diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   Resevoir capacitors for Chip Amps (http://www.diyaudio.com/forums/power-supplies/240955-resevoir-capacitors-chip-amps.html)

portreathbeach 17th August 2013 10:00 PM

Resevoir capacitors for Chip Amps
 
I know there are a lot of threads about PSUs and capacitor sizes etc. but I wanted to know if anyone has 'scoped' the +-DC from the PSU when running their chip amps.

Peter Daniel from audiosector.com uses only 1500uF caps per rail right at the chip, others use thousands of uF on the PSU board and others use snubberized supplies. Everyone claiming their particular setup is 'the best' implementation.

Has anyone out there built several different PSUs and tested them with the same chip amp? A test like putting a 25Hz sine wave into the input and see what the DC is doing at different volumes, then 50Hz, 100Hz, 200HZ, 400Hz etc.

Jsixis 17th August 2013 11:15 PM

I'm new to the scope usage so thanks for some ideas.

abraxalito 17th August 2013 11:55 PM

I didn't scope it but I did record it with my Sony PCM recorder (96k sample rate). I put some plots up on this thread - Possibly the most frugal high-end sounding amp? - Page 3

From the results of these recordings, I've gone over to massively paralleled banks of small caps :D

sofaspud 18th August 2013 12:04 AM

Basically, what you'll see on the rails is an increase in ripple proportional to the current drawn from the power supply, combined with an inverse proportionality to the amount of reservoir capacitance.
Quote:

Everyone claiming their particular setup is 'the best' implementation.
But of course. :)
And they may all be correct.
Different circuits and chip amps may have different PSRRs. Your test method isn't invalid, but what is more important than ps DC output is the effect on chip amp output, i.e. with a clean 25Hz 1V signal in, and a gain of 10, the chip amp should output a clean 25Hz 10V signal. Increasing the volume (which increases the load on the ps) will eventually cause the output to distort (ripple exceeds what the circuit can handle) or clip (the supply has reached its physical limit).
I'm not an expert, though, so there's sure to be holes in that explanation.

abraxalito 18th August 2013 12:11 AM

The major hole in that explanation is that music doesn't look very much like a single sinewave and that 'clean' doesn't have a precise definition. But your other points are good ones - I for one don't know how much PSRR is called for when music's being played and the amp is used with feedback (DS plots are shown in the open-loop case).

portreathbeach 18th August 2013 01:18 AM

Hello again abraxalito, thanks for the link. I'm currently building a small amp for my conservatory with a 15-0-15 120VA toroid I had lying around and I'm going to use a simple PSU using 2 bridge rectifiers and 2 x 3300uF caps per rail with 2 x 1500uF caps on my Peter Daniel's LM4780 board. I'm not going to use the snubber on the PSU as I have in the past, so it'll be interesting to see if the sound is any different.

Also, I normally use simple bridge rectifier diodes in my PSUs, I see that lots of people use the MUR860 diodes, do these make that much difference?

Has anyone had any good success with SMPS? Surely with enough capacitance at the output, these could work well.

lanchile 18th August 2013 01:34 AM

Quote:

Originally Posted by portreathbeach (Post 3599587)
Hello again abraxalito, thanks for the link. I'm currently building a small amp for my conservatory with a 15-0-15 120VA toroid I had lying around and I'm going to use a simple PSU using 2 bridge rectifiers and 2 x 3300uF caps per rail with 2 x 1500uF caps on my Peter Daniel's LM4780 board. I'm not going to use the snubber on the PSU as I have in the past, so it'll be interesting to see if the sound is any different.

Also, I normally use simple bridge rectifier diodes in my PSUs, I see that lots of people use the MUR860 diodes, do these make that much difference?

Has anyone had any good success with SMPS? Surely with enough capacitance at the output, these could work well.

