost endless ocean of shit.
On the topic of PCB-design or maybe electricity in general, I've heard it being said that 'the energy is in the fields/the dielectric etc etc'. Which kind of sounds profound but I can't help but feel it kinda sounds a bit like saying 'the energy in a waterpump and pipe system is in the pipes' because there is only pressure and direction of flow because the pipes prevent the water from going everywhere. I was wondering if you had any thoughts on this?
I have little to no experience in RF electronics so maybe im just talking nonsense, but yeah.
That is a big compliment, thanks!
I haven't thought of the 'water equivalent' for RF yet. I will be making another video that explains RF a little bit further again, hopefully I can post that next week.
Compare standard FR4 to Rogers 4003C or 4350B. Rounded vs right-angle vs mitered track corners.
Thanks for the idea's, I wanted to try rogers, but man: That stuff is really expensive. Once my channel is monetized maybe I can invest a bit more. I did use rogers in the past when I was evaluating a 4-5GHz radar chip, it works wonders!
@@HansRosenberg74 You could probably ask Rogers for material if you show results on RUclips
@@rfdspguru2 It would be incredible if Rogers would join the "I want to thank Altium and PCB Way for sponsoring this video" gang 😂
I love you man! This is so gooooood!!!!!
thanks!
Send me the checklist. Thank you very much.
link is in the description
Will a row of vias, closly spaced, below the shield, can prevent the direct transmission you mentioned?
Yes, it should help and I'm planning to make a video on that. Crosstalk above the pcb will continue unless you put a shield there.
Can I get the checklist, please? Thank you! I really enjoyed your explanations.
You're welcome, You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Hi. I enjoy your videos a lot. Thanks for that. Please send me the checklist.
you're welcome
You can get it here
www.hans-rosenberg.com/checklist
BR, Hans
Send me the checklist.
I'm a PCB Designer, too. I'm always looking to improve my work. Thank you very much!
Thanks a lot. Link is in the description. Best regards hans.
Hi Hans, some top ground planes are already connected to the bottom groundplanes via the 'vias' next to the connector or am I mistaken? Great videos and thanks for sharing.
Best regards,
Ron
You are correct! I'm just adding the ground connections of the circuit itself.
Hi...what is your opinion on the Libre-VNA? I'm thinking of buying one.
I have no idea, I only know the NanoVNA (and the very expensive 'real' stuff from hp/agilent/keysight... I wonder when they're going to change their name again :-D )
Nice videos !!
Please send me the checklist!
I have some coupling troubles with my homemade domotic: when I switch some lamps on or off, it also switches another one that has a secondary switch in the same area 😅 and it quite bothers my wife 😅😅
I'm pretty sure good grounding will fix this, sounds like a typical grounding problem. Also make sure you have good supply decoupling!
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Very usefull information. I would also prefer some practical exemple. Take a design and design it with 2 layer and 4 layer. Anyway Interesting theoretical staff. How about the book, some exemple înșine?
Please check part 1 and 2, that covers differences between 2 and 4 layer designs with practical examples. The book does not have these examples, that would be something for if I turn it into a real book :-)
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Please send me the checklist. Thanks for the physical explanation of the causes of crosstalk.
You're welcome,
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Hey
Very informative, and well explained.
If the book is half as detailed, then I'm going to love it!
Please send me a copy, let me know how to get in touch.
Thanks a lot! This book is about the steps in a full pcb design process and common errors that can be avoided.
You can get it here:
www.hans-rosenberg.com/checklist
Best regards, Hans
Hey i also start electronic at 8. it´s my life
hah, nice, how old are you now, I just got 50 :-)
This is super interesting!
It has so much of the opposite problems I can face when dealing with transimpedance amplifiers. For me currents coupled capacitively make problems and having more distance between planes can help. Frequencies are a lot lower though. Any thoughts on this? Digital signals coupling to analog amplifier inputs where slight currents become amplified.
Digital 'aggressors' attacking analog 'victims' is very common, since digital signals have a very high amplitude and lots of harmonics. If you really have a crosstalk problem, the best way to deal with that is expensive: Shield cans across your analog or digital circuitry. The absolute minimum you need is a very good ground plane, without that, nothing will work.
@@HansRosenberg74 Optical sensor -> transimpedance amplifier 22bit ADC gives 5pA sensitivity😝 Leave it on a dark table facing down and it still picks up vibrations from you walking on a concrete floor... Needless to say, any signals ON the same board can be picked up too. Things are differential to help, but I am sure more can be done on the layout side. Reading out the data from FIFO is also a worry as reading fast creates higher speed flanks/frequencies but allows more sleep between reads for battery life. Slow gives less noise from readout but consumes battery; and I am sure I can spot noise from the system "running" in the data as well (albeit faint).
