janneman said:
Well, that is exactly the point and is well stated by you.
Basically, you do not have perfect components in this current loop. In fact, you have some of the most compromised components generally in use. Since the load current does pass directly through the filter caps, that essentially means that you usually have an electrolytic capacitor of some kind directly in the loop. How many people jump at the chance to use electrolytic caps directly in any of the audio current loops? It sure would save a lot of effort and time to just use a large value electrolytic between the output stage and the speaker, for example, wouldn't it? No more worries about DC levels, which might eliminate the need for split rail power systems. In the case of a simple blocking cap, there also wouldn't be any superimposed rectifier output pulses, either. Unless you have a perfect cap in the ps filter you do get all kinds of non-linearities.
Now, I suppose that an amplifier stage with perfect PSRR would eliminate any and all of this. Just how do you attain that over the range of the rectifier output spectrum? If you measure the spectral output of a typical or even very good solid-state rectifier as applied in a typical power supply, you find all sorts of signal components up to and some times beyond a MHz. Unless you have this perfect PSRR, however achieved, these components will combine non-linearly within the audio amplifier stages. After all, in a typical amplifier using feedback, how much effective feedback might you have above 100 KHz? In other amplifier designs, different techniques might be used to attain PSRR, but they often diminish in effectiveness with frequency.
The same reasoning could be used with regard to CMRR as well. If you examine the various current loops within an entire system, you find that a lot of the coupling is via common mode paths. My observation is that within a typical system there is little or no emphasis placed on the overall CMRR. How many users actually make sure that the impedances in all the attached components are optimized for best CMRR?
Without meaning to be disrespectful here to janneman or anybody else, if all these amplifier and system imperfections are easily solved with obvious engineering techniques, just why is anybody devoting any time at all to discussing any of this?
Andy
You can reference C1 to bootstraped cascode emitter as well, I have some good experience with that and yes, internal compensation is a must.
Adam
You can reference C1 to bootstraped cascode emitter as well, I have some good experience with that and yes, internal compensation is a must.
Adam
CG said:
How about SMPS with small capacitance transforer and PFC which keeps the drawn current a sine?
CG said:
The ripple that is caused by the load current through the imperfect supply caps is ripple on the supply voltage and is attenuated by the PSRR. That means that shortcomings of the power supply caps can be *fixed* by better PSRR, better caps, etc. Engineering. A perfect PSRR *would* 100% fix this. Now, I didn't say we can have perfect PSRR. Please read my post carefully. I said, you can get arbitrarily close depending on how much money you want to spend. I am assuming that you are not looking for 1000dB PSRR from 1Hz to 1Mhz, but something like 120dB across an extended audio band would make all supply junk below all other distortion products. Engineering.
The *signal* voltage that is generated across an output cap in series with the speaker is subtracted (or added depending on the phase) to the output voltage and is not corrected for. It can only be minimized by getting the best possible cap. So it directly influences the sound.
That is the difference. The non-perfect output cap impacts directly the output voltage. The power supply cap, although in the current loop, does not, except at a secondary level via the power supply rejection ratio, which can be optimized.
Then you mention EMI and rectifier noise. A completely different subject.
Lastly, you wonder why all this discussion if it is *just* engineering. I hope you forgive me for passing on this one. 😉
Jan Didden
janneman said:
PSRR was discussed in the context of the decoupling capacitors. From the GEB perspective nothing changed, your argument is just that: :bs:
OTOH, the devotion of the GEB sometimes almost brings tears to my eyes; somebody willing to pay $1,936.00 (one thousand nine hundred thirty six dollars and no cents) for a 3.3uF/600V silver foil capacitor http://www.partsconnexion.com/catalog/CapacitorsFilm.html deserves no less than respect.
Jan;janneman said:
I believe you are being silly, No one disputes that it can be made workable, in fact it is done all the time. However this entire thread is about subtleties. How to get that last increment of performance where most audiophiles claim to hear a difference.
