Bob Cordell Interview: Power Supplies

Status
Not open for further replies.
The mean voltage is higher, the larger cap discharges more slowly as well as charging more slowly, but the overall ripple is less and the mean voltage less variant -- this example shows a switched load of 1 to 4 amps -- simplified just to show the effect of the capacitor and not energy stored in a transformer core or inductor:

An externally hosted image should be here but it was not working when we last tested it.
 
Bob Cordell said:
Note that a second level of main power supply RC filtering, say 10,000 uF and 0.22 ohms, will have a corner frequency of 60 Hz, meaning that the nasty fundamental and harmonics of the sawtooth will be attenuated by 6 dB starting at 120 Hz and rolling off at 6 dB/octave after that if the 10,000 uF cap has a good ESR. This will really round off the sawtooth wave. Note, that any kind of power supply rejection tends to decrease at high frequencies, so the increased rolloff of the RC filter can really help. We'll have 20 dB of attenuation from this additional filtering stage at 600 Hz. This is also the capacitor that likes to be right near the output stage power transistors to close the current supply loop locally. The presence of the 0.22 ohm resistor will cause only a fairly negligible reduction in available supply voltage.
This reduction of high freq content on the supply lines seems like a great goal to achieve, but I didn't understand why you'd need a second entire cap bank and resistor. Can't the same goal be achieved by adding one small high-freq cap, say of 0.1uF, after the main cap bank? Won't such a cap just short all high freq components to ground?

I'm new to amplifiers, so please excuse any lack of understanding. 🙂
 
Hi Tcpip,
The capacitor following the resistance of the transformer forms a low pass filter. One can calculate the -3db turn over frequency of this filter from F=1/(2 PiRC).
If you add 0.1uF to 10,000uF you get 10,000.1uF. The filter frequency has hardly changed.

If one were to cascade two filters RCRC then the first and second filters can be calculated independantly and the effects of cascading the two filter stages can be added together.

To have a significant effect in removing 50Hz or 60Hz the F-3db needs to be down near the mains frequency, or better still, well below Mains F. That demands large C if one wants to use a small R.

If, as Bob has suggested, one were more interested in removing the harmonics then a smaller C (or smaller R) is acceptable, but the bigger the RC time constant the more attenuation one can achieve.
 
Re: PSU Snubbers

gpapag said:
Spectrum Analyser
With just a cheap scope it is often possible to zoom in sufficiently to the location on the AC wave where diodes turn off and see the ringing, then try different snubber values to remove it.

tcpip said:
A bit complicated; I'm still trying to understand it.
It's actually simpler, since you only have one snubber across the secondary, instead of across each diode. The only thing complicated is the calculating the correct values (or rather, the measurements needed as inputs to the calculation). However, that would apply just as much if you had snubbers across each rectifier. Using generic values for various power supplies without checking the result on the scope is in my (limited) experience not the best way to proceed.
 
Re: PSU Snubbers

Nixie said:
.... The only thing complicated is the calculating the correct values (or rather, the measurements needed as inputs to the calculation). However, that would apply just as much if you had snubbers across each rectifier. Using generic values for various power supplies without checking the result on the scope is in my (limited) experience not the best way to proceed.
I decided to try this without measurement or calculation and found that a bad snubber choice exaggerated an instability problem in a bad output Zobel capacitor.
 
Well, i have finished to read the Bob Cordell Interview thread.

The real gurú.... really you are great Bob.

There are many "stunts"...some people searching for fame and others with real doubts...they had discussed a lot, and made the tread loss its rithm....well, they have their own doubts...but i found more pleasant and usefull to jump all them and read only your comments.

Hey guys!...let the man talk!

Thank you to share your knowledge.

Carlos
 
lumanauw said:
Snubbers usually needs 2 to 5watt resistor, they are hot. The values can be approximated by calculation first (or more easily, look at other schematic, what value do people use). But the final tuning needs scope, it is impossible to do it without scope.

The power rating of the snubber resistor is a function of its resistance and the frequency and voltage across it. The frequency in this case is the ringing frequency of the CL circuit formed by the diode and the transformer.

I had thought that one snubber across the secondary was going to work -- but found in real life that each diode had to be snubbed -- I am going to check my work to make sure, however, wouldn't want to have folks needlessly led astray.
 
janneman said:
...For me there really isn't any advantage above 20kuF or so, and the disadvantage (and space and cost!) start to come into play.

Jan Didden

agent.5 said:
...I did not try to go to either extreme of no capacitance or so much capacitance that only NASA can afford to experiment with such setup...


