Fotios need help with power supply smoothing capacitors.

[fotios]

In this part I am addressed to persons that have good experience with power supplies. Even if it exists already a permanent thread of Bob Cordell around from this subject, I apologize that I open one separate thread; however it should you occupy also my own place, that I cannot read one so much long thread in that probably are found the answers that I ask. This subject concerns the relation between the power of supply transformer and the suitable value of the smoothing capacitors. I have read enough posts in the Bob Cordell thread however I did not find a satisfactory answer in my thoughts.
Suppose that we have a transformer of 250 VA with a split secondary of 2 X 42,5 Vac a usually used value due to standard 63Vdc capacitors. A good amplifier supplied from a such type energy source can produces hypothetically an output power: P = V²/R = 42,5²/8 = 225,78Wrms in a load of 8Ù. In practice the average output power in whole spectrum it does not exceed the 170Wrms/8Ù. Of course in higher frequencies it touches the peak value of 225,78Wrms indeed. My query is the following: How much is enough the capacitance value of smoothing capacitors in this case of the 250VA transformer. I have made experiments certainly in order to find the minimal possible capacitance who serves satisfactorily this power level. The only reliable test that I could think is the injecting of square waves of 50% duty cycle between 20 to 100 Hz in the input of amplifier and the observation with a scope of its output how much is decreased the tilt of the duty portion of square according to the increase of capacitance. My conclusions are that further increase from 5600ìF does not give nothing more in the reduction of the output square tilt. An other ascertainment that I made was that the particular transformer even if it was special order with enamelled core and with extra enamelled windings and finally fixed in a metal case with epoxy resin and before it is used remained for one month in the shelf so that dry well all the resins (I wonder what can becomes more from them) during the demanding of peak power under 100Hz was heard the light cracking of his core to my big disappointment. After this I decided to make the same experiment supplying however this time the amplifier from my 1KVA laboratory power supply in order to is dissolved any doubt it had remained to me. It should also I mark that the particular amplifier it is of class AB it has an input capacitor of 22ìF and a feedback AC coupling capacitor of 1000ìF. But also this time I ascertained the same thing precisely; that or with 250VA transformer, or with 1KVA transformer it does not become further reduction of tilt with capacitors above 5600ìF. With other words is not increased the supply of energy from the amplifier to the load. My speculation of course does not have relation with the cause of this phenomenon. He is simple and has relation with the expenses of buying big capacitors; for what reason? I have tried also except Panasonic or Nichicon caps the better quality Aerovox with the smallest resistance it exists, or a scheme with 2 or 4 paralleled caps instead one to reduce furthermore the resistance. The result was the same. I would want too much to read the comments and your experiences and also if it escapes from my attention something about this subject. I am very distressed with this subject and I do not know what to make. It should I place capacitors of 15000 to 22000 ìF in order to cause impression or place those that showed my experiments and be the truth? I request you however write something that would have value at your crisis and in well formulated phrases because many times I do not bring out the inner meaning from what you write when you write hastily. I quote bellow some schematics in order to you occupy better my thoughts.



In the first row at top it is the input signal and in the bottom it is the output signal. In the second row at the top and the bottom are the same signals represented in the S-plain. In the third row the shaded areas are the energy which cannot give the amplifier in his output as it follows the input swing.

[Cloth Ears]

Ù is ohms, correct?
ìF is micro Farads, correct?
For class AB amplifier, correct?

Nelson Pass (in his Power Supplies: Commentary For Consumers article) indicates enough capacitance to reduce the ripple current to 0.6v. This for a class A amplifier, mind you.
Dejan Veselinovic (in his Solid State Power Amplifier Supply (Part 3) article on the TNT-Audio site) indicates 40,000uF/ch for a 300VA transformer. Dejan also mentions, elsewhere, that a single Sikorel is 'better' (my word) that multiple lesser capacitors.
Rod Elliot (in his Linear Power Supply Design article) provides a formula which works out at around 10,000/ch for the figures you quoted. Rod indicates (I think in this article) that multiple smaller value capacitors are better than a single larger item (both cost and benefits).

Originally posted by fotios
It should I place capacitors of 15000 to 22000 ìF in order to cause impression or place those that showed my experiments and be the truth?

So, it appears there is no "truth", only variations based on opinion. One thing that most seem to agree on is that capacitors are no substitute for a "too small" transformer. Over-specifying both your transformer and your smoothing capacitors will not hurt - but under-specifying may.

Is this the type of wishy-washy answer that you were looking for?

