I'm wondering about power supply design and PSRR

[pjpoes]

So I've been told before, especially around here, that most solid state amps have such good PSRR that a really low noise supply isn't needed. To this end, they often would say that capacitance didn't need to be very high, and that most commercial designs over did it for advertising reasons (my amp has more uf's than your amp). Now I've also read some people argue that PSRR is an over rated statistic and that quieting the power supply can make both a measurable and noticeable difference.

In my own experience, I built a few solid state amps using the Aussie Amp boards. These have a pretty high PSRR, and it was suggested that 20,000-40,000 uf's would be plenty. I tried it this way, using 600va tororoidals, and 44,000 uf's per rail. There was a very slight but noticeable hum, which was measurable on my oscilloscope too. I had used all sorts of grounding schemes to remove the noise, but nothing worked.

Then I saw those Viola and Cello amps using inductors, and so I purchased some 11mh 12 amp inductors from surplus, which has very low esr. This power supply had virtually unmeasurable ripple. As a comparison, the 44mf filter was around 1 volt of ripple with a 20 ohm load, vs around 3 millivolts with the CLC power supply. This design was noticeably quieter, in fact, I posted around here that now all I heard was hiss, and that I wanted some advice to get rid of the hiss. Turned out to be Shot noise, which is typically masked by power supply noise in my experience.

Then came project number two. I built a more powerful amplifier, and this time, I tested everything using a huge bank of over 120,000uf's of capacitance (I bought a large amount, around 50 of the 22mf, 47mf, and 66mf Hitachi caps surplus), and while things were very quiet, there was a slight bit of power supply spurious measurable on the o-scope. Again, tried different things to get rid of it, but it always remained. It wasn't audible in my rudimentary listening tests, but the actual design was going to use a switch mode power supply from Cold Amp. This is a much stiffer and quieter supply than the cap only supply would be, and when I inserted it, sure enough, all I could hear was shot noise again. The amp modules in this unit were better designed and have a lower noise floor, so again, things even quieter than before.

These two experiences lead me to believe that what people say about PSRR might be true, maybe it is over rated. I mean, according to what I had been told, this should have made no audible or measurable difference, but it did? I won't even try to say that I listened to music and blacks were blacker, and things were better separated. It's not like that, we are talking about hooking it up to test, putting my ear to the tweeter, and hearing noise. I'm wondering if people have thoughts or comments. I mean, I'm guessing that what most would argue is that these difference are so small as to be imperceptible during music, and would be masked by so many other types of noise and distortion, that the expensive involved in the power supplies aren't worth the effort. I mean, the main reason for the Switching supply is that its small and light, and when I priced a large enough transformer, this was so close in price, I decided to try this, see if I liked it. I mean, if the amps were more efficient designs I could probably put them in a .5 U rack enclosure and it would weigh 5lbs, but pump out 300 watts into 8 ohms. However, they are normal Class AB amps with a pretty high bias current, and need a lot of heat sink area, so they are in a huge enclosure, of which the inside is mostly empty space.

[lineup]

Hi.

Different amplifiers can have very different PSRR.
So they are differently depending on clean power supply.

All amplifiers benefit from clean supply.
But some amplifiers with lower PSRR benefit much more.

CLC = capacitor - inductor - capacitor, is a very good power supply filter.
Much better than the usual only C or CC (= 2 capacitors in parallell)

Nelson Pass has used CLC filters in some of his www.passdiy.com projects.
See for example in http://www.passdiy.com/pdf/sonofzen.pdf

BUILDING THE POWER SUPPLY
The Son of Zen requires a serious power supply.
-------
The air core inductors L1 and L2 have nominal values of two milliHenries,
and should be heavy gauge to carry as much as three amps DC.
These are available from MCM as part numbers 50-375 (MCM 1-800-824-8324).

See also this recent post by me.
With an image of an air core inductor, the size of a FOOTBALL.
Pass Labs >Power supply inductors (what specs?)
http://www.diyaudio.com/forums/showt...88#post1563088
It is this image:
http://www.geocities.com/super_bq/Hi...uctor_Side.JPG


Regards Lineup

[pjpoes]

Sorry, I should have noted that I was talking more about Class AB. I know Class A amps, like those from Nelson Pass, use CLC supplies. It's the high power solid state amps that you don't see much beyond a normal C or CC or CCCCCCCCCCCC supply.

I noticed that I had asked about something similar once before, and was told that Class A and Class AB with high quiescent currents can be a bit noisy (audible Hum) with no or low signal present. The poster said this is why the better supplies are used in those systems. I'm wondering if, since I am using higher quiescent currents, this is the reason I noticed the difference. I will say though, even with the lower recommended current its still noisy (I think).

[lineup]

I never mentioned any different Class.
I just confirmed, that using so called 'PI'-filtering: CLC
is more effective than CC or CRC

And that all amplifiers benefit from clean and well filtered supply.
But, some amplifiers with lower PSRR benefit more,
regardless of Class.

