Power Supply Resevoir Size

[Terry Given]

Daniel @ #1213

you are right re. the cap ESR, but as Frank points out above, ESR varies strongly with time, temperature and sock colour (among other factors). its a commonly used trick, but because of this variation its not a very robust one. look at an electrolytic ESR for example - good cap data sheets specify ESR at 20C and at -10C - this is because ESR rises sharply at low temperatures. this can (and does) cause stability and even startup problems with regulators that rely on the ESR*C zero for stability.

A much more robust, predictable method is to use very low ESR caps, then add your own series resistor to get the desired zero/damping. pay close attention to peak pulse power dissipation though - not all resistors are created equal.

This specifically includes wire-wound resistors - quality WW resistors pack the wound mandrel with potting material before placing it in the ceramic body & encapsulating it; cheap nasty WW resistors drop the wound mandrel into the ceramic body and cover over the hole. the result is plenty of air around the resistive element, which hugely decreases its peak power handling capability - I have seen pretty orange flashes of light coming from the resistive element when given a good thump, before the resistor fails. Vitrohm for example dont do this, but all the WW resistors I can get from hobbyist stores in NZ are made like this. you can check a WW resistor by smashing it open with a hammer, providing hours of entertainment and making a mess

there are also some seriously good SMT resistors with astonishing peak pulse power ratings - Panasonic ERJP and Vishay CRCW-HP are awesome.

[gootee]

Originally Posted by gootee

Here is some more data, the 100W case to go with the 25W, 50W, and 75W cases in post # 1175, which is at Power Supply Resevoir Size .

I also attached the spreadsheet with all of this. Just remove the .txt from the filename after (or while) downloading it.

It looks like "2500 uF/Amp", using RMS Amps into the load at 2X the maximum rated output power level, would be more than sufficient, for these four 8-Ohm cases, for which the transformer is well over-sized. That gives about 5% as the maximum drop in rail voltage (the drop is relative to its average level), in each case, and about 0.005% average RMS error, in each case, in the square wave that draws 2X the rated power. But you could go as low as 1600-1800 uF per amp, and even lower when rated output power is less than 100W. But for higher than 100W rated output power with the same transformer ratings, more uF/amp would probably be required. I will try to run some other cases, as I get time.

Nico,

The "uF per Amp" for 1X the rated output power (sine), instead of 2X the rated output power (square), would be √2 times the ones I gave using the 2X cases.

So instead of 2500uF/Amp when considering the max load current at 2X the rated power, it would be about 3500uF/Amp.

Or, you could go as low as 1600-1800uF/Amp (using load current at 2X rated power) which translates to 2200-2500 uF/Amp, using the load current for the rated max output power.

The output currents for the 25W, 50W, 75W, and 100W 8-Ohm cases would be √(P/R) = 1.77A, 2.5A, 3A, and 3.5A.

Using 2200 uF/A, their reservoir capacitances could be 3900uF, 5500uF, 6600uF, and 7700uF, which should give roughly 15% worst-case ripple and 0.0075% Avg RMS Error for a square wave at 25 Hz, at 2X the rated output power (i.e. w/same peak output voltage as a sine at 1X the rated output power).

Using 3500 uF/A, their reservoir capacitances could be 6200uF, 8750uF, 10500uF, and 12250uF, which should give roughly 10% worst-case ripple and 0.005% to 0.0055% Avg RMS Error for a square wave at 25 Hz, at 2X the rated output power (i.e. w/same peak output voltage as a sine at 1X the rated output power).

For output power ratings of less than 100 Watts, less capacitance could be used, but the ripple and distortion percentages would then be higher.

Using the same "uF per Amp" factor keeps the ripple percentage and the output distortion percentage about the same, for any output power rating, unless the transformer is too small.

But again, for the 100W case, if the transformer is rated at less than 44-0-44 Volts, or less than 480 VA (240 VA per secondary), then 2200-2500 uF per Amp might not be sufficient. (But, assuming the transformer ratings were "reasonable", it probably would be sufficient, for most people, most of the time, unless they want to run the amp at its max rated power with a guarantee of absolutely zero clipping.)

I will try to run the 100 Watt case with a series of transformer sizes, to see what effects that will have.

It seems that choosing the transformer VA and Vrms ratings is at least as important as choosing the total reservoir capacitance.

It would be nice to have a simple formula relating Vrms, VA, Pout, Rload, and Creservoir, that would also assume at least a bare minimum of 3V to 6V for the amplifier, between the bottom of the worst-case rail dip voltage and the top (peak output voltage) of the load.

Maybe someone can figure out (or already knows) how to add the VA and Vrms limitations to what we already have for Pout, Rload, and Cres.

Cheers,

Tom

[OnAudio]

Originally Posted by Terry Given

I've been playing around with a conceptual SYMEF layout, and making the local decoupling on the amp look like this really is difficult - the layouts that Harrison (OnAudio) has done are pretty damn good, given all the constraints - they are much better than all the audio amp layouts I have looked at to date.

Thank you Terry for your kind words, I am humbled, I am learning . Looking forward to your layout

[OnAudio]

Originally Posted by gootee

Enjoy! (That reminds me... there's still some beer in my refrigerator.)

But how do you have time to watch football?

We won .

So as not to grow old too fast, do many activities, make many friends , take it easy

[OnAudio]

Heres the tickle

We have many measurements we can make including TIM, PSRR, THD, Slew rate, Dumping, square wave tests etc. However the question remains whether we interpret their importance
correctly and whether they are actually valid for what we are trying to achieve. Lets have fun. We have the formulas, are they valid ? What do the speakers and ears think ?

