Hi Jan,
I had no intention of discouraging your comments, but merely wished to add some of my own experiences here.
Having discovered very many years ago that the construction and power-ratings of some resistors can have more effect on sonic results than their measured values, I felt it pertinent to mention this example. Especially as the discussion was related to the effects of impedance in a PS, whether by a resistor or by a regulator, and the differences I observed were most likely due to the signal modulating the resistances involved.
I also have "scopes and meters" for testing purposes, but sometimes they don't show these 'aural' differences to the extent which I would wish them to. However, I prefer to take advantage of what I hear, rather than disregarding this, and (occasionally!) end up with some improvements as a result.
Kind regards, 🙂
I had no intention of discouraging your comments, but merely wished to add some of my own experiences here.
Having discovered very many years ago that the construction and power-ratings of some resistors can have more effect on sonic results than their measured values, I felt it pertinent to mention this example. Especially as the discussion was related to the effects of impedance in a PS, whether by a resistor or by a regulator, and the differences I observed were most likely due to the signal modulating the resistances involved.
I also have "scopes and meters" for testing purposes, but sometimes they don't show these 'aural' differences to the extent which I would wish them to. However, I prefer to take advantage of what I hear, rather than disregarding this, and (occasionally!) end up with some improvements as a result.
Kind regards, 🙂
john curl said:
John,
Thanks for the clarification, I appreciate it.
A few comments if I may.
On the feedback circuitry, I understand your point, but I think with some care we can design perfectly stable feedback supply circuits that don't ring or overshoot. Often designers push to far in a quest to very low static Zout numbers (I know, I've been there I must confess!) and get too close to the instability domain, and then get the problems you mention. But it is not necessarlily always so. I guess we could agree on that one largely.
I also agree with you on the use of the single stage buffer in class A having relatively low supply current variation, so any impressed riple will be low and benign. Possibly this is as great an advantage for class A circuits than the class A operation in itself.
Still, in such a case, I really cannot see the need for any resistor. Even if the supply Zout is relatively high, a few ohms, this would still be better than an artificial increase by 100 ohms or so. Even if the supply Zout varies from, say 1 ohms at lf to 20 ohms at 10kHz, I still don't see what is gained by increasing it with 100 ohms or more to get the % variation down to 101 ohms at lf and 120 ohms at 10kHz.
The Zout of a regulated supply is mostly inductive in nature (that's why it increases with frequency of course) but adding a resistor only increases the total value and the inductive part stays as it is.
I remember that Philips, at about the time I worked for them, many, many years ago, made a point of decoupling all their opamps in audio equipment with 4.7 ohms to the supply. At the time I assumed that it was to decouple stages from each other as mostly these circuits used unregulated supplies, but to my surprise none of the people I talked to had a clear explanation, it appeared to be a case of "that's how we do this here". After all these years, I'm STILL in the dark 🙁
Jan Didden
Hi, Bobken,
Maybe the difference will not appear in 20hz-20khz, but will show up in 10's of Mhz freq? Means different intermodulation towards signals of 20hz-20khz?
Jan, thank you. You have gotten me to rethink a new power supply buffer that I was designing, in order to remove certain reliability problems---Mostly, if you 'look wrong' at one of my Vendetta low noise buffers, it breaks and it is a pain to fix.
John,
To have an idea and better understanding of what you are talking about, a schematic may come handy.
To have an idea and better understanding of what you are talking about, a schematic may come handy.
What usually happens is that I accidently short the output with my test probe, and it blows the series fets, instantly. I hope that this ends your confusion over this matter.
Courage, I do not publish schematics and have not done so since 1981, which is 25 years now.
Courage, I do not publish schematics and have not done so since 1981, which is 25 years now.
That's a problem of a common additional transistor filter (capacitance multiplier), you short the out for a while and a new solid-state grave is born ... 😀 . I know it.
courage said:
Welllll... It's hugely off-topic, but I beg to disagree. One knows (unconciously) much more than one knows (conciously), if you get my drift. People who are placed in uncommon or very stressfull situation can have reaction based on things they know but were not aware of that they knew.
Jan Didden
power supply feedback/damping
On the point of power supply overshhot/undershhot I have found that building the feedback so that it is critically damped is the issue here.
Adding some series swamps the reactance and overdamps the output.
IMO this is fine in a constant current environment, however in real world envionment there is some current variation.
However obtaining a PSU that is critically damped over the required frequency range is not trivial, yet can cope with the current ripple and dialectric issue of a LF design.
IMO (again) my approach is to move the PSU control loop above the audio band. Now the other issue is positive and negative supplies behaving symetrically. But for most of my designs I am running a SMPS. I am having a lot of sucess with PFC->positive and negative regulators DC/DCs. then linear stages.
On the point of power supply overshhot/undershhot I have found that building the feedback so that it is critically damped is the issue here.
Adding some series swamps the reactance and overdamps the output.
IMO this is fine in a constant current environment, however in real world envionment there is some current variation.