Lineup

[CBS240]

Hi

Some amp topologies are more susceptible to PS noise than others. CRLC filter works quite good IMO. This allows the large pulse currents from the power transformer to be on a separate conductor than the pulse currents of the amplifier output so they don't affect each other. Small film caps are good for noise suppression and to counteract the inductance of the large filter caps for higher frequency pulses. I place them parallel with the second bank of caps in the CRLC filter. Also rectifier diodes make noise and EMI when they slam shut. This can create noticeable noise. Small film caps (10nf) in parallel with the rectifier diodes helps with this. As for the hiss, it may be inherent to the amplifier, just noticeable when the noise floor from the PS is lower.


I question how much better those "high end", extra expensive caps really are when the circuit and lay out is proper. (bet I'll catch flack for that one )


In my amps, I began using two tier voltage supplies when I started using hexfet source follower outputs, which require higher gate drive. The higher one being low current for all but the output stage. This allows me to use low power voltage regulator circuits, IOW a separate amplifier that amplifies a steady DC reference voltage. This way PS noise can be significantly reduced without building a high power regulator. IMO, at a comfortable home listening level, the lower the noise floor, the more detail can be heard. At a high level, your ears can't discern the details anyway. Anyone here can spend a lot of and parallel a lot of outputs to get lots of power and volume. I've found it to be more challenging to build a high quality lower power amp (~100W rms) with an "undetectable" noise floor. Good power supply design is critical for this.

[CBS240]

I'm wondering if, since I am using higher quiescent currents, this is the reason I noticed the difference.

More current, more ripple. This is another advantage of two supplies.

[Conrad Hoffman]

I've never seen a difference between reasonable ps cap values, and way overkill. Most amps will have good low frequency psrr, and if you've got hf noise it needs to be taken care of with local hf bypassing, and bypassing at the rectifiers. If there's any residual hum at the outputs it's usually because of a grounding error or some magnetic coupling or eddy current issue with the transformer(s). You can certainly design an amp with poor lf psrr, but I can't see a good reason to take that path.

[pjpoes]

Alright Conrad, as your response was the first to really directly address the issue with an answer I was expecting, I would like an explanation for what I hear. I know I know, you can't really account for taste or subjective assessments, and I"m not asking you to. I would like to know why, for instance, there is hum regardless of grounding scheme (I use star grounding, but have tried all sorts), and that simply upgrading the supply to a design with lower ripple gets rid of the noise.

Recently I took my Acurus A200 amplifier apart and added an outboard power supply using the same transformer used originally, but greater capacitance, and this actually reduce but did not eliminate the noise. I then made a CLC supply for this one and its now gone, no hum. Yes there is hiss, it hasn't eliminated the noise floor to nill, but there is no more power supply noise. Again, an analysis of the noise on an oscilliscope shows it to be 120hz, so thats not magnetic coupling, right? Plus magnetic coupling shouldn't be fixed by larger caps, right?

Simply adding small bypass caps doesn't fix the hum at all, though I do feel that subjectively they improve the sound of the amp.

There are certain things with science vs subjective impressions that I will accept as being potentially a placebo effect. Even sometimes I will accept that things I can measure I probably can't hear. However, this is not one of them, this is not only something I can hear myself very clearly (120hz hum isn't exactly hard to miss), I can meausre rather easily, and I can remove.

Since I have no other explanation for this, my best guess would be that PSRR might not be quite what some believe it to be, and that Hum could be detected at output for reasons other than grounding or magnetic coupling, such as too much ripple on the supply lines.

I even experienced this with the Chip amps, and thought I must be doing something wrong, until I finally heard someones chip amp which multiple people referred to as dead quiet. I heard hum and was told, well yeah but its dead quiet compared to most amps at the listening distance (Which isn't good enough for me, sorry).

I'm really not trying to bate an argument here, I just want a better explanation as to why my experience doesn't match the commonly held wisdom on this subject.

[Andre Visser]

Originally posted by Conrad Hoffman
I've never seen a difference between reasonable ps cap values, and way overkill.

You must listen for a difference, not look for one.

[wg_ski]

Hiss and hum in power amps isn't usually the power supply's fault. Or even the inherent PSRR of the amp. Most of the time it comes from violating one of those 11 or so cardinal rules that Self outlined in his book.

Originally posted by CBS240

Anyone here can spend a lot of and parallel a lot of outputs to get lots of power and volume. I've found it to be more challenging to build a high quality lower power amp (~100W rms) with an "undetectable" noise floor. Good power supply design is critical for this.

With 100 watt amps, that only have 1 or 2 output pairs, with everything for both channels on one PCB, it's relatively straightforward to get it dead quiet. When you truly can have a star ground the diameter of a 12d nail and have it handle the current. To get the same out of big amps - with buss bars for distribution, lots of parallel paths that probably shouldn't be, and the only place for the star ground is back between the filter caps six inches away from where it should be - is the real challenge.