[danielwritesbac]

Originally Posted by gootee

I also attached the spreadsheet with all of this.

Thanks!! Very effective and informative!

Originally Posted by Terry Given

Daniel @ #1213
you are right re. the cap ESR, but as Frank points out above, ESR varies strongly with time, temperature and sock colour (among other factors).

Well, I mentioned that it wasn't as good as the diode trick (last seen in post 1115). However, in all other cases where I've been able to add additional huge caps to the amp board, without dull audio consequences, the preexisting amp decoupling caps were superb signal grade (good enough to use for NFB/Input cap), in the range of 100u to 330u and the Dick Smith ESR Meter regards the amp decoupling caps as a dead short to AC. However, instead of relying on specific model numbers of especially chosen capacitors that get discontinued without notice, I gave the example of 100u//100u//100u low esr because that is similar to a perfect cap approach without actually requiring totally perfect capacitors. That example doesn't utterly block variety, but looks more like a stumble in the right direction.
EDIT:
The length of the power supply interlink cable also has an effect.
Shortest possible = dullest possible with most bass
Very long = clearest possible with least bass.
However,
This is not a lot different than changing the size of the amp board's decoupling caps, smaller for a combination of more forwards and clearer -or- larger for a combination of duller and more laid back. Tuning this size is a really common sport with chip amplifiers. . .
AND, as previously mentioned, results on ideal amp board decoupling cap size all differ somewhat depending on power supply interlink/umbilical cable length.

Originally Posted by Terry Given

A much more robust, predictable method is to use very low ESR caps, then add your own series resistor to get the desired zero/damping. pay close attention to peak pulse power dissipation though - not all resistors are created equal. . .

Thanks! A more solid answer would be great; but, I couldn't envision it, so I'm still in the dark on how to do your series resistors with low ESR caps for amp decoupling example. Sorry to ask, but got a sketch?

Originally Posted by Terry Given

There are also some seriously good SMT resistors with astonishing peak pulse power ratings - Panasonic ERJP and Vishay CRCW-HP are awesome.

This must be the kind of resistor that the Chinese use for burning down cars, boats and RV's with their popular 1156 LED refit bulbs. Those have 3 of 3v diode (9v normal, 10v when maxed) and the rest of the 13.5 to 14.8 or 18v (the common half wave charger) ends up in a very small resistor that is durable enough to damage everything except the surprisingly sturdy SMD resistor.

These 3v ingan really go berserk in the presence of noise while there's no serious overheat with cleaner power. Well, there's a lot nicer uses for this instead of burning down RV's. For example, bleeder resistors that attack more noise than clean power (or attacking almost only noise if adding a stout zener), or so much like riding lawnmower races, the t-amp loudness wars in my signature below.

[Terry Given]

Daniel,

stick the resistor in series with the low-ESR capacitor

----[external resistor]-----[capacitor with low ESR]----

gives RC "damper" with C = C_capacitor, R = R_external + ESR

the idea is to use low ESR caps so that ESR*(temperature factor)*(time factor)*(batch factor)*(WTF factor) is small compared to R_external.

that way even when ESR changes a lot, the total resistance doesnt change much.

example: 1uF + 4R. we could use a JPCON 1uF 50V cap with ESR = 3.8R and that would sort of work. but at -10C the ESR will be at least 15R (probably higher - these are lousy caps so I imagine the ESR-vs-temperature curve is a lot worse than a Rubycon ZLH, which increases 4x). And as the cap ages ESR will go up too. god knows what the cap-to-cap variation is, but I betcha its not great

this means damping depends strongly on actual cap, temperature & time. yuck. quadrupling the resistor is almost certain to have a noticeable (not in a good way) effect.

but if you use a 1uF MKP cap with ESR ~ 10mOhm + 3R9 resistor, then even if the cap ESR changes by a factor of 10 (+900%) the damping resistance goes from 3R91 to 4R = +2.3%

[AndrewT]

Originally Posted by gootee

The output currents for the ............ 100W 8-Ohm cases would be √(P/R) = .............. 3.5A.


Using 3500 uF/A, their reservoir capacitances could be.............. 12250uF, .....................
For output power ratings of less than 100 Watts, less capacitance could be used, but the ripple and distortion percentages would then be higher.
.................
But again, for the 100W case, if the transformer is rated at less than 44-0-44 Volts, or less than 480 VA (240 VA per secondary),....................

I will try to run the 100 Watt case with a series of transformer sizes, to see what effects that will have.

I am sad to have to admit I have given up reading this Thread, mainly because the posts are so long and it was just taking me too long to read and follow all the technicalities.

So I just pop in now and again to see the latest.

The above has me confused.
I use a 35+35Vac transformer for a 100W into 8r0 amplifier.
Why the reference to 44-0-44 Volts transformer?

Is that a peak of the AC waveform? or the AC voltage after applying the transformer regulation, or taken literally as a 44Vac-0-44Vac transformer giving supply rails when powering a quiescent output stage of very approximately +-60Vdc.

I am lost and this may be due to not following the Thread in detail.

[Terry Given]

Andrew,

is it possible for you to measure & post the parameters of the 35+35Vac transformer you use? (using the technique Tom gave about elleventy-seven posts ago, reposted here for convenience)

I'd like to Per-Unitise them & compare with Toms Toroid - IMO the secondary resistance is very high, as is the secondary leakage. which leads of course to bags of LF droop.

[danielwritesbac]

Hey Andrew, you might like to have a look at Tom's new spreadsheet
http://www.diyaudio.com/forums/attac...-copy.xlsx.txt
(save and then rename/remove the ".txt" so that the file name reads PSU_Square_Load_Data_PU2.xlsx and then a spreadsheet can open it)