Some regulator series (for example *1117 from SOME manufacturers) require ESR of capacitors on the output to be within some range, for example 0.1 ohm -- 1 ohm. In such scenario, using an electrolytic capacitor can be better than using a ceramic capacitor. If using a ceramic capacitor is required due to size constraints or other reasons and a better regulator is not possible, then one could add a resistor in series with the capacitor to introduce some ESR into the circuit. Also, some datasheets say something similar to "a 10uf tantalum capacitor is enough" ... keep in mind they may say tantalum because the ESR of tantalum capacitors was usually quite low (in the hundreds of mOhm) but not as low as ceramics and way back then on the first revisions of those datasheets electrolytic capacitors may have been much worse specs wise, so if you want to use electrolytic or polymer capacitors instead, you may have to go with a slightly higher capacitance value. Tantalums are often more expensive and they have some drawbacks, like potentially burning up / exploding on overvoltage (hence it's often recommended to use tantalum capacitors rated for at least 50% more than regular expected voltage)
Agreed, except be careful with polymer caps as the lowest ESR caps are actually polymer ones. They really are made to be mainly used in very modern high frequency dc-dc converters requiring significant currents like the ones you find in GPU powering circuitry. And they are lower max voltage, higher cost, lower availability anyway, so actually more disadvantages. When you are dealing with very sharp voltage drops as demonstrated in this video you are more after lower equivalent inductance (including parasitic) than capacitance. And for that purpose all application notes say that smaller footprint SMD ceramics placed as close as possible (obviously not on breadboard) outperform bigger and higher capacitance ones due to physics of internal structure. Best of both worlds is to put like 0.1-1uF caps very close in pair with 4.7-10uF bulk input cap where available nearby. When you see in schematics a bunch of same small value non-polar caps in parallel, this is the same thing but with compromise for consolidation of nominal values. Output side may or may not desire low value cap depending on the following load factors.
great subject, great watch. Even though I understand these principals beforehand, it is still appreciated to see the real-world waveforms as it actually saves considerable amounts of time performing these experiments on ones own.
Ha ha. That audiophile effort, as with the rest of the quackery that goes on in that industry, is hilarious. In response to your question of whether we'd like to see more detail on smps filtering, I for one would be very interested. I'm particularly interested in the implications of driving high current solenoids (about 4A). Thanks for your intelligent videos. Much appreciated.
Anything that has an effect on the waveform will affect the sound the amplifier makes, the goal for audiophiles isn't make a well functioning power supply it's to make a nice sounding amplifier.
Great vid. It is good to have visual confirmation and explanation of the conclusions I have drawn after many experiments where adding more capacitance never seemed to have the desired effect. Thanks. +1 switching output cap discussion
I tend to have been using LDO's for their efficiency and tend to use about 10uf ceramic in parallel with the local and bulk decoupling. I only know 1000uf's are only required after rectification for smoothing. Other than large fast loads like you mentioned.
Well, any reactance on the output introduces a pole which in the worst case violates the Nyquist stability criterium. You showed nicely the impact on transient load changes where oscillatory behavior became apparent already. In principle a regulator is as you say just the equivalent to an op-amp. With the proper compensation you can achieve better slew rates than just with the .20dB/dec response. In a power supply you want to enforce stability which as you demonstate can easily be destroyed. Anyway in your drawing looks like that likely by mistake you switched the inputs of the op-amp - thus adding the compensating capacitance to the non-inverting input and output which surely causes instability given its value.
What value was the ceramic capacitor you used? The mlcc waveform looked horrible though the conventional wisdom is to put small ceramic capacitors on the output.
@@sdgelectronics Ok those are the values of mlcc I usually use on the output of 78xx regulators, I don't have a scope though so I could never check if it was working. Datasheets usually recommend 100nf ceramic on the output maybe 100nf's the largest mlcc can be without interfering with the feedback loop? Either way I'm going to stick with a 1uf+ tantalum in parallel with a 100n mlcc now hopefully that should be ok.
