I read this on diyAudio. ( Got interested, then confused.)
Take a guy who has decided to provide 20,000 uF per rail. He can do it one of three ways, at fairly similar cost. He can put in one big 20,000 uF cap; or he can parallel a pair of smaller 10,000 uF caps; or, he can put in two 10,000 uF caps in parallel, but separated by a 0.22 ohm resistor. If you do the ESR math, the last approach is about 10 times better at filtering the nasty stuff, although total energy storage and p-p 120 Hz ripple are similar in all three cases.
How can "the last approach be 10 times better at filtering ....although....p-p 120 Hz ripple are similar in all three cases" ?
What ESR math ?
Take a guy who has decided to provide 20,000 uF per rail. He can do it one of three ways, at fairly similar cost. He can put in one big 20,000 uF cap; or he can parallel a pair of smaller 10,000 uF caps; or, he can put in two 10,000 uF caps in parallel, but separated by a 0.22 ohm resistor. If you do the ESR math, the last approach is about 10 times better at filtering the nasty stuff, although total energy storage and p-p 120 Hz ripple are similar in all three cases.
How can "the last approach be 10 times better at filtering ....although....p-p 120 Hz ripple are similar in all three cases" ?
What ESR math ?
Turn it into a homework problem:
Find a definition of "THE NASTY STUFF" such that bridge->cap1->0.22R->cap2 does indeed exhibit 10x better rejection of "THE NASTY STUFF", than bridge->(cap1 parallel cap2) . Or, prove that there is no such definition; prove that it is impossible for each and every definition of "THE NASTY STUFF"
Find a definition of "THE NASTY STUFF" such that bridge->cap1->0.22R->cap2 does indeed exhibit 10x better rejection of "THE NASTY STUFF", than bridge->(cap1 parallel cap2) . Or, prove that there is no such definition; prove that it is impossible for each and every definition of "THE NASTY STUFF"
As this is DIY audio, I will try not to spoil all the fun by doing the work for you. I highly recommend downloading LTSpice and running the analysis there for yourself. It's pretty easy to do and pretty fun. There are some youTube videos out there that can help you. There is an example amplifier schematic in the LTSPice folder as well as examples for different analysis. For instance you could add your power supply to the example amplifier (40V AC power with some series resistance to model the transformer, rectifiers, capacitors ) and look at the noise at the output with the transient analysis. You can put in the capacitors and resistor and then look at the transfer function to see the attenuation. For comparison a coil between the capacitors forms what's called a pi filter because the schematic looks a bit like the pi symbol. With some values of the coil and capacitor tuned to reject the noise it can be very effective. Depending on what type of amplifier is being powered, there will be some amount of power supply noise rejection. So that is an important thing to look at in determination of how much power supply ripple can be tolerated.
I just want to know wether this makes sense ? "10 times better" although "similar" 🤣
Do you have answers to my question in the OP.
I am well aware of maths and Spice simulation.
Do you have answers to my question in the OP.
I am well aware of maths and Spice simulation.
Doesn't "10 times better" refer to an improved high frequency noise reduction?
Quite a separate issue to ripple reduction which could be similar.
The question may be how do you calculate the cut off frequency for a low pass pi filter in which a resistor is used in place of an inductor?
Quite a separate issue to ripple reduction which could be similar.
The question may be how do you calculate the cut off frequency for a low pass pi filter in which a resistor is used in place of an inductor?
It appears to be purely theoretical, not a real life problem encountered in a Do-It-Yourself project already underway. So I conclude there is no urgency. Therefore I choose to wait seven or eight days before posting my opinions and "answer" . . . . if it hasn't already been posted by someone else. Like many here, I am aware of maths and Spice simulation.
RC filters work. It makes sense. Getting a specific answer as to how well it works at what frequency and if it would make any difference to a specific amplifier requires a calculation. So I can do it or you can do it.
F3 = 1 / ( 2 * pi * R * C) filter cut off frequency
1 / (2*3.14*0.22 * 10000e-6) = 72 Hz for that 0.22 ohm, 10,000 uF capacitor
It will attenuate noise 3dB starting at 72 Hz and the attenuation will increase 20 dB/decade
with rising frequency.
If you want better performance, use a coil and get 40 dB / octave roll off.
Investing a bit of time you could learn to use LTSpice to DIY . A simple back of the envelope calculation would be to get a 10:1 ratio of voltage you would need a 10:1 ratio of impedance. If you had Capacitors with ESR of 0.022 and you then put ten times that resistance between two of them, as mentioned 0.22 ohms, ( zero math wildly wrong approximation ) I could imagine getting the "10 times better" or specifically one tenth the noise at the output of the filter referred to the input at some frequency.
