Excellent video! I hope a lot of people see this video. I have actually seen a design where someone have placed all the 100nF decoupling capacitors in one corner of the PCB "because it looked nicer and didn't mess up the routing" 🤯😱
I work in a company where we have dedicated layout engineers so the schematic designers will always put decoupling capacitors next to their IC in the schematic. You’re right that it sometimes can get in the way of readability but often the schematic can be organized in such a way that the reader’s eyes still follows the signal flow appropriately. Schematic readability really is an art form.
I have to confess that since I'm mostly making PCBs for my own use, I tend not to put a ton of time into making the schematic pretty. If I was working at a company, of course, I'd operate differently. (I do make sure that the labels aren't overlapping, etc).
Interesting, I've been binge watching the Altium channel. Eric Bogatin has some interesting things to say about capacitors, Rick Hartley too. What Type and Quality of capacitors makes more of a difference than what a schematic says. Schematics are the outline, the Rubber meets the road with the PCB designer.
just found your videos and love them all, thank you for the work you've put into making them intuitive and helpful! re: schematic placement, I like to place IC power pins into their own little corner (labeled something like "cv summing amp power") for clarity and section them off via dashed lines on the schematic - my schematics aren't open source (yet) but I try to organize my schematics with "future me, just woke up, looking at this schematic 6 months after shelving the project initially because I already have 50 other half finished concepts in progress" in mind
I work in an environment where multiple people can be working on a schematic or pcb design. Putting all of your bypass caps away from their intended place on the schematic is just asking for trouble. If you are worried about clean schematics, just increase your page size or separate the design into multiple pages. You don't need to put bypass caps in the signal flow, but the power pins and caps should be near the circuit they are intended to bypass
Definitely, if I was working at a company I'd go with whatever the "house style" is. And large projects have different demands than me laying out a small synth module.
I agree with many comments below. Putting your bypass caps next to incoming power filter is confusing. I also agree that directly attaching them to symbol is clutter. I place the cluster bypass cap on the same schematic page but label them showing IC ref designator.
@@Lantertronics For a large design, I typically have a cover page, power supply, CPU, analog, I/O pages in my schematics. For a small one, I have cover page then sub-divide page 2 into the same categories.
Nice video! An interesting follow-up video may be from 'w2aew', #257: his video on power supply decoupling & filtering. He adds a nice demonstration of what placement of such a 100nF cap exactly does.
Some PCB designers advise that you should use the fewest capacitors you can get away with. Their reasoning is that it shortens the BOM, resulting in lower costs in both parts and assembly. They also argue it increases reliability of the circuit given capacitors have the highest failure rate of any PCB component. If the failure mode is a short, it also poses a fire hazard, requiring careful risk assessment and mitigation such as adding circuit protection, which in turn increases manufacturing and design costs.
That's Ian Fritz's position (from the synth DIY mailing list); he uses fewer decoupling caps. If I had Ian's expertise and experience, I'd do that, but I honestly don't have enough knowledge to figure out where those fewer capacitors should go. "What you can get away with" is a bit slippery -- every one in a while you'll get a weird motorboating effect or something else strange without bypassing, but quite often you can just leave the bypass capacitors out and test your module and not notice a difference in casual listening (I usually don't bother with the bypass caps when breadboarding, for instance, unless I run into something weird). It's sort of targeting those edge cases, I guess.
Super helpful video! Thank you! Although I have learned it over time, this was indeed a big mystery in the beginning, especially the thing with moving power pins and caps to somewhere else in the schematic. :-) Now it is my preference as it brings clarity, but it was initially really confusing. :-D
I'd prefer to see a logical separation between power filtering for the board and power delivery to the IC, even in a simple circuit like this. The two fill different enough roles that they shouldn't be grouped in my opinion. Consider the situation where you need to introduce some additional complexity. Say your ferrite beads and the capacitors after them start to oscillate in testing requiring you to introduce a damping resistor or RC network to prevent the power filter from oscillating. Because power filtering for the board doesn't have it's own space it could be ambiguous from the schematic what part of the circuit those filter elements belong to unless you understood enough of the circuit to know the exact problem they were there to solve. If everything related to power filtering for the board is on its own then that isn't an issue. It makes the revision process more cumbersome as well because there are very few circumstances where you'd want to change the board's power filtering and the decoupling for an IC on the board simultaneously. Say you want to change the way power is delivered to your board. For example you want to alter the board to operate from a different supply voltage by introducing a switching regulator to produce the positive and negative supply voltages. In that circumstance you will want to tear out everything related to power filtering but would almost certainly be keeping the decoupling capacitors exactly the same. On the other hand if you decide you want to replace an IC with an alternative which has additional or fewer power pins you will need to change how you're decoupling that IC which has nothing to do with the boards power filtering. I don't mind when all the decoupling caps are lined up in the corner assuming there's something about the schematic which makes it clear what components those caps belong to. For example if an MCU is off on its own page and in the corner of that page there is a line of caps then it's clear that the MCU and those caps are related. Grouped sections of a circuit which connect via wires rather than net labels should be logically related though and not just functionally related. The decoupling caps and power filter are certainly performing similar functions, in that they are both part of the PDN, but there are very few situations where you would logically group them during design or layout. It really is such a minor issue that I'd never bring it up but since you're trying to teach the correct way to do things, insofar as a correct way even exists, then that's the approach I'd personally advocate for.
