PCB Layout & Decoupling - Clarifications
SockThief , 06-28-2022, 05:51 AM
Hi!
I have been looking into "noise" lately (related topic: https://designhelp.fedevel.com/forum...-what-is-noise) and as part of that, I ended up looking at @robertferanec's series of YouTube Videos
1. Why the circuit in the thumbnail is wrong? Do you know? (https://www.youtube.com/watch?v=TVSFg94k1J0)
2. PCB Layout & Decoupling - Explained why it's so complicated (Part 1) https://youtu.be/5Ca0Eah7eKI
3. PCB Layout & Decoupling - Understanding Impedance (Part 2) https://youtu.be/Tt8X6_maj6c
4. PCB Layout & Decoupling - Measuring and Fixing (Part 3) https://youtu.be/hZSOhVdzqZk
As I understand it (and please correct me where I am wrong) - when switching an output pin, by necessity the MCU will draw more current through the Vcc pin. Due to internal resistances, parasitic capacitance and inductance the more current drawn, the higher the voltage drop on Vcc. We can reduce this by adding bulk-capacitors to stabilize the voltage rails and well-regulated power supplies. We can also add bypass capacitors (though, to be honest I really don't like this name, as it really doesn't adequately describe clearly the function). If we are switching an output with some frequency, say a PWM or serial line, then this will create a ripple on the Vcc line at the frequency we are switching the output.
So far, so good!
However, the impedance looking from the Vcc pin back towards the power supply represents a complex impedance - complex in that it is both resistive and reactive, but complex in that there are many, many factors that contribute to the total impedance; not only components like bypass capacitors, and ferrite beads, but also stray and parasitic capacitance from traces and multi layers etc. At the end, you will need to simulate and actually measure the PDN. In doing this we see that there are peaks at various frequencies and Robert + guests show that if you switch your output at the resonant frequency of the input impedance you can get some major issues. At a 10.000 foot level I get this, makes sense - however I missed how the output switching frequency is transferred to the Vcc line - I can think of two ways, but want to clarify
- is the issue caused by the Vcc ripple being the same frequency as the resonant frequency; i.e.: the output switching causes ripple on the Vcc, and when that frequency is the same as a peak in the PDN, there will be increased noise (makes sense, but not sure if this is righ)
- in one video Florian says "the square wave is made up of all the frequencies" - if i remember right, Florian is also showing the edge transition of the output and the noise on Vcc - so this somewhat makes sense as well!
Lastly, in the final video, Robert "solves" the problem by replacing the Ferrite in question with a 0-ohm resistor, but unfortunately we miss the discussion of the consequence of doing this - the Ferrite must be there for a reason, and so removing it must have some impact (though I have a separate question on Why Ferrites? https://designhelp.fedevel.com/forum...-ferrite-beads)
I have been looking into "noise" lately (related topic: https://designhelp.fedevel.com/forum...-what-is-noise) and as part of that, I ended up looking at @robertferanec's series of YouTube Videos
1. Why the circuit in the thumbnail is wrong? Do you know? (https://www.youtube.com/watch?v=TVSFg94k1J0)
2. PCB Layout & Decoupling - Explained why it's so complicated (Part 1) https://youtu.be/5Ca0Eah7eKI
3. PCB Layout & Decoupling - Understanding Impedance (Part 2) https://youtu.be/Tt8X6_maj6c
4. PCB Layout & Decoupling - Measuring and Fixing (Part 3) https://youtu.be/hZSOhVdzqZk
As I understand it (and please correct me where I am wrong) - when switching an output pin, by necessity the MCU will draw more current through the Vcc pin. Due to internal resistances, parasitic capacitance and inductance the more current drawn, the higher the voltage drop on Vcc. We can reduce this by adding bulk-capacitors to stabilize the voltage rails and well-regulated power supplies. We can also add bypass capacitors (though, to be honest I really don't like this name, as it really doesn't adequately describe clearly the function). If we are switching an output with some frequency, say a PWM or serial line, then this will create a ripple on the Vcc line at the frequency we are switching the output.
So far, so good!
However, the impedance looking from the Vcc pin back towards the power supply represents a complex impedance - complex in that it is both resistive and reactive, but complex in that there are many, many factors that contribute to the total impedance; not only components like bypass capacitors, and ferrite beads, but also stray and parasitic capacitance from traces and multi layers etc. At the end, you will need to simulate and actually measure the PDN. In doing this we see that there are peaks at various frequencies and Robert + guests show that if you switch your output at the resonant frequency of the input impedance you can get some major issues. At a 10.000 foot level I get this, makes sense - however I missed how the output switching frequency is transferred to the Vcc line - I can think of two ways, but want to clarify
- is the issue caused by the Vcc ripple being the same frequency as the resonant frequency; i.e.: the output switching causes ripple on the Vcc, and when that frequency is the same as a peak in the PDN, there will be increased noise (makes sense, but not sure if this is righ)
- in one video Florian says "the square wave is made up of all the frequencies" - if i remember right, Florian is also showing the edge transition of the output and the noise on Vcc - so this somewhat makes sense as well!
