For a long time that I know this excellent article on power supplies:
http://www.tnt-audio.com/clinica/ssps1_e.html
Lately I've been thinking...
Why the hell don't these chips like high capacitance?
Why do they sound bad?
The reason is so simple that in the end I just wanna kick myself.
Bigger caps have higher inductance, and the amp will have lower damping factor at mid/high fequencies.
The thing is: these chips are really sensitive to this.
Go higher than 1500uf per rail/chip and you start noticing that the midband and treble magic goes away.
The more capacitance you add, the worse it sounds.
The solution is...
A snubber.
I've been testing tonight on my main system, with a new LM4780 amp that is
in' in for 3 days on my bench, and it really works.
The original unregulated test PSU ended at C5/C6 on my schematic.
For some time ago - since I tested and use regulated PSUs - that I use 100uf on the chip's PSU pins (not 1,000~1,500uf), and as I reported off-topic on another thread, this works better even with unregulated PSU.
I tested adding to the PSU on the schematic 10,000uf (after C5/C6) and it sounded bad, as usual, but then I tested several snubber values until I settled with this.
It sounded better and better as I lowered the value of C9/C10, the resistors were always 1R.
This is preliminary, more and more tests can be made, but I don't feel that going to a much lower value for C9/C10 will improve things even further.
This is very old stuff, I'm surprized nobody ever tested this with chip amps, or had an explanation for these chips prefering low capacitance...
Here's the schematic, have fun.
http://www.tnt-audio.com/clinica/ssps1_e.html
Lately I've been thinking...
Why the hell don't these chips like high capacitance?
Why do they sound bad?
The reason is so simple that in the end I just wanna kick myself.
Bigger caps have higher inductance, and the amp will have lower damping factor at mid/high fequencies.
The thing is: these chips are really sensitive to this.
Go higher than 1500uf per rail/chip and you start noticing that the midband and treble magic goes away.
The more capacitance you add, the worse it sounds.
The solution is...
A snubber.
I've been testing tonight on my main system, with a new LM4780 amp that is
in' in for 3 days on my bench, and it really works.The original unregulated test PSU ended at C5/C6 on my schematic.
For some time ago - since I tested and use regulated PSUs - that I use 100uf on the chip's PSU pins (not 1,000~1,500uf), and as I reported off-topic on another thread, this works better even with unregulated PSU.
I tested adding to the PSU on the schematic 10,000uf (after C5/C6) and it sounded bad, as usual, but then I tested several snubber values until I settled with this.
It sounded better and better as I lowered the value of C9/C10, the resistors were always 1R.
This is preliminary, more and more tests can be made, but I don't feel that going to a much lower value for C9/C10 will improve things even further.
This is very old stuff, I'm surprized nobody ever tested this with chip amps, or had an explanation for these chips prefering low capacitance...
Here's the schematic, have fun.
Attachments
Good work
As this is a real quantitative and qualitative claim, it would be good to actually have measurement data on it.
It's great that high capacitance finally sounds good for you.
Now why not make it reproducible for everyone?
What inductance is acceptble, what value is bad, which capacitor is the best with what snubber, etc.
I'm not into questioning your results, I just never experienced a problem with different capacitor sizes, nor any overall improvement with reducing the value below 4.7uF. While minorly improving the midrange at low levels, small sizes always worsened many other aspects of the sound. I'd like to see a systematic approach to wether the snubber can really be a cure in all situations.
Cheers,
Sebastian.
As this is a real quantitative and qualitative claim, it would be good to actually have measurement data on it.
It's great that high capacitance finally sounds good for you.
Now why not make it reproducible for everyone?
What inductance is acceptble, what value is bad, which capacitor is the best with what snubber, etc.
I'm not into questioning your results, I just never experienced a problem with different capacitor sizes, nor any overall improvement with reducing the value below 4.7uF. While minorly improving the midrange at low levels, small sizes always worsened many other aspects of the sound. I'd like to see a systematic approach to wether the snubber can really be a cure in all situations.
Cheers,
Sebastian.
Upupa Epops said:Carlos, little bit of theory : damping factor not depend only on impedance of PS
Pavel, read my post more carefully, as I didn't talk about impedace at all.