However obtaining a PSU that is critically damped over the required frequency range is not trivial, yet can cope with the current ripple and dialectric issue of a LF design.
IMO (again) my approach is to move the PSU control loop above the audio band. Now the other issue is positive and negative supplies behaving symetrically. But for most of my designs I am running a SMPS. I am having a lot of sucess with PFC->positive and negative regulators DC/DCs. then linear stages.
Dear bscally
How is your SMPS built like? Do you use transformer based feedback, transoptor based feedback or swither regulator on the secondary?
Furthermore could you tell something more about how you achieve critical damping, i.e. do you use dominant pole feedback loop compensation or something. 😉
You got me interested. 😀
best regards
How is your SMPS built like? Do you use transformer based feedback, transoptor based feedback or swither regulator on the secondary?
Furthermore could you tell something more about how you achieve critical damping, i.e. do you use dominant pole feedback loop compensation or something. 😉
You got me interested. 😀
best regards
An interesting fact is that a regulated power supply can have its frequency compensation entirely done at its output using a capacitor. Gerard Perrot (father of the Lavardin amps) published such a circuit in an article with some pertinent considerations ("Vieilles recettes ou nouvelle cuisine pour nos alimentations") in the french magazine 'L'Audiophile". He obtained nearly perfect square waves, indicating a resistive impedance in a wide frequency band.
anatech said:
Yes, VERY interesting. Forr, do you have a reference? I have most of the L'Audiophile issues in my files.
Jan Didden
my PSU design
OK so here is a little more information on the scheme.
The SMPS is based on 3 stages.
1) PFC - This allows a stable base to work from and reduces noise pollution and I also use the same PSU for my power amp (~120Khz switching 450V)
2) "Isolation" This is a transforme coupled opto isolated stage delivering ~120v. I use one isolation per channel (200-300Khz)
3) DC/DC dominant pole compensated feedback. (400kHz)
I went this way because I had trouble getting the optoisolated or transformer based isolation stage to deal with the response wanted. I think if there is any lesson from John C's PSU it is that make something that does it's job well, If it takes two logical blocks well it takes t logical blocks.
Also the final stage is only crittically damped to about 200Khz or half the switching frequency
I am working on a design for an isolated PFC supply delivering 120V output, however the PFC ICs are not being very understanding. But that is a discussion for another thread.
There are still improvements to be made in the 3 stage design. I have issues with current lumps being drawn from the DC/DC causing an increased noise on it's input.
Also ( on the preamp) there are 3 linear stages afer the SMPS.
I have to thank John C for the idea of adding a parallel stage to help flatten the current profile.
Brian
darkfenriz said:
OK so here is a little more information on the scheme.
The SMPS is based on 3 stages.
1) PFC - This allows a stable base to work from and reduces noise pollution and I also use the same PSU for my power amp (~120Khz switching 450V)
2) "Isolation" This is a transforme coupled opto isolated stage delivering ~120v. I use one isolation per channel (200-300Khz)
3) DC/DC dominant pole compensated feedback. (400kHz)
I went this way because I had trouble getting the optoisolated or transformer based isolation stage to deal with the response wanted. I think if there is any lesson from John C's PSU it is that make something that does it's job well, If it takes two logical blocks well it takes t logical blocks.
Also the final stage is only crittically damped to about 200Khz or half the switching frequency
I am working on a design for an isolated PFC supply delivering 120V output, however the PFC ICs are not being very understanding. But that is a discussion for another thread.
There are still improvements to be made in the 3 stage design. I have issues with current lumps being drawn from the DC/DC causing an increased noise on it's input.
Also ( on the preamp) there are 3 linear stages afer the SMPS.
I have to thank John C for the idea of adding a parallel stage to help flatten the current profile.
Brian
Hi Janneman,
Sorry, I do not have the exact reference for the Hephaistos's articles. Here the summary of all the published articles by L'Audiophile but it too lacks the reference of the issues :
http://ndaviden.club.fr/index.html
Sorry, I do not have the exact reference for the Hephaistos's articles. Here the summary of all the published articles by L'Audiophile but it too lacks the reference of the issues :
http://ndaviden.club.fr/index.html
If I've got it right...
L'Audiophile Ancienne Série :
Part 1 : N° 23 / Feb 1982 / page 15
Part 2 : N° 24 / May 1982 / page 51
Part 3 : N° 30 / Nov 1983 / page 55
Part 4 : N° 41 / 3rd Q 1987 / page 51
Hope this helps...
BTW, if someone has electronic copies of it, may I have a share ? 😉
L'Audiophile Ancienne Série :
Part 1 : N° 23 / Feb 1982 / page 15
Part 2 : N° 24 / May 1982 / page 51
Part 3 : N° 30 / Nov 1983 / page 55
Part 4 : N° 41 / 3rd Q 1987 / page 51
Hope this helps...
BTW, if someone has electronic copies of it, may I have a share ? 😉
- Status
- Not open for further replies.