Good video. For linear regulators, such as 78xx, it is not recommended to put low esr capacitors at the output. This can cause generation. As well as high-capacity capacitors. But if you need a large capacitor capacity in the circuit, and it stands after the linear regulator, then a 4.7-15 ohm resistor must be installed between them.
As always, then, follow the spec sheet! I'd be very interested to see exactly the same, but for LDO regulators. How is all this affected by the distributed supply capacitance in a complex circuit with local decoupling on multiple devices? Do we simply rely on the local supply inductance to allow the regulator the freedom to regulate?
I have an electronic/mechanic background, although most of the electronics were plug and play. I am now wanting to build and design my own projects now that I'm retired. Some of this was over my head for where I'm at understanding wise. Could you suggest some videos that may start more with laymen terms? I think you do a great job, just some of it's over my head..
Nice video. Regarding ringing with the ceramic capacitor on the output, I think you said it was due to the low ESR, but to solve it you increased the capacitance as well as using an electrolytic. Could you instead have added some resistance in series with the ceramic capacitor, would that have worked?
You could, but then it loses the high frequency performance. It might be possible to reduce the problem with an actual PCB since the breadboard used in this video does add some parasitics.
Excelent video, very well demonstrated, the capacitor in your diagram, inbetween the opamp output and the non-inverting input, what is it used for. Also could a similar test setup be used to get an estimation of the Phase Margin of the feed back loop by observing the oscilations on the step impulse at the output of the regulator. Thanks again, a great video.
I did the drawing slightly incorrect, swapping the + and - symbol around on the op-amp. The capacitor limits the bandwidth to prevent unstable behaviour at high frequency
I might be wrong but I'm sure XraytonyB had some problem with ringing in a tone circuit when he changed the electrolytics for ceramics and was asking if anyone knew what the reason was for it, is this a related issue as he changed them back and the ringing stopped.
Without all the output caps the inrush current will have no where to go. ;-) an often forgotten part of regulators is the input ripple rejection within the audio band.
The ESP8266 is a difficult IC. 3.3 volts and a few mA, but high spikes when it connects to the wifi. Not every linear regulator can do this. Usually the 3.3 volt is generated from a 5 volt. That's why I put enough uF in front of the 3.3 volt regulator as a buffer. In the project I had a 24-30 volt supply voltage, from which a stepdown generated 5 volts, then the 3.3 volt linear regulator.
@@suncrafterspielt9479 It's another graphic figure(symbol) for representing two transistors with common bases tied to one collector, aka current mirror. It takes less space. Regarding current mirrors, take a look at this: th-cam.com/video/VnJHXQCPIvs/w-d-xo.html th-cam.com/video/B9PJsJsK3mk/w-d-xo.html
Maybe for the input but I thought part of the message here was that having too much capacitance on the output can be detrimental, so you probably don't want to go overboard with too many different types in parallel there?
Some regulator series (for example *1117 from SOME manufacturers) require ESR of capacitors on the output to be within some range, for example 0.1 ohm -- 1 ohm. In such scenario, using an electrolytic capacitor can be better than using a ceramic capacitor. If using a ceramic capacitor is required due to size constraints or other reasons and a better regulator is not possible, then one could add a resistor in series with the capacitor to introduce some ESR into the circuit.
Also, some datasheets say something similar to "a 10uf tantalum capacitor is enough" ... keep in mind they may say tantalum because the ESR of tantalum capacitors was usually quite low (in the hundreds of mOhm) but not as low as ceramics and way back then on the first revisions of those datasheets electrolytic capacitors may have been much worse specs wise, so if you want to use electrolytic or polymer capacitors instead, you may have to go with a slightly higher capacitance value. Tantalums are often more expensive and they have some drawbacks, like potentially burning up / exploding on overvoltage (hence it's often recommended to use tantalum capacitors rated for at least 50% more than regular expected voltage)
Agreed, except be careful with polymer caps as the lowest ESR caps are actually polymer ones. They really are made to be mainly used in very modern high frequency dc-dc converters requiring significant currents like the ones you find in GPU powering circuitry. And they are lower max voltage, higher cost, lower availability anyway, so actually more disadvantages.