I talked about the inductance of larger caps.
http://zero-distortion.com/start.htm
Read the article "Designing your own power supply", on the last pages you have the snubber (although with different values) and an explanation on why to use it.
Same author (Dejan V. Veselinovic), same artice, a little more detailed.
sek said:Good work
As this is a real quantitative and qualitative claim, it would be good to actually have measurement data on it.
Sebastian, I have a broken scope
in my basement and I didn't find the time to look at it yet.The part that burned out is on a such a difficult place to get access that I have to dismantle the whole thing, including the screen.
As I said, this is a preliminary with an existing test PSU, it may not be a final schematic, or you may even have to tune it in every implementation.
What I say is that the effect of the snubber is quite amazing, and I've got to a value that works fine with this PSU/amp.
Everyone is welcome to test, measure, suggest, listen.
Who dares?
PS: I'm sorry, I've been very busy at work, so lately I've had less time to DIY (so many things to test
) and please be patient if I take a little longer to repply.sek said:I'd like to see a systematic approach to wether the snubber can really be a cure in all situations.
One more thing:
It's so easy to test this that I hope you guys don't start asking for proofs no end.
The 10,000uf caps I added and also the snubbers are connected to the PSU with aligators.
The amp doesn't even have a case yet.
Neither does the PSU.
If this wasn't worth a try I would not even open this thread, even after testing it.
But it's worth a try.
Hi,
Le'me start with a little introduction:
The supply capacitors basically form an RC filter network between transformer and load (in our case: amplifier), consisting of a resistance R and a capacitance C. As the resistances in the power wiring (transformer, cables, etc.) should be as low as possible, we have a very low value for R in the equation of the RC filter.
Basically, the product of R and C in the supply wiring forms a so called 'time constant' t = R*C. Now, t corresponds to the frequency 1/t where the RC filter starts to - well - filter.
Ideally, we want to have the filter work from frequencies of 0Hz and up, so that no noise or hum can pass the RC filter. This would then mean that the supply voltage would always be constant and contain no hum or noise. So, in order to let the filter work at very low frequencies, we have to compensate for the low value for R with a high value for C.
From another point of view, there's also the problem that the value of R changes, depending on the power that the amplifier draws. Accordingly, a higher value of C would compensate for a higher change of R to be still tolerable.
Simplified, the whole filter thing determines the 'regulation' of the supply voltage, e.g. how much (or better: less) the supply voltage sags under load. Consequentially, the more powerful the amplifier, the larger our C has to be for an equally good regulation. Or the other way 'round: With a larger C and thus better supply regulation, our amplifier gets less influenced by how much power it consumes (or 'how loud it plays').
As for the objective benefit of larger C:
When high current is drawn into the (loudspeaker) load, a poorly regulated power supply sags significantly, reducing it's voltage. Now, enough voltage is unfortunately exactly what we need in high power situations. The situation seems paradox: as the output power (and simplified: the output voltage) goes up, the supply voltage goes down!
This could lead to amplifier clipping once the output voltage reaches a certain amount of the supply voltage, as now the amplifier has no voltage left to 'supply itself'...
Clipping means distortion, but distortion can already occur way before clipping. E.g. when a bass signal draws much power, the weak supply takes a while to 'recover'! Thus, the music signals that occur during this 'recovery' find the amplifier in a situation of 'power compression', where it can't work up to spec and distorts and weakens the output signal...
As for the subjective improvement on the sound:
Less clipping means better sound, that's easy. But that easier 'recovery' from high power situations (with less power compression) also improves the sound.
For a better understanding, it should be noted that both power compression and clipping due to supply sagging are easier perceived at higher power levels, because the effect is stronger when more power is drawn. Thus, the improvement is something for us loud-listeners. Bass is stronger and more natural. Loudness of mids and highs doesn't get affected by the bass. Overall sound is more natural and impressive.
But that doesn't mean that listening at lower levels can't get improved by a better supply regulation. Even at low power levels, slight power compression and distortion could occur with signals of high dynamics, such as attacks of instruments. Preventing this by using large capacitors would not be so obvious, as the quantitative amount is lower, but the sound could become less coloured and listening less 'tiring' to the ear even in longer listening sessions.