Last edited:
Ideally, you would also have to take into account the loading: when the 0.22 resistor is added, the output of the first cap sees a larger impedance compared to 10mF, and this will somewhat increase the ripple.
The net effect will still be beneficial, but less than you might expect.
Of course, matters improve as the frequency increases
The net effect will still be beneficial, but less than you might expect.
Of course, matters improve as the frequency increases
Don’t you have to consider the diodes and transformer winding resistances ?
May be they come in to swamp the expected benefit.
May be they come in to swamp the expected benefit.
Yes you also have to consider the transformer winding resistance. But if there is any 120Hz ripple rejection because of the winding resistance, you simply are lowering you DC voltage. Because of the internal resistance. From whatever perspective you look at it, it is a bad thing for a power supply.
Adding of the 0.22 ohms resistor has the same effect. The Ri of the power supply gets worse. Drawing more current will decrease the output voltage. With 0.22 ohms that is not dramatic at all. Drawing 10 A give you a voltage decrease of 2.2V. Which is still good for practical purposes.
OTOH previous replies indicated that this filter won't do much with a Fc of 72 Hz.
So in order to do anything, the resistor should be 10x as high. 2.2 Ohms. But then you experience a voltage drop of 22V (!) at spikes of 10A. And that is not acceptable. Let alone that in a stereo amplifier the other channel will suffer severely by this high Ri.
A little bit of 120Hz ripple is not that serious. Power amplifiers have a CMRR which reject much of the power supply ripple. AND at low output power there hardly is any ripple because no current is drawn. Once you draw high current at high output power the ripple increases. But any audible effects are totally masked by the high output level.
Adding of the 0.22 ohms resistor has the same effect. The Ri of the power supply gets worse. Drawing more current will decrease the output voltage. With 0.22 ohms that is not dramatic at all. Drawing 10 A give you a voltage decrease of 2.2V. Which is still good for practical purposes.
OTOH previous replies indicated that this filter won't do much with a Fc of 72 Hz.
So in order to do anything, the resistor should be 10x as high. 2.2 Ohms. But then you experience a voltage drop of 22V (!) at spikes of 10A. And that is not acceptable. Let alone that in a stereo amplifier the other channel will suffer severely by this high Ri.
A little bit of 120Hz ripple is not that serious. Power amplifiers have a CMRR which reject much of the power supply ripple. AND at low output power there hardly is any ripple because no current is drawn. Once you draw high current at high output power the ripple increases. But any audible effects are totally masked by the high output level.
Like I said.
If you want a complete answer it is easy to get. Just model the circuit, including the power station if you like, using LTSpice and you will have all the effects including the switching of the diodes and the PSR of the amplifier if you throw that into circuit model. All the by gosh and by golly you didn't include X, Y or Z can be handled in one tiny LTSpice model.
You can measure the self resonant frequency of a ten thousand microfarad filter capacitor, as specifically called for in post #1 of this thread. The easy way is to use an LCR meter with continuously adjustable test frequency, like the ET4510. While monitoring the magnitude of Impedance, sweep frequency. Note the frequency at which Impedance is minimum; that's the S.R.F.
Another way to measure S.R.F. is discussed by YouTube teacher W2AEW, in the video below.
Once you know the S.R.F. you are able to calculate the self inductance of the capacitor, just working backwards through the resonance equation: (1 / omega) = SQRT(L * C)
Then the LCR meter will measure the capacitor's ESR and boom, yer done. Victory. You've got L, you've got R, you've got C. Now you can include all of these effects in your circuit simulation runs.
Another way to measure S.R.F. is discussed by YouTube teacher W2AEW, in the video below.
Once you know the S.R.F. you are able to calculate the self inductance of the capacitor, just working backwards through the resonance equation: (1 / omega) = SQRT(L * C)
Then the LCR meter will measure the capacitor's ESR and boom, yer done. Victory. You've got L, you've got R, you've got C. Now you can include all of these effects in your circuit simulation runs.
You seem to be the person proposing the '10 times better' comment, and to have done the ESR math to support the claim ?? If not, then perhaps provide the reference to where you read the comment, to give a better context to why the comment may have been made.
Do you appreciate the likely impedance spectrum of a 10,000uF (of unknown voltage rating) e-cap, and seen any such spectrum plots in e-cap datasheets/app guides? The impedance of such a large e-cap is certainly not the same as that of say a small ceramic cap as tested in the youtube link in post #12.
Do you have any more definition to what you mean by 'nasty stuff', and were you referring to nasty stuff originating from the mains AC and/or the rectification process ?