In all my tries, increasing capacitor never gives better sounding. My 'religion' would to suppress capacitors. Understand, minimize negative effects of big capacitors. To do this I introduce the 'Shared current power supply' (SCPS) concept.
I explain it on my web site http://tech.juaneda.com/en/articles/powersupply.html

I think best sounding is obtain when amplifier is supplied by transformer, not by capacitors.

Eric Juaneda
 
Eric Juaneda said:





In all my tries, increasing capacitor never gives better sounding. My 'religion' would to suppress capacitors. Understand, minimize negative effects of big capacitors. To do this I introduce the 'Shared current power supply' (SCPS) concept.
I explain it on my web site http://tech.juaneda.com/en/articles/powersupply.html

I think best sounding is obtain when amplifier is supplied by transformer, not by capacitors.

Eric Juaneda

Big capacitors can be highly inductive -- this is the problem that you have to tackle -- that big cap becomes "detached" from ground as frequency rises. this is where it is helpful to have a network analyzer to look at the actual impedance of the supply as you stress the amplifier -- try bypassing a big cap with two polypropylenes -- one of which is snubbed with a few ohms resistance -- (you can sim this in LT Spice and look at the bode plot.) The two small caps filter high frequency noise and the resistor broadens the Q of the resonant circuit.

As Andrew states - a bad bypass or snubber can make a circuit unstable -- you can form a resonant circuit with a narrow peak in response if you aren't careful.

EDIT -- again -- look at what happens when an amplifier is near clipping at high frequencies -- if you have a good spectrum analyzer you see a "blossoming" of high frequency harmonics -- being high frequency (compared to audio) they propagate easily and can cause all sorts of headaches with semiconductor circuitry.
 
AndrewT said:
The capacitor following the resistance of the transformer forms a low pass filter. One can calculate the -3db turn over frequency of this filter from F=1/(2 PiRC).
If you add 0.1uF to 10,000uF you get 10,000.1uF. The filter frequency has hardly changed.
I can understand this math, but I'm sure there's more to it than that. If it was just a case of calculating the equivalent capacitance by adding the nominal values of each device, then one would never need to bypass an electrolytic with a ceramic, right? I believe this bypassing of one cap with another is done because the large and inexpensive electrolytics do not have high capacitance at high frequencies, hence you need to parallel them with very low-inductance low-ESR small-value caps in order to ensure that the filtering remains effective even at MHz frequencies.

I am trying to achieve just the same effect by adding a small-value ceramic after the bank of high-value electrolytics.

To have a significant effect in removing 50Hz or 60Hz the F-3db needs to be down near the mains frequency, or better still, well below Mains F. That demands large C if one wants to use a small R.

If, as Bob has suggested, one were more interested in removing the harmonics then a smaller C (or smaller R) is acceptable, but the bigger the RC time constant the more attenuation one can achieve.
Yes, this part makes sense. I am quite clear that Bob's approach will work. I was wondering why mine wouldn't.
 
Eric Juaneda said:
In all my tries, increasing capacitor never gives better sounding. My 'religion' would to suppress capacitors.
Understand, minimize negative effects of big capacitors.
To do this I introduce the 'Shared current power supply' (SCPS) concept.
I explain it on my web site http://tech.juaneda.com/en/articles/powersupply.html
I think best sounding is obtain when amplifier is supplied by transformer, not by capacitors.
Eric Juaneda

Dear Mr Juaneda,

I have to say that I am extremely impressed by your revolutionizing theory.
This is an extremely important issue for me that of the needed capacitance in a amp's power supply.
First of all I have to say that I am everything but an expert on electronics, but I have always tried to understand the basics to make some good choice in equipment selection.
Up to today, I have always thought that the more capacitance the better (and of course quality of capacitance, low esr and so on, were also important).
Are you saying that it is possible to avoid completely the caps in the power supply of an amp ?
In this case I would be wiiling to mod some amps of mine replacing the existing transformer with a very big new one and eliminate any caps after the diode bridge.
Of course I looked at your excellent web site but a lot of technical talk are too difficult for me.

Thank you very much indeed.
Kind regards,

beppe
 
tonyptony said:
Bob, I've heard though that the boutique ultra-fast, "soft recovery" diodes should not necessarily have snubbers added. Something about the way they operate already mitigating a lot of the radiated stuff.


I'm sure that those diodes improve matters, but I've never heard that the addition of snubbers would not improve matters further still. Keep in mind that no matter how fast and perfect the diode is, quite high currents will still be switched on and off very quickly as a fundamental part of the rectification process. The switching currents are inevitably taking place in a circuit environment that contains inductance and capacitance that can resonate and radiate. The snubber very strongly damps these resonant circuits, especially if an RC snubber is used.