I think this is a "what is good enough" situation as you can always find things with high sensitivity, ultra low noise sensors.
Oef, I get your problem now :-) Sounds like a nice project.
Nice series - even if I'm not really into RF (except for an occasional NFC antenna) I still enjoyed it. But I'm more into doing stupid things with a handful, or two handfuls, of 74-series logic ICs. And being a bit on the cheap side I usually go for two-layer PCBs just like they did in the 1980s. Do you have any tips for power distribution (especially when using faster families like 74AC). And also does crosstalk matter in a digital bus, like if you do the but at 5/5mil or space the bus out to say 8/16mil?
Crosstalk is not really a problem for a hobby project. However, if you have a real product, and you don't have a ground plane and you 'forget' decoupling caps close to the power pins of your IC and don't put a ground via very close to the other side of the decoupling cap: You've build an excellent interference transmitter :-) This will get you into trouble with certification tests.
Haven't watched all of your videos, so I'm sorry if you've tested this already:
What would the result of having a ground pour on both outer layers of the board (surrounding the signal traces), as opposed to having the ground plane be on the opposite side of the board, look like? You could essentially make two ground planes with some via stitching. Just curious since I didn't see you mention pours in this video. Thanks!
Check out part 4 of this series
What would happen if you had a slot in the board in between the two connections and attached the shield directly to the bottom ground plane on the two-layer board? My apologies if this has already been suggested or thought of. One other thing with separating the ground plains, I know you're against that, but wouldn't the ground be dirty on the opposing side, and would it make sense to double the shield with an insulator in between and split the ground just at that point?
Maybe checkout my part 5 video, I try to explain the theory behind split ground planes and why it is a bad idea.
@@HansRosenberg74 yes I did that's how I know it you are against it. Well you don't have to have a split ground to have a small slot where the shield can be soldered to the bottom plane. Your thoughts on that.
yes, a shield like that should help! I'm also thinking about a via-wall, that will shield signals inside the pcb then, not above it.
@@HansRosenberg74 I wonder what would be cheaper vias + shield or a slot + shield. You're going to use a shield anyway then a slot would hold it real nice.
I think productionwise a slot with a shield inside it might be more expensive, but I'm not sure of that.
Awesome video! Dumb question, but what exactly are you measuring in the plots of crosstalk? Is it an S12 plot? Or have I misunderstood entirely?
The transfer from port 1 to port 2. So If there were no crosstalk between these two loops, this would be - infiniti dB, but it's not.
Fantastic! Please send me the checklist too!
Thank you!
Hi, the link is in the description. Sorry for the late reply, I was on the road for 2 days. Best regard, Hans
An Idea is to add a premade shielding can over the problematic parts of the board and connect it to the surface gnd. Parts that are influenced by each other woudl then be separated by a "grounded wall". singal traces are then routed on the other side to pass under the shielding can for an uninterrupted conenction of the can to gnd.
That would definitely work, however, 4 shielding cans on a filter would get rather expensive 🙂 In my next testboard I'm going to try some other measures to deal with it and improve it. Curious to see how this is going to work, I still have to make that board.
@@HansRosenberg74 What about using a nonconductive epoxy encapsulant like Stycast? In a 2-layer board where the shortest path to GND isn't necessarily between the signal trace and plane, would exchanging air for a higher Dk material help? Not saying this is practical from a cost of manufacturing standpoint, of course.
Hello Hans, would you please send me the checklist. Thank you so much for your videos
Hi, the link is in the description. Sorry for the late reply, I was on the road for 2 days. Best regard, Hans
Interesting! What if you milled a slot for the shield plate, and soldered it continously straight to the ground plane?
That should close off the x-axis leakage path completely (and also probably be way more expensive to manufacture, just a thought experiment).
Are the peaks in the 4 layer response with shield perhaps somehow related to the small leakage path that is still under the shield plate?
(I guess some minor signal can still go the long way around the ends of the plate, unless you go full monty and completely faraday cage the connections.)
PS. I'm a complete RF noob, still mostly black magic to me.
Milling a slot and putting in a shield will definately solve it. Those peaks are very common in RF. Stuff very easily starts resonating, only very clean controlled structures will not do that. This is not very 'sophisticated' so it will show this kind of behavior. Best regards, Hans
thicker ground traces? thicker copper layers?
Thicker copper traces/layers (for instance 70um instead of 35um) does not help due to the skin effect. It will help for DC (less resistance), but not for RF. Best regards, Hans