[flg]

Originally posted by fotios
The only reliable test that I could think is the injecting of square waves of 50% duty cycle between 20 to 100 Hz in the input of amplifier and the observation with a scope of its output how much is decreased the tilt of the duty portion of square according to the increase of capacitance. My conclusions are that further increase from 5600ìF does not give nothing more in the reduction of the output square tilt.

fotios, I use this test also. Very simple quick check for problems
However, I have seen the "distortion" you speak of before. Would you have been doing your tests on an amp with a capacitance coupled output? If that is the case, the coupling caps in the output are the limiting factor not the power supply caps...

[anatech]

Hi fotios,
Your experiment lines up with what I see.

It depends on how sensitive your amplifier is to supply ripple (as in class A types) for one. Then simply your maximum output current plus some to cover capacitor tolerances and age. A few capacitors in parallel are better than one big one on average, but at 5,600 uF (design spec maybe 6,800 uF), you just need to bypass it.

I see real problems in stuffing really large capacitors into normal style class AB amplifiers. The same holds true for tube equipment. Supplies can be handled with more finesse than brute force and turn out to be better.

-Chris

[fotios]

Originally posted by flg


fotios, I use this test also. Very simple quick check for problems
However, I have seen the "distortion" you speak of before. Would you have been doing your tests on an amp with a capacitance coupled output? If that is the case, the coupling caps in the output are the limiting factor not the power supply caps...

Hi FLG
No i not use capacitor in the output. I forgot to refer that the amplifier it is full complementary from input to output so there is not the need od a A.C. coupling capacitor in the output. If i not understand right that you wrote, tell me further.
Thanks
Fotios

[fotios]

Originally posted by Cloth Ears
Ù is ohms, correct?
ìF is micro Farads, correct?
For class AB amplifier, correct?

Nelson Pass (in his Power Supplies: Commentary For Consumers article) indicates enough capacitance to reduce the ripple current to 0.6v. This for a class A amplifier, mind you.
Dejan Veselinovic (in his Solid State Power Amplifier Supply (Part 3) article on the TNT-Audio site) indicates 40,000uF/ch for a 300VA transformer. Dejan also mentions, elsewhere, that a single Sikorel is 'better' (my word) that multiple lesser capacitors.
Rod Elliot (in his Linear Power Supply Design article) provides a formula which works out at around 10,000/ch for the figures you quoted. Rod indicates (I think in this article) that multiple smaller value capacitors are better than a single larger item (both cost and benefits).

So, it appears there is no "truth", only variations based on opinion. One thing that most seem to agree on is that capacitors are no substitute for a "too small" transformer. Over-specifying both your transformer and your smoothing capacitors will not hurt - but under-specifying may.

Is this the type of wishy-washy answer that you were looking for?

Hi Cloth Ears
Indeed it is such by some way ( wishy-washy ) if i understand rightly the mean of this phrase. The difference in price between a 5600mF and a 10000mF cap (snap type) it is only 0,7 Euros in my place. Of course i am not stupid and i use a 10000mF pair for each channel. What is the difference? 2,8 Euros only for the total supply. Also the difference in price between a 250VA and a 300VA toroid it is only 5 Euros. The cost it is increased significantly only from the extra works from the standard offered (resin dipping etc). And for the absolute ripple rejection the main work it is performed from the capacitors of 100 to 220mF mounted in the PCB of main amplifier close to output transistors as i know and i have experimented. My speculation it is about from the consumerinsm caused thought of people. How many of them thinked the role of the power of the transformer, the role of bridging rectifier and the role of the shared capacitance of the cables of voltage rails?
Thanks for the reply
Fotios

[Nordic]

did you just swap out diffirent value capacitors, or did you also try parallel pairs to see what the lower ESR does...?

Regards
Nico

[fotios]

Originally posted by anatech
Hi fotios,
Your experiment lines up with what I see.

It depends on how sensitive your amplifier is to supply ripple (as in class A types) for one. Then simply your maximum output current plus some to cover capacitor tolerances and age. A few capacitors in parallel are better than one big one on average, but at 5,600 uF (design spec maybe 6,800 uF), you just need to bypass it.

I see real problems in stuffing really large capacitors into normal style class AB amplifiers. The same holds true for tube equipment. Supplies can be handled with more finesse than brute force and turn out to be better.

-Chris

Hi Chris
You don't believe that i use 5600mF capacitors. You know me from previous discussions.
Fotios

[fotios]

Originally posted by Nordic
did you just swap out diffirent value capacitors, or did you also try parallel pairs to see what the lower ESR does...?

Regards
Nico

Hi Nordic
I have tried also the two cases.
Thanks
Fotios

[anatech]

Hi Fotios,

You don't believe that i use 5600mF capacitors. You know me from previous discussions.

True. But guess what? I keep my supplies in line with what is needed, as I stated in my post. I will use slightly higher if that is all I have.

You do need to consider the effects of reducing your conduction angle by increasing cap values. You may be creating problems where there were none.

-Chris