The amount of improvement of course depends on many factors. But in general, the usually recommended 1,000uF for a Gainclone are considered already extremely low. 4,700uF is probably what an engineer would settle for a commercial 50W amplifier. 10,000 is probably where audiophiles would start, whatever that means...
------
As a bottom line, It can be assumed that most people actually know the bad influence of a weak supply on the sound. Remember what happens to the volume and sound quality of a 'ghetto-blaster' when the batteries drain?
Cheers,
Sebastian.
Le'me start with a little introduction:
The supply capacitors basically form an RC filter network between transformer and load (in our case: amplifier), consisting of a resistance R and a capacitance C. As the resistances in the power wiring (transformer, cables, etc.) should be as low as possible, we have a very low value for R in the equation of the RC filter.
Basically, the product of R and C in the supply wiring forms a so called 'time constant' t = R*C. Now, t corresponds to the frequency 1/t where the RC filter starts to - well - filter.
Ideally, we want to have the filter work from frequencies of 0Hz and up, so that no noise or hum can pass the RC filter. This would then mean that the supply voltage would always be constant and contain no hum or noise. So, in order to let the filter work at very low frequencies, we have to compensate for the low value for R with a high value for C.
From another point of view, there's also the problem that the value of R changes, depending on the power that the amplifier draws. Accordingly, a higher value of C would compensate for a higher change of R to be still tolerable.
Simplified, the whole filter thing determines the 'regulation' of the supply voltage, e.g. how much (or better: less) the supply voltage sags under load. Consequentially, the more powerful the amplifier, the larger our C has to be for an equally good regulation. Or the other way 'round: With a larger C and thus better supply regulation, our amplifier gets less influenced by how much power it consumes (or 'how loud it plays').
As for the objective benefit of larger C:
When high current is drawn into the (loudspeaker) load, a poorly regulated power supply sags significantly, reducing it's voltage. Now, enough voltage is unfortunately exactly what we need in high power situations. The situation seems paradox: as the output power (and simplified: the output voltage) goes up, the supply voltage goes down!
This could lead to amplifier clipping once the output voltage reaches a certain amount of the supply voltage, as now the amplifier has no voltage left to 'supply itself'...
Clipping means distortion, but distortion can already occur way before clipping. E.g. when a bass signal draws much power, the weak supply takes a while to 'recover'! Thus, the music signals that occur during this 'recovery' find the amplifier in a situation of 'power compression', where it can't work up to spec and distorts and weakens the output signal...
As for the subjective improvement on the sound:
Less clipping means better sound, that's easy. But that easier 'recovery' from high power situations (with less power compression) also improves the sound.
For a better understanding, it should be noted that both power compression and clipping due to supply sagging are easier perceived at higher power levels, because the effect is stronger when more power is drawn. Thus, the improvement is something for us loud-listeners. Bass is stronger and more natural. Loudness of mids and highs doesn't get affected by the bass. Overall sound is more natural and impressive.
But that doesn't mean that listening at lower levels can't get improved by a better supply regulation. Even at low power levels, slight power compression and distortion could occur with signals of high dynamics, such as attacks of instruments. Preventing this by using large capacitors would not be so obvious, as the quantitative amount is lower, but the sound could become less coloured and listening less 'tiring' to the ear even in longer listening sessions.
The amount of improvement of course depends on many factors. But in general, the usually recommended 1,000uF for a Gainclone are considered already extremely low. 4,700uF is probably what an engineer would settle for a commercial 50W amplifier. 10,000 is probably where audiophiles would start, whatever that means...
------
As a bottom line, It can be assumed that most people actually know the bad influence of a weak supply on the sound. Remember what happens to the volume and sound quality of a 'ghetto-blaster' when the batteries drain?
Cheers,
Sebastian.
Personally I would use 10x 1000uF instead of 1x 10000uF, connected using a sheet bus, and as close as practical to the chip supply pins.
Paralleling all those ESR's and ESL's really works in your favour, ripple current handling goes right up, and you get to choose from the ultra-low impedance types from Nichicon, Rubicon etc.
Paralleling all those ESR's and ESL's really works in your favour, ripple current handling goes right up, and you get to choose from the ultra-low impedance types from Nichicon, Rubicon etc.