I would use at least 2 X 6800uf caps.
Long time ago, I build Peter Daniel LM3875 and I end up using 2 X 10000uf with excellent results.
PS: I try many values from 10uf to 20000uf per rail.

abraxalito 18th August 2013 01:35 AM

Very good results with SMPSUs yes, provided they're adequately filtered for common-mode and differential mode noise. I wouldn't go back to a linear supply for a power amp now. No experience with playing with bridge diodes - the RF output is what to watch for. Use RC snubbers across them.

gootee 18th August 2013 03:40 AM

3 Attachment(s)
Quote:

Originally Posted by abraxalito (Post 3599538)
The major hole in that explanation is that music doesn't look very much like a single sinewave and that 'clean' doesn't have a precise definition. But your other points are good ones - I for one don't know how much PSRR is called for when music's being played and the amp is used with feedback (DS plots are shown in the open-loop case).

"A WHOLE LOT" of PSRR is called for, when music is playing. :)

The music IS THE CAP CURRENT, almost all of the time. And the cap voltage is proportional to the integral of the cap current. So the louder the music, the bigger the ripple, and the more PSR you need. (I attached part of the intro from the song "Highway to Hell", in an LT-Spice simulation of a power amp and PSU, showing the reservoir cap currents and the output voltage across the speaker. Very enlightening.)

For all but the very high frequencies, the reservoir and decoupling (at the chip) caps are seen as one capacitance, by the chip's power pins. So maybe Peter Daniel has more capacitance near the chip amps, to make up for the crazy-low value at the PSU.

You can't just pull the capacitance value out of the air, or base it on your "opinion" of "the sound", and expect to have an amp with the max rated output power that the rail voltage would be capable of providing. Too little capacitance and the output will clip, well before the rail voltage limit. You can and should calculate the minimum acceptable capacitance, for a desired max rated power. If you then decide to use less than that, then you have to calculate the new (lower) rated max power spec.

The peak voltage of the output signal can only go up to the point where there's still room for the ripple voltage and the voltage across the amplifier itself to sit between the signal peak and the max rail voltage. Higher than that and ripple-shaped chunks get gouged out of the output signal. That's the onset of clipping. Higher still and the entire top of the output signal will get sheared off. The spray of high frequencies inherent in the sharp edges has been known to burn up tweeters, and blow the ribbon tweeter fuses of my Magnepan MG-3.6/R speakers, unless a higher-rated amp is used, so that it isn't approaching clipping at the desired (high) loudness.

We can also easily prove, mathematically, that too little capacitance means that there is a bass frequency below which the capacitors will run out of charge, before the next charging pulse comes along. The capacitance value needed, to be able to reproduce down to a particular bass frequency, also depends on the rated maximum RMS output power, which defines the peak output signal voltage and current if a sine wave is assumed.

But as abraxalito mentioned, music doesn't usually look like a sine wave (except maybe when there's a flute solo).

The most bullet-proof way to calculate the MINIMUM required capacitance is to assume that the signal could be ANY shape, and could be anywhere up to the peak output voltage and current that are implied by the rated maximum output power. To do that, we should always assume that the output signal could be constant DC, at the peak value. Then the amp will never be able to clip, ever, and will be able to handle even the lowest bass frequencies without ever running out of current.

Once you know the max rail voltage (from transformer output voltage and rectifier voltage drop at max current) and the load impedance and the "Vclip" voltage for the amplifier itself (given in LM3886 and LM3875 datasheets, as a plot versus rail voltage), then the minimum required capacitance and the max rated output power are redundant, i.e. if you have one, you can calculate the other. And if one changes, then so does the other.

I attached the formula for calculating the absolute minimum reservoir capacitance.

But it's a whole lot easier to just let a spreadsheet calculate for you. So I attached the spreadsheet version of the formula. You can change any of the numbers in the BLUE cells, and see what happens. (Save the file with the ".txt" removed from the end of the file name.)

Cheers,

Tom

portreathbeach 18th August 2013 05:52 PM

I just gave the spreadsheet a try and it seems to work quite well. Thanks


All times are GMT. The time now is 11:47 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2