Hi,
Sorry, some basic questions. Looking for feedback mainly on peoples subjective experiences with regards to power supplies for power amplifiers.
1. For an amplifier with high PSRR, which is dead silent with just an all C’s power supply, does choice of cap still affect sound (e.g Panasonic vs Mundorf caps)?
2. For the same amplifier, with a high PSRR, which is completely silent on tweeters on sensitive speakers is there any merit going for ‘fancier’ power supplies. E.g CLC, CLCLCC etc? I ask because I have found double regulation to be beneficial elsewhere, for example DACs, DDC’s - e.g taking ripple down to nanovolt levels.
3. For an amplifier with moderate, to low, PSRR, let’s say the noise floor is -100dB with a CLC power supply. If I moved to CLCLCC with bypass caps and so forth will the noise floor drop even further, say down to -110dB. How do you know when the power supply design has reached the limit? How would you know if self generated noise of the amplifier itself means there no further merit because a fancier power supply cannot overcome it?
Thanks for helping me work through these points!
Gavin
Sorry, some basic questions. Looking for feedback mainly on peoples subjective experiences with regards to power supplies for power amplifiers.
1. For an amplifier with high PSRR, which is dead silent with just an all C’s power supply, does choice of cap still affect sound (e.g Panasonic vs Mundorf caps)?
2. For the same amplifier, with a high PSRR, which is completely silent on tweeters on sensitive speakers is there any merit going for ‘fancier’ power supplies. E.g CLC, CLCLCC etc? I ask because I have found double regulation to be beneficial elsewhere, for example DACs, DDC’s - e.g taking ripple down to nanovolt levels.
3. For an amplifier with moderate, to low, PSRR, let’s say the noise floor is -100dB with a CLC power supply. If I moved to CLCLCC with bypass caps and so forth will the noise floor drop even further, say down to -110dB. How do you know when the power supply design has reached the limit? How would you know if self generated noise of the amplifier itself means there no further merit because a fancier power supply cannot overcome it?
Thanks for helping me work through these points!
Gavin
All capacitors seems to affect the sound in some way, depending on their type and function.
Including even in circuits with high PSRR. But not necessarily because of ripple or noise.
But if you increase the supply impedance between the rectifiers and the capacitors by adding some
series R or L, then you will rely more on the final capacitor for sinking the amplifier signal currents.
Including even in circuits with high PSRR. But not necessarily because of ripple or noise.
But if you increase the supply impedance between the rectifiers and the capacitors by adding some
series R or L, then you will rely more on the final capacitor for sinking the amplifier signal currents.
Useful comment here: https://www.head-fi.org/threads/question-about-esr.802144/post-12434753
I think my question now relates to how do you evaluate the overall (self-induced?) background noise of an amplifier, so you can design an appropriate power supply. For example, assuming good components are use there may be no audible difference between CLC, vs CLCLCC on some amps and it would be a waste of money to go beyond CLC.
I think my question now relates to how do you evaluate the overall (self-induced?) background noise of an amplifier, so you can design an appropriate power supply. For example, assuming good components are use there may be no audible difference between CLC, vs CLCLCC on some amps and it would be a waste of money to go beyond CLC.
Does noise that is uncorrelated to the music signal matter? If you think it does, you should understand why and then you’ll answer your own question.
I found this reference from the Pass F5 manual:
"The noise of a properly laid out circuit is about 30 uV with a quiet supply, and raw supply with two 29,000 uF capacitor and 70 mV ripple will give about 100 uV noise (measured in the band from 20 to 20,000 Hertz)."
https://www.firstwatt.com/pdf/prod_f5_man.pdf
So 5mV of ripple as suggested should result in the amplifier circuit being the dominant source of noise, with the power supply only having a minimal effect (much less than the 70mV of ripple given in the example above).
Assuming I could find out the noise of an amplifier circuit from the designer, how would I calculate the proportional contribution caused by the power supply for other amplifiers, similar to what Nelson has done?
The other way of looking at it might be that so long as 5mV of ripple is maintained (even at decent peaks in music) then there's not much point in going any further? Even Purifi 1ET400A has 11.5μV of noise, so 5mV of ripple also seems to be appropriate there (subjectively speaking). No point going to nV levels because the noise of the amplifier is dominant.
Hopefully my understanding is correct.
You are giving, IMO, to much consideration to noise adding effect from a power supply. For a power amplifier, PS noise is not that critical as with DAC or preamp. It’s good to aim for PS with les noise and ripple, but not necessarily to uV levels.
Each amplifier has attenuation factor between power rails and output (PSRR). There are some designs with low PSRR, but many power amplifiers have high enough PSRR. That’s why PS noise, which is mainly 100/120 Hz ripple is attenuated by 50, 100, 1000 or more times, depending on the amp PSRR. Amplifier published noise levels are not calculated but measured.
I have on the bench one simple amplifier circuit, which measures 2.45 uV of 100 Hz ripple at output, while PS rail has 17 mV ripple. Total noise is 50 uV. So, this amplifier would be dead silent even with 170 mV rails ripple, as no one can hear 25 uV signal (that’s 100 dB below 1W) from loudspeakers.
To calculate how much of PS noise will end at amplifier output, you should know or measure amplifier PSRR.
Measured noise at amplifier output, doesn’t consist only of 100 or 120 Hz ripple, which we hear as hum. Measured total value, consists of summed contribution of various noise peaks (harmonics of main 100/120 Hz ripple, various electromagnetic interference and base circuit noise) over the whole audio frequency band. Here is an example of measured noise spectrum and total amount, at one amplifier output.
Each amplifier has attenuation factor between power rails and output (PSRR). There are some designs with low PSRR, but many power amplifiers have high enough PSRR. That’s why PS noise, which is mainly 100/120 Hz ripple is attenuated by 50, 100, 1000 or more times, depending on the amp PSRR. Amplifier published noise levels are not calculated but measured.