When you are dealing with very sharp voltage drops as demonstrated in this video you are more after lower equivalent inductance (including parasitic) than capacitance. And for that purpose all application notes say that smaller footprint SMD ceramics placed as close as possible (obviously not on breadboard) outperform bigger and higher capacitance ones due to physics of internal structure. Best of both worlds is to put like 0.1-1uF caps very close in pair with 4.7-10uF bulk input cap where available nearby. When you see in schematics a bunch of same small value non-polar caps in parallel, this is the same thing but with compromise for consolidation of nominal values. Output side may or may not desire low value cap depending on the following load factors.
+1 would love to learn more about output capacitance on switching regulators
great subject, great watch.
Even though I understand these principals beforehand, it is still appreciated to see the real-world waveforms as it actually saves considerable amounts of time performing these experiments on ones own.
Careful the Audiophool Mafia will be after you :D
Don't open Pandora's box 😂. Neither talk about the 32bit audio idiotness or picosecond jittering!
Ha ha. That audiophile effort, as with the rest of the quackery that goes on in that industry, is hilarious. In response to your question of whether we'd like to see more detail on smps filtering, I for one would be very interested. I'm particularly interested in the implications of driving high current solenoids (about 4A). Thanks for your intelligent videos. Much appreciated.
Anything that has an effect on the waveform will affect the sound the amplifier makes, the goal for audiophiles isn't make a well functioning power supply it's to make a nice sounding amplifier.
Great vid. It is good to have visual confirmation and explanation of the conclusions I have drawn after many experiments where adding more capacitance never seemed to have the desired effect. Thanks. +1 switching output cap discussion
Great video..........often ignored subject.
Yes pls do a video such as this for switchers. 👍🏽
Great tutorial, I for one would love more like this...cheers.
This was an excellent video.
Learned something new today! Thanks.
Very interesting. Also suprising to see that having a smaller/lower ESR capacitance is not necessarily better for a linear regulator.
Uhmm it's stated in the datasheet for an reason.
I tend to have been using LDO's for their efficiency and tend to use about 10uf ceramic in parallel with the local and bulk decoupling.
I only know 1000uf's are only required after rectification for smoothing. Other than large fast loads like you mentioned.
Excellent, thanks…
Well, any reactance on the output introduces a pole which in the worst case violates the Nyquist stability criterium. You showed nicely the impact on transient load changes where oscillatory behavior became apparent already. In principle a regulator is as you say just the equivalent to an op-amp. With the proper compensation you can achieve better slew rates than just with the .20dB/dec response. In a power supply you want to enforce stability which as you demonstate can easily be destroyed. Anyway in your drawing looks like that likely by mistake you switched the inputs of the op-amp - thus adding the compensating capacitance to the non-inverting input and output which surely causes instability given its value.
Thanks, yes I swapped the inputs to the op-amp accidentally.
Ótimo vídeo gualidade total parabéns guerreiros bazuka Foz do Iguaçu pr Brasil 🇧🇷🇧🇷🦈🦈🦈🇧🇷🇧🇷
Great Info! Thanks
Nice video , this it is a useful information .
Great video as always
Great episode 👏👏👏👏
What value was the ceramic capacitor you used? The mlcc waveform looked horrible though the conventional wisdom is to put small ceramic capacitors on the output.
I think it was 1 or 10 uF
@@sdgelectronics Ok those are the values of mlcc I usually use on the output of 78xx regulators, I don't have a scope though so I could never check if it was working. Datasheets usually recommend 100nf ceramic on the output maybe 100nf's the largest mlcc can be without interfering with the feedback loop?
Either way I'm going to stick with a 1uf+ tantalum in parallel with a 100n mlcc now hopefully that should be ok.
This topic never gets old. The question is why is there so much misconception?
Great video !!!
ive had the issue several times with a classic 78 79 dual rail supply and opamps that the positive rail wont start and just sits at -0.5V or something
I'm pretty sure I've seen a diode across the output (cathode to + side) of each regulator to stop that.