Lastly, in the final video, Robert "solves" the problem by replacing the Ferrite in question with a 0-ohm resistor, but unfortunately we miss the discussion of the consequence of doing this - the Ferrite must be there for a reason, and so removing it must have some impact (though I have a separate question on Why Ferrites? https://designhelp.fedevel.com/forum...-ferrite-beads)
qdrives , 06-28-2022, 04:40 PM
The true specialist on these topics is Steve Sandler ( @Steve.Picotest ). If you are lucky, he will answer your questions, but probably will mention he has already made a post about a similar topic somewhere as you are not the first to ask the question.
The problem with power supplies (as far as I know) related to your questions are:
- Anti-resonance between capacitors.
- Resonance frequency between inductor (as a ferrite bead also is one) and the capacitor.
A lot of people think that ferrite beads are perfect for filtering. I was in that category too.
However, there are serious drawbacks if you are not careful with the design. Either have enough resistance to have a flat impedance or add a snubber to dampen the resonance.
Can you replace a ferrite bead by a low ohmic resistor? Often yes.
The problem with power supplies (as far as I know) related to your questions are:
- Anti-resonance between capacitors.
- Resonance frequency between inductor (as a ferrite bead also is one) and the capacitor.
A lot of people think that ferrite beads are perfect for filtering. I was in that category too.
However, there are serious drawbacks if you are not careful with the design. Either have enough resistance to have a flat impedance or add a snubber to dampen the resonance.
Can you replace a ferrite bead by a low ohmic resistor? Often yes.
SockThief , 06-29-2022, 06:14 AM
thanks as always @qdrives
as is always the case, I strongly doubt I am the first to ask! the answers are normally out there *somewhere* - it can just be difficult filtering the noise (see what I did there?!) to find them - i assure you I *always* try and do as much research and reading as I can before asking! but please if this has been answered somewhere else, please let me know. however, not only is this about finding an answer to my specific question, its also some feedback to @robertferanec on the video on how to clarify an ambiguous section - otherwise it was an excellent series of videos.
i mentioned in private to robert that other than the technical content, one of the best parts of these videos was his humility. Many of us look up to Robert, especially in his position as an educator. However even with that he was open to say "i learned this" or "i am so embarrassed that I made that mistake - but i left it in the video" and that he actively seeks experts in areas he sees as areas he can improve, with Florian he gave his design and said "here! break this and show me what i did wrong"; this level of openness, honesty and transparency is rare - and it gives people like me the confidence to say "perhaps I should know this, but i'm not sure - could someone help me out" - and he should be commended and celebrated for that.
as is always the case, I strongly doubt I am the first to ask! the answers are normally out there *somewhere* - it can just be difficult filtering the noise (see what I did there?!) to find them
- i assure you I *always* try and do as much research and reading as I can before asking! but please if this has been answered somewhere else, please let me know. however, not only is this about finding an answer to my specific question, its also some feedback to @robertferanec on the video on how to clarify an ambiguous section - otherwise it was an excellent series of videos.i mentioned in private to robert that other than the technical content, one of the best parts of these videos was his humility. Many of us look up to Robert, especially in his position as an educator. However even with that he was open to say "i learned this" or "i am so embarrassed that I made that mistake - but i left it in the video" and that he actively seeks experts in areas he sees as areas he can improve, with Florian he gave his design and said "here! break this and show me what i did wrong"; this level of openness, honesty and transparency is rare - and it gives people like me the confidence to say "perhaps I should know this, but i'm not sure - could someone help me out" - and he should be commended and celebrated for that.
qdrives , 06-29-2022, 12:36 PM
Perhaps it is ambiguous as there still is a lot of unknowns in that area. But I too must admit I do not know everything.
SockThief , 06-30-2022, 12:55 AM
I am reading some articles on the topic, Clean Power for Every IC (https://www.allaboutcircuits.com/tec...ss-capacitors/) - there is a section dedicated to the signal transition and ringing for an inverter. What I am quickly learning is that it is an extremely complex topic; decoupling, bypass and noise... it is a very deep rabbit hole!
i believe I have a better understanding of the topic now, though far from perfect; I need to focus on building my experience and keep learning from people like Robert.
I might have to go back and watch the series again with new eyes!
i believe I have a better understanding of the topic now, though far from perfect; I need to focus on building my experience and keep learning from people like Robert.
I might have to go back and watch the series again with new eyes!
robertferanec , 07-08-2022, 12:01 AM
@SockThief Thank you for nice words. Yes, I keep learning and there are number of things and videos where I was wrong and I am sure there are things where I am still wrong. But the interviews I make with different people help me a lot.
Use our interactive Discord forum to reply or ask new questions.