0.1uF was the “rule of thumb” back in the old TLL days of logic boards where clock frequencies were in the few MHz range. At audio frequencies a 0.1uF has a reactance of ~21 ohms at 60Hz. So it would be of little benefit. You would be better off with a larger value such as 1.0uF, IMO. Also, bypassing close the to pin as you first showed is ok IF there is a ground plane. Otherwise locate the caps top and bottom so the ground between them is as short as possible. You run the ground up the middle of the IC. This keeps circulating currents “tightly” constrained. I could go on but you get the idea. It can be complicated for some designs that have multiple frequencies involved like clocks and LFOs.
Hmmmm.... out of the hundreds of synth circuits I've looked at, I don't think I've ever seen one with local bypass caps as high as 1.0uF. That's a lot of uF. I just checked the TL07x data sheet, which explicitly says to use 0.1 uF. I don't think it's targeting audio frequencies per se; higher frequency gunk can wind up "detected" by nonlinearities (think of the way radio broadcasts sometimes show up output by a guitar amplifier.) I'm just spitballing, though.
Ah -- I should have clarified when I was showing the different places to put the cap that I could choose one based on whatever other traces needed to be routing. (There's so many conflicting design principles when laying out a PCB).
@@Lantertronics as I said, it was the rule of thumb and it is better than nothing! Many data sheets will also list the power supply rejection ratio and as far as the TL072 data sheet, they specifically mention “to reduce errors” meaning high frequency Common mode noise. I just know from experience designing logic and analog circuits back in the 80’s that everything got a 0.1uF for primarily 3 reasons. 1. Power supplies were just traces on the board so local bypassing was crucial. (It also increased the impedance of the supply making 0.1uF more effective and the TL072 data sheet specifically mentions “high impedance power supplies). 2. 0.1uF was about the largest value cap in a small package that was available and had low ESR. Higher values were either electrolytic or tantalum which increased cost significantly. 3. Because power and ground were just traces on the board, high frequency coupling was significantly more troublesome than low frequency. Plus you got the benefit of distributed capacitance. A lot has been learned over the years (the TL072 data sheet for TI was first published in 1978 with minor updates to parameters since then) and the designs I’ve worked on most recently will have a multiple bypass caps from 1.0uF-10uF up to 10-100pF to get broad range coverage as opamps today frequently have gain-bandwidths in the 100s of MHz. Low ESR ceramic caps in the uF range are pretty common place now. I think using power planes and careful routing to control ground currents will benefit a lot more than a local 0.1uF, but you can’t go wrong by placing one locally. I didn’t mean to imply anything you said was “wrong” by any means, my intent was to add context but I could have worded it better and I probably dug myself into a hole. My attempted clarification may have only succeeded in digging it deeper. If so, I apologize in advance.
Also I have personally found errors in datasheets as well as found they will change the test conditions to optimize the parameter they are testing so you can expect one or the other but not both at the same time! So I’ve learned to only use the data sheet as a rough estimate and if a particular parameter is important, be sure to do the research and understand it thoroughly before accepting it as truth. I’ve been burned too many times! Lol.