Carlos
time to invest in a network analyzer --
most of the time the ESL of a capacitor is stated in nano-henries -- quite important for switching regulators where di/dt is quite large -- or if you are trying to couple an RF circuit since the inductance can swamp the capacitance.
i wonder whether the effects you experience with the LM338 regulator for your GC's isn't compensating for a transformer which has poor coupling, or is overheating -- i don't think that the series inductance of the filter caps is going to materially affect the power supply impedance at the kind of frequencies we are talking about.
time to invest in a network analyzer --
most of the time the ESL of a capacitor is stated in nano-henries -- quite important for switching regulators where di/dt is quite large -- or if you are trying to couple an RF circuit since the inductance can swamp the capacitance.
i wonder whether the effects you experience with the LM338 regulator for your GC's isn't compensating for a transformer which has poor coupling, or is overheating -- i don't think that the series inductance of the filter caps is going to materially affect the power supply impedance at the kind of frequencies we are talking about.
m0tion said:
Glad I could help.
Well, that's the culprit! They should, but many report they aren't. Not because what I describe doesn't apply to chip amps, too. But because the sound in the midrange is reportedly compromised with a higher capacitance PSU. That's why Carlos is so happy about his results with a compensation circuit for higher capacitances. I suspect we'll have to wait until further investigation has been done... (unfortunately I don't have the time, nor the need at the moment)
There's also a neat article on TNT audio about designing PSUs. It also describes the method discussed in this thread.
Sebastian.
jackinnj said:
The first GC I made was with a 2x24V 384VA custom made toroid, unregulated PSU with MUR860s and 1,000uf per rail on each LM3875 board.
A typical GC, then.
This toroid didn't ever get hot, did never vibrate.
And it was quite a good trafo for two channels.
Unfortunately this amp, although good sounding, was not able to drive my speakers properly.
Bass was not tight, even at low volume.
I added 2x4,700uf caps on the PSU diodes and at first it seamed to sound better. At least bass was tighter.
Listening more carefully a week later, I found out that the amp was sounding like a cheap one. Low-level detail, midband, treble, "air", "ambience", soundstage, all went to the toilet.
I then made plenty of tests with capacitance and got my conclusions.
Then I made another modules with LM3886 and never looked back again to the LM3875.
But with the low capacitance unregulated PSU it was still not there as I wanted. There was more to be gained in driving ability, and I knew it.
Regulation did solve this problem in a way that I call my power amp a mini-Krell.
This test now with high-capacitance unregulated PSU is just because I felt that it could be made to work much better.
I don't say it's better than to regulate, in my oppinion it's not, but it will be enough to use with a much more variety of speakers tham the typical GC can cope with.
If one can't detect a difference for the worse in changing caps for bigger ones on the unregulated PSU with these chips, then there's a lack of transparency in the system that prevents it to be detected.
I don't ever take conclusions with my test speakers, the final and cruel test is always in my main system, where it gets all exposed.
Of course, I'm so used to it that I can detect any change for better or for worse.
carlosfm said:
I am having trouble making out the capacitor value in the figure 4 at the top of part Part 2. http://www.tnt-audio.com/clinica/ssps2_e.html
Does any body know what value the writer was recommending?
Carlosfm, I know you have tested and have a recommended value you like for chipamps, but I am curious as well about what the author of the article was recommending for (presumably) solid state amps.
carlosfm said:
This one was addressed to m0tion. Although I could have choosen my words better, I only wanted to recommend him to really read it (as it appeared he hadn't, assuming he might not have noticed).
So, just add "in case you didn't notice" to my previous post...
Sorry to be redundant,
Sebastian.
Re: Re: The (high-cap.) unregulated PSU for chipamps
He suggests a starting point, but he also says it can be fine-tuned in every implementation between 220nf and 560nf.
Check the link I posted on post #4 of this thread, as that article is a little more detailed.
I started with 560nf, then 330nf, then 120nf and...
That's it for now.
It'sin'-in, sounds good.
moving_electron said:
He suggests a starting point, but he also says it can be fine-tuned in every implementation between 220nf and 560nf.
Check the link I posted on post #4 of this thread, as that article is a little more detailed.
I started with 560nf, then 330nf, then 120nf and...
That's it for now.
It's
in'-in, sounds good.- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.