Good video. For linear regulators, such as 78xx, it is not recommended to put low esr capacitors at the output. This can cause generation. As well as high-capacity capacitors. But if you need a large capacitor capacity in the circuit, and it stands after the linear regulator, then a 4.7-15 ohm resistor must be installed between them.
4:20 this looks like one of that retarded "fever level" audiofool regulator/filter module from Ali ;D
I would have liked to see also a test with datasheet recommended values and cap types
As always, then, follow the spec sheet! I'd be very interested to see exactly the same, but for LDO regulators. How is all this affected by the distributed supply capacitance in a complex circuit with local decoupling on multiple devices? Do we simply rely on the local supply inductance to allow the regulator the freedom to regulate?
I have an electronic/mechanic background, although most of the electronics were plug and play. I am now wanting to build and design my own projects now that I'm retired. Some of this was over my head for where I'm at understanding wise. Could you suggest some videos that may start more with laymen terms? I think you do a great job, just some of it's over my head..
At 5:22 the +- inputs of the op amp are inverted.
My thoughts exactly..
Apologies I missed that
Nice video. Regarding ringing with the ceramic capacitor on the output, I think you said it was due to the low ESR, but to solve it you increased the capacitance as well as using an electrolytic. Could you instead have added some resistance in series with the ceramic capacitor, would that have worked?
You could, but then it loses the high frequency performance. It might be possible to reduce the problem with an actual PCB since the breadboard used in this video does add some parasitics.
Excelent video, very well demonstrated, the capacitor in your diagram, inbetween the opamp output and the non-inverting input, what is it used for. Also could a similar test setup be used to get an estimation of the Phase Margin of the feed back loop by observing the oscilations on the step impulse at the output of the regulator. Thanks again, a great video.
I did the drawing slightly incorrect, swapping the + and - symbol around on the op-amp. The capacitor limits the bandwidth to prevent unstable behaviour at high frequency
@@sdgelectronics Cheers, great video again, keep it up.
I might be wrong but I'm sure XraytonyB had some problem with ringing in a tone circuit when he changed the electrolytics for ceramics and was asking if anyone knew what the reason was for it, is this a related issue as he changed them back and the ringing stopped.
You didn't test the regulator noise under load, which is I think why they do add excessive output capacitors.
Nice video. Is there a general rule on Input:Output cap ratio? - datasheets typically show between 3:1 to 10:1 from memory.
good video
Say if you were Photonicinduction and wanted to blow a 5kA fuse you'd want a large cap bank ROFL!
Without all the output caps the inrush current will have no where to go. ;-) an often forgotten part of regulators is the input ripple rejection within the audio band.
The ESP8266 is a difficult IC. 3.3 volts and a few mA, but high spikes when it connects to the wifi. Not every linear regulator can do this. Usually the 3.3 volt is generated from a 5 volt. That's why I put enough uF in front of the 3.3 volt regulator as a buffer. In the project I had a 24-30 volt supply voltage, from which a stepdown generated 5 volts, then the 3.3 volt linear regulator.
2:40 Can anyone pls explain Q11?
Q11 is the current source(current mirror) for Q15_Q16 Darlington output regulator.
But why the double collectors? Or how does it look like in silicon? PNPP?
@@suncrafterspielt9479
It's another graphic figure(symbol) for representing two transistors with common bases tied to one collector, aka current mirror. It takes less space. Regarding current mirrors, take a look at this:
th-cam.com/video/VnJHXQCPIvs/w-d-xo.html
th-cam.com/video/B9PJsJsK3mk/w-d-xo.html
Ok, now it makes a lot more sense, thank you
The solution is to put a number of different caps on both the input and output lines.
Maybe for the input but I thought part of the message here was that having too much capacitance on the output can be detrimental, so you probably don't want to go overboard with too many different types in parallel there?
Over capasitance on the output is bad and parallel isn't going to fix that.
I use 0.1uf+47uf output of 7805
First. :-)
I'm not the last : )
First comment to sink to the bottom.