I remember the "Point one for every chip" design days, but mostly the 0.1uF "shortages" = price increases. 0.1uf became the most expensive value of Thru-hole Ceramic Cap in every vendors catalog back then. Luckily I noticed that the price of 0.082uF and 0.12uF were still being priced at their original cost - indeed ~75% less expensive than the jacked-up prices of 0.1uF caps. Most importantly those 2 other values were _always_ in stock at every vendor, and all 3 are the same board sizes! (ColoGraphicComputers Atlanta, Ga 1980's)
Why use Z5U and Y5V dielectrics in this day? Especially for 100nF caps it's not worth the hassle of taking into account the variation of capacitance with temperature, bias voltage etc Just use X7R for everything! Regarging schematics, personally I prefer the power supply pins (and the respective caps) for opamps separate, but close to the opamps. If you gather the supply pins for all the opamps together it can quickly get very confusing and hard to tell which opamp uses what power rails.
Really great videos you are putting out. Subscribed. One request if possible, is there a way to invert the color scheme on your circuit modeling software? I find it hard to see and read with all that bright white and dark lines. Even if that's not possible, great videos and thank you.
That sort of assumes that everything is operating linearly with no intermodulation distortion going on. I really should make a video on this. In the mean time, consider this: sometimes guitar amps will pick up radio transmissions, which are well above the audio range, and you can hear the transmission coming out the speaker. Contemplate how that happens.
And I'm stretching here, but think about a MHz range parasitic oscillation appearing in a synth op amp circuit. Can you hear it? No. But it can still eat up your dynamic range and make your audio sound weak and crappy. (This can happen when people "roll op amps" and start plugging in op amps designed for video signals into their audio amp without really knowing what they're doing).
The main google hit is "The Truth About Ferrite Beads Will Shock You" -- and the author of that is overlooking a lot. I started using them because Paul Schreiber of MOTM did in sort of a "cargo cult" fashion. I didn't really follow his reasoning until much later.
we had to put in ferrite beads in some of our machines the other day otherwise they shut down, ferrute beads really do work if they are specified correctly.
I usually put all the op-amp / whatever power pin symbols with bypass caps attached on the power supply schematic page. EDIT: Oh that's what you're doing hahaha
Excellent video! I hope a lot of people see this video. I have actually seen a design where someone have placed all the 100nF decoupling capacitors in one corner of the PCB "because it looked nicer and didn't mess up the routing" 🤯😱
HaHaHa!
Wow!
I work in a company where we have dedicated layout engineers so the schematic designers will always put decoupling capacitors next to their IC in the schematic. You’re right that it sometimes can get in the way of readability but often the schematic can be organized in such a way that the reader’s eyes still follows the signal flow appropriately. Schematic readability really is an art form.
I have to confess that since I'm mostly making PCBs for my own use, I tend not to put a ton of time into making the schematic pretty. If I was working at a company, of course, I'd operate differently. (I do make sure that the labels aren't overlapping, etc).
Interesting, I've been binge watching the Altium channel. Eric Bogatin has some interesting things to say about capacitors, Rick Hartley too. What Type and Quality of capacitors makes more of a difference than what a schematic says. Schematics are the outline, the Rubber meets the road with the PCB designer.
@@Dazza_Doo Interesting coincidence I just downloaded Altium (educational license) to try out.
@@Lantertronics I'm so jealous
your videos are so helpful I'm still watching them even after I graduated, Dr. Lanterman!
Thank you for your kind words!
this was my first mistake on pcb design, and the output madly produced self-oscillation.
thanks for pointing out.
Beads they just abide, like the Dude! I loved this reference! :D
just found your videos and love them all, thank you for the work you've put into making them intuitive and helpful!
re: schematic placement, I like to place IC power pins into their own little corner (labeled something like "cv summing amp power") for clarity and section them off via dashed lines on the schematic - my schematics aren't open source (yet) but I try to organize my schematics with "future me, just woke up, looking at this schematic 6 months after shelving the project initially because I already have 50 other half finished concepts in progress" in mind
I work in an environment where multiple people can be working on a schematic or pcb design. Putting all of your bypass caps away from their intended place on the schematic is just asking for trouble. If you are worried about clean schematics, just increase your page size or separate the design into multiple pages. You don't need to put bypass caps in the signal flow, but the power pins and caps should be near the circuit they are intended to bypass
Definitely, if I was working at a company I'd go with whatever the "house style" is. And large projects have different demands than me laying out a small synth module.
I agree with many comments below. Putting your bypass caps next to incoming power filter is confusing. I also agree that directly attaching them to symbol is clutter. I place the cluster bypass cap on the same schematic page but label them showing IC ref designator.
Ah for multiple page designs that's an excellent idea, as is adding some clarifying labels.
@@Lantertronics For a large design, I typically have a cover page, power supply, CPU, analog, I/O pages in my schematics. For a small one, I have cover page then sub-divide page 2 into the same categories.