Cheers,
Bob
 
AndrewT said:
Hi Tcpip,
The capacitor following the resistance of the transformer forms a low pass filter. One can calculate the -3db turn over frequency of this filter from F=1/(2 PiRC).
If you add 0.1uF to 10,000uF you get 10,000.1uF. The filter frequency has hardly changed.

If one were to cascade two filters RCRC then the first and second filters can be calculated independantly and the effects of cascading the two filter stages can be added together.

To have a significant effect in removing 50Hz or 60Hz the F-3db needs to be down near the mains frequency, or better still, well below Mains F. That demands large C if one wants to use a small R.

If, as Bob has suggested, one were more interested in removing the harmonics then a smaller C (or smaller R) is acceptable, but the bigger the RC time constant the more attenuation one can achieve.


Well-stated, Andrew. Thanks.

Bob
 
Re: Well, i have finished to read the Bob Cordell Interview thread.

destroyer X said:

The real gurú.... really you are great Bob.

There are many "stunts"...some people searching for fame and others with real doubts...they had discussed a lot, and made the tread loss its rithm....well, they have their own doubts...but i found more pleasant and usefull to jump all them and read only your comments.

Hey guys!...let the man talk!

Thank you to share your knowledge.

Carlos


Thanks for your kind words, Carlos. There are many very smart people here, and my interactions with all of you has taught me many valuable things as well.

Cheers,
Bob
 
Eric Juaneda said:
.....the 'Shared current power supply' (SCPS) concept.
I explain it on my web site http://tech.juaneda.com/en/articles/powersupply.html

I think best sounding is obtain when amplifier is supplied by transformer, not by capacitors.
Hi Eric,
just had a look.
If I bring my three phase (415Vac) into my listening room, how do I wind a three phase toroid?

I would like you to tell me if you have solved the excessive noise problem of the half wave rectified, no capacitor supply to your amplifiers yet?

I would like you to describe the sound of an amplifier with ZERO volts on the supply line when the no capacitor half wave rectified supply is between supply on periods.
 
Hi all
Reffering to Eric Juaneda's use of much oversized x-formers, a lot of others corresponders in many threads have reported improvement of sound.
My own experience aligns with Eric: the higher the VA of the x-former on a given circuit, the more the subjective improvement in sound.
Why this happens I don't know.
Rod Elliot has a three part article on transformers
here
trying to explain things.
On the other hand, when it comes to PSU capacitors, brute force does not provide satisfactory results for me. For low power circuits, I am more happy using a pre-regulator followed by the end regulator, low capacitance overall. But this is something you all know

Regards
George
 
tcpip said:

This reduction of high freq content on the supply lines seems like a great goal to achieve, but I didn't understand why you'd need a second entire cap bank and resistor. Can't the same goal be achieved by adding one small high-freq cap, say of 0.1uF, after the main cap bank? Won't such a cap just short all high freq components to ground?

The same goal cannot be achieved by substituting RCRC with RCc.
They do something different.

For the sake of simplicity (for me as well as maybe you) I shall assume that the Big C only works at lower frequencies and the little c only works at higher frequencies without any overlap.

Do the lower frequency calculation and determine the attenuation rate of the RC part of the filter.

Now do the same at higher frequency for the Rc filter and calculate the attenuation rate.

The little c is very effective at very high frequencies making it very good as a bypass (next to the device that consumes the current and CHANGES current requirement rapidly) capacitor.
The big C will still do it's best to supply near DC to the rails and a little c at the smoothing end will help to attenuate the very high frequencies, but it does almost nothing to attenuate the harmonics of the mains frequency, even upto the tenth or the hundreth (about 5000 or 6000Hz) Assume the transformer R=0r1 and c=100nF, F-3db=16MHz.

The little c would be better placed on the primary side (where the long supply cables provide some inductance) and across the secondary, then RC snubbers across the diodes and if CarlosFM is right then RC snubber across the PSU output pads

BTW,
I did it your way 30years ago in total ignorance, using a mix of HF electro//PC//PS and fortunately had an amp with a good PSRR to hide my mistake.
 
Please excuse me Sirs, but my question is very basic.
Can a power amp work without smoothing caps in the power supply?
I mean just a very big transformer and a diode bridge after it ?
Have anyone tried this ?
It seems not possible to me. Never heard of such a PS.
Thanks and regards,

beppe
 
Status
Not open for further replies.