I have on the bench one simple amplifier circuit, which measures 2.45 uV of 100 Hz ripple at output, while PS rail has 17 mV ripple. Total noise is 50 uV. So, this amplifier would be dead silent even with 170 mV rails ripple, as no one can hear 25 uV signal (that’s 100 dB below 1W) from loudspeakers.
To calculate how much of PS noise will end at amplifier output, you should know or measure amplifier PSRR.
Measured noise at amplifier output, doesn’t consist only of 100 or 120 Hz ripple, which we hear as hum. Measured total value, consists of summed contribution of various noise peaks (harmonics of main 100/120 Hz ripple, various electromagnetic interference and base circuit noise) over the whole audio frequency band. Here is an example of measured noise spectrum and total amount, at one amplifier output.
I didn’t pay attention to the first post, so my explanation of PSRR was unnecessary. 🙂
Keep I mid that so low noise levels, as in my examples, are not expected from most amplifiers. What was advised by other members is a good guide. Several mV of PS ripple is usually enough to have perfectly noise free amplifier. There are other sources that can spoil results, like grounding and inadequate shielding.
Keep I mid that so low noise levels, as in my examples, are not expected from most amplifiers. What was advised by other members is a good guide. Several mV of PS ripple is usually enough to have perfectly noise free amplifier. There are other sources that can spoil results, like grounding and inadequate shielding.
Papa's baby amps have amusingly poor PSRR, but don't throw out the baby with the bath water.
My recent experience with an F5 going from crc to crcCLCC (and ripple went from 50mV to 2mV):
https://www.diyaudio.com/community/...en-power-amplifier.379587/page-2#post-6855323
Design the amp for the load. Build the amp with the 'standard' PS, but leave room for many more caps and inductors. Don't pay for more PS caps than you can hear.
My recent experience with an F5 going from crc to crcCLCC (and ripple went from 50mV to 2mV):
https://www.diyaudio.com/community/...en-power-amplifier.379587/page-2#post-6855323
Design the amp for the load. Build the amp with the 'standard' PS, but leave room for many more caps and inductors. Don't pay for more PS caps than you can hear.
Thanks for that spectrum from a 50Hz AC mains, which appears to be DOMINATED by those 50Hz-100Hz harmonics at least to 1kHz. Every spike sits on a 50Hz multiple. So squish that 50Hz, and the rest goes down along with it. KISS.
Last edited:
Alternatively, consider a regulator. May save cost and will virtually eliminate mains artefacts. Depends on how much V and average I you need.
Micro Audio do top notch SMPS - I already own some.
What regulator would you suggest looking at?
What regulator would you suggest looking at?
I would, but not used yet... another active member here has used both a traditional linear and Micro-audio SMPS but was not happy to share subjective feedback/comparison. Fair enough I guess, as each solution could sound very different depending upon the rest of your rig.
Last edited:
Ha, ha, definitely not so simple. 😉
FFT graph I posted is good example how high measurement resolution and graph scale can lead to impression that there is anything left to improve. Notice that measured levels, are in the single digit uV range. What would be benefit for a power amplifier to have supply with 2 uV of 100 Hz ripple, compared to 10 uV? That 10 uV is 109 dB below voltage required for 1 W at 8Ω load. Other questions are where that uV range ripple comes from, is it really there or is this EMI pickup by measurement gear or dirty ground is to be blamed.
Let’s follow the trail of measurement to see if we can come to some conclusion. Measurements were taken on my amplifier which has Jung type super regulators. Regulator noise is around 2.5 uV total if measured on the bench, with no transformers in vicinity, as shown:
Measured regulator PSRR is 106 dB at 100 Hz. At 3 A load, I have 245 mV rms ripple before regulator. Regulator should pass to his output only 1.23 uV of 100 Hz ripple. However, measured rail on the FFT graph below, has 10 uV ripple at 100 Hz and there is 50 Hz as well. Where those come from?
There are two sources. First is EMI pickup by measurement equipment. Despite having all shielded and using coaxial cables, at such low levels, gear can pickup an odd uV from power transformer. It is enough that I just switch on power bar with no load, several meters apart and measurement FFT graph will jump at 50 Hz to 0.1 uV level, so transformer will add more.
Amplifier case has steel base. Power transformer 50 Hz stray field propagates though this steel plate and induces small voltages at PS PCB tracks and all amplifier grounding wires. So, ground reference used for measurements is actually floating by microvolts and affecting regulator output and measured values. Induced 50 Hz is also adding harmonics, first being 100 Hz.
Comparing rail noise to amplifier output, we can see that rail values are not the same as amplifier output. Amplifier is reducing rail ripple and noise at his output, but has other sources of 50 Hz noise and related harmonics. Amplifier PCB tracks and input signal cable are also picking some of transformer 50 Hz stray field. And, I have additional source: input transformer used as voltage amplification stage. This transformer is directly picking 50 Hz and inserting this as signal (with all harmonics) to amplification chain. I could add shielding to input signal line transformers, but there is no reason as noise levels are ridiculously low.
Based on all this, I have my conclusion on do I need to make any improvement on power supply. I leave it to you to make your own conclusions that may differ from mine. 😀
Great information!
Jung type super regulator. I've been led to believe that unregulated sounds better... perhaps this relates to class A/B only, but full class A regulated can make sense, especially if there a continuous 5A draw with no peaks beyond that. What's your solution?
Jung type super regulator. I've been led to believe that unregulated sounds better... perhaps this relates to class A/B only, but full class A regulated can make sense, especially if there a continuous 5A draw with no peaks beyond that. What's your solution?