Nice video! An interesting follow-up video may be from 'w2aew', #257: his video on power supply decoupling & filtering. He adds a nice demonstration of what placement of such a 100nF cap exactly does.
Thanks, I will check it out!
Some PCB designers advise that you should use the fewest capacitors you can get away with. Their reasoning is that it shortens the BOM, resulting in lower costs in both parts and assembly. They also argue it increases reliability of the circuit given capacitors have the highest failure rate of any PCB component. If the failure mode is a short, it also poses a fire hazard, requiring careful risk assessment and mitigation such as adding circuit protection, which in turn increases manufacturing and design costs.
That's Ian Fritz's position (from the synth DIY mailing list); he uses fewer decoupling caps. If I had Ian's expertise and experience, I'd do that, but I honestly don't have enough knowledge to figure out where those fewer capacitors should go. "What you can get away with" is a bit slippery -- every one in a while you'll get a weird motorboating effect or something else strange without bypassing, but quite often you can just leave the bypass capacitors out and test your module and not notice a difference in casual listening (I usually don't bother with the bypass caps when breadboarding, for instance, unless I run into something weird). It's sort of targeting those edge cases, I guess.
Super helpful video! Thank you! Although I have learned it over time, this was indeed a big mystery in the beginning, especially the thing with moving power pins and caps to somewhere else in the schematic. :-) Now it is my preference as it brings clarity, but it was initially really confusing. :-D
I'd prefer to see a logical separation between power filtering for the board and power delivery to the IC, even in a simple circuit like this. The two fill different enough roles that they shouldn't be grouped in my opinion.
Consider the situation where you need to introduce some additional complexity. Say your ferrite beads and the capacitors after them start to oscillate in testing requiring you to introduce a damping resistor or RC network to prevent the power filter from oscillating. Because power filtering for the board doesn't have it's own space it could be ambiguous from the schematic what part of the circuit those filter elements belong to unless you understood enough of the circuit to know the exact problem they were there to solve. If everything related to power filtering for the board is on its own then that isn't an issue.
It makes the revision process more cumbersome as well because there are very few circumstances where you'd want to change the board's power filtering and the decoupling for an IC on the board simultaneously. Say you want to change the way power is delivered to your board. For example you want to alter the board to operate from a different supply voltage by introducing a switching regulator to produce the positive and negative supply voltages. In that circumstance you will want to tear out everything related to power filtering but would almost certainly be keeping the decoupling capacitors exactly the same. On the other hand if you decide you want to replace an IC with an alternative which has additional or fewer power pins you will need to change how you're decoupling that IC which has nothing to do with the boards power filtering.
I don't mind when all the decoupling caps are lined up in the corner assuming there's something about the schematic which makes it clear what components those caps belong to. For example if an MCU is off on its own page and in the corner of that page there is a line of caps then it's clear that the MCU and those caps are related. Grouped sections of a circuit which connect via wires rather than net labels should be logically related though and not just functionally related. The decoupling caps and power filter are certainly performing similar functions, in that they are both part of the PDN, but there are very few situations where you would logically group them during design or layout.
It really is such a minor issue that I'd never bring it up but since you're trying to teach the correct way to do things, insofar as a correct way even exists, then that's the approach I'd personally advocate for.
I only use one capacitor and because i always use sockets, it allows me to place the 1206 smd capacitor right in the center of the opamp.
nice tip!
0.1uF was the “rule of thumb” back in the old TLL days of logic boards where clock frequencies were in the few MHz range. At audio frequencies a 0.1uF has a reactance of ~21 ohms at 60Hz. So it would be of little benefit. You would be better off with a larger value such as 1.0uF, IMO.
Also, bypassing close the to pin as you first showed is ok IF there is a ground plane. Otherwise locate the caps top and bottom so the ground between them is as short as possible. You run the ground up the middle of the IC. This keeps circulating currents “tightly” constrained.
I could go on but you get the idea. It can be complicated for some designs that have multiple frequencies involved like clocks and LFOs.
Hmmmm.... out of the hundreds of synth circuits I've looked at, I don't think I've ever seen one with local bypass caps as high as 1.0uF. That's a lot of uF. I just checked the TL07x data sheet, which explicitly says to use 0.1 uF. I don't think it's targeting audio frequencies per se; higher frequency gunk can wind up "detected" by nonlinearities (think of the way radio broadcasts sometimes show up output by a guitar amplifier.) I'm just spitballing, though.
Ah -- I should have clarified when I was showing the different places to put the cap that I could choose one based on whatever other traces needed to be routing. (There's so many conflicting design principles when laying out a PCB).
@@Lantertronics as I said, it was the rule of thumb and it is better than nothing! Many data sheets will also list the power supply rejection ratio and as far as the TL072 data sheet, they specifically mention “to reduce errors” meaning high frequency Common mode noise. I just know from experience designing logic and analog circuits back in the 80’s that everything got a 0.1uF for primarily 3 reasons.
1. Power supplies were just traces on the board so local bypassing was crucial. (It also increased the impedance of the supply making 0.1uF more effective and the TL072 data sheet specifically mentions “high impedance power supplies).
2. 0.1uF was about the largest value cap in a small package that was available and had low ESR. Higher values were either electrolytic or tantalum which increased cost significantly.
3. Because power and ground were just traces on the board, high frequency coupling was significantly more troublesome than low frequency.
Plus you got the benefit of distributed capacitance.
A lot has been learned over the years (the TL072 data sheet for TI was first published in 1978 with minor updates to parameters since then) and the designs I’ve worked on most recently will have a multiple bypass caps from 1.0uF-10uF up to 10-100pF to get broad range coverage as opamps today frequently have gain-bandwidths in the 100s of MHz. Low ESR ceramic caps in the uF range are pretty common place now.
I think using power planes and careful routing to control ground currents will benefit a lot more than a local 0.1uF, but you can’t go wrong by placing one locally.
I didn’t mean to imply anything you said was “wrong” by any means, my intent was to add context but I could have worded it better and I probably dug myself into a hole. My attempted clarification may have only succeeded in digging it deeper. If so, I apologize in advance.
Also I have personally found errors in datasheets as well as found they will change the test conditions to optimize the parameter they are testing so you can expect one or the other but not both at the same time! So I’ve learned to only use the data sheet as a rough estimate and if a particular parameter is important, be sure to do the research and understand it thoroughly before accepting it as truth. I’ve been burned too many times! Lol.
I remember the "Point one for every chip" design days, but mostly the 0.1uF "shortages" = price increases.
0.1uf became the most expensive value of Thru-hole Ceramic Cap in every vendors catalog back then.
Luckily I noticed that the price of 0.082uF and 0.12uF were still being priced at their original cost - indeed ~75% less expensive than the jacked-up prices of 0.1uF caps.
Most importantly those 2 other values were _always_ in stock at every vendor, and all 3 are the same board sizes!
(ColoGraphicComputers Atlanta, Ga 1980's)
Why use Z5U and Y5V dielectrics in this day?
Especially for 100nF caps it's not worth the hassle of taking into account the variation of capacitance with temperature, bias voltage etc
Just use X7R for everything!
Regarging schematics, personally I prefer the power supply pins (and the respective caps) for opamps separate, but close to the opamps.
If you gather the supply pins for all the opamps together it can quickly get very confusing and hard to tell which opamp uses what power rails.
Really great videos you are putting out. Subscribed. One request if possible, is there a way to invert the color scheme on your circuit modeling software? I find it hard to see and read with all that bright white and dark lines. Even if that's not possible, great videos and thank you.
Ah, hmmmm... there must be a way to do that, and that's a good idea.
From my understanding, "beads" on power lines have only esotheric value
That sort of assumes that everything is operating linearly with no intermodulation distortion going on. I really should make a video on this. In the mean time, consider this: sometimes guitar amps will pick up radio transmissions, which are well above the audio range, and you can hear the transmission coming out the speaker. Contemplate how that happens.
And I'm stretching here, but think about a MHz range parasitic oscillation appearing in a synth op amp circuit. Can you hear it? No. But it can still eat up your dynamic range and make your audio sound weak and crappy. (This can happen when people "roll op amps" and start plugging in op amps designed for video signals into their audio amp without really knowing what they're doing).
The main google hit is "The Truth About Ferrite Beads Will Shock You" -- and the author of that is overlooking a lot. I started using them because Paul Schreiber of MOTM did in sort of a "cargo cult" fashion. I didn't really follow his reasoning until much later.
we had to put in ferrite beads in some of our machines the other day otherwise they shut down, ferrute beads really do work if they are specified correctly.
I usually put all the op-amp / whatever power pin symbols with bypass caps attached on the power supply schematic page.
EDIT: Oh that's what you're doing hahaha
Power corrupts, absolute power ... *looks at room number* ... sorry wrong class! 😅
Hah!
squoosh