The capacitor/power supply rejection issue is just one more item where great care results in a better design. It's not black and white, all or nothing; it is paying attention to details that makes a given design not only measure better but sound better as well.
Most people are merely saying, watch out for the pitfall here, use a low ESR cap, maybe parallel the cap with a film cap, maybe add a shunt regulator and most of all don't let a ground loop introduce those current spikes into the amplifier gain stages.
It's not rocket science, but it is a place where newbies can too easily visualize an ideal set of functions such as; this block supplies power, this block supplies gain, this block drives the load. In reality all those functions interact at some level, a good designer minimizes those interactions.
hermanv said:
Some are unable to get it. No wonder, it needs hard work, lot of comparative tests, prototypes, experience.
7 years ago I would argue same as Jan and many others.
salas said:
A cap input filter will not work the same as a choke input filter. The choke input filter output voltage is .707 X the RMS supply voltage less rectifier drops. The cap input is 1.414 X RMS. You can't be sort of one or the other. The current in the choke is continuous or it isn't. I have built both and I prefer the choke input BUT the AC still has current peaks on it from the rectifiers switching. And if its becomes unloaded the supply quickly rises to the cap input condition.
I have looked at, listened to and measured both. Careful design of the layout around the rectifiers is critical to getting performance from either.
PMA said:
Make it less than two years. Less than two years ago, your contributions on this forum were all about engineering. Certainly something happened, since you completely changed the team. Care for my opinion as of why? 😀
In fact you are quite right, 7 years ago I returned to audio, after a very long gap. As you, and others, I considered many discussed items as superstitions. As I spent a lot of time in prototyping and listening tests, I started to get many similar experimental results as John and Charles and others.
I do not find myself subjectivist or objectivist, IMHO this division is a nonsense. I am sure there are methods that correlate with subjective perception, just we need time to find a define them.
BTW John C., have you received my link to dynamic crossover video? Excellent work of a colleague of mine here.
I do not find myself subjectivist or objectivist, IMHO this division is a nonsense. I am sure there are methods that correlate with subjective perception, just we need time to find a define them.
BTW John C., have you received my link to dynamic crossover video? Excellent work of a colleague of mine here.
I'll refer everyone to this paper authored by Paul Brokaw, generally considered to be a knowledgeable engineer:
http://www.analog.com/static/imported-files/application_notes/135208865AN-202.pdf
I'm not at all sure why this a different topic and how they can be separated. The rectifier "noise" passes through the very same supply capacitor as the signal uses as one of its load current return paths. If the capacitor is perfect, with zero impedance across the entire band and no other non-linearities, I guess it is a moot point. But, if not, then high frequency components of this noise beyond the audio range - where the PSRR of the amplifier isn't so great - will enter the amplifier through the supply terminal and may well generate intermodulation products within the audio band due to non-linearities of the active devices within the amplifier. If this were *not* the case, then why would some people go to such extremes to make very, very wide bandwidth power supply voltage regulators?
This exchange has all the makings of an endless argument. So, to spare everyone's time, I will withdraw at this point.
http://www.analog.com/static/imported-files/application_notes/135208865AN-202.pdf
janneman said:
I'm not at all sure why this a different topic and how they can be separated. The rectifier "noise" passes through the very same supply capacitor as the signal uses as one of its load current return paths. If the capacitor is perfect, with zero impedance across the entire band and no other non-linearities, I guess it is a moot point. But, if not, then high frequency components of this noise beyond the audio range - where the PSRR of the amplifier isn't so great - will enter the amplifier through the supply terminal and may well generate intermodulation products within the audio band due to non-linearities of the active devices within the amplifier. If this were *not* the case, then why would some people go to such extremes to make very, very wide bandwidth power supply voltage regulators?
This exchange has all the makings of an endless argument. So, to spare everyone's time, I will withdraw at this point.
darkfenriz said:
Good question. Potentially, this could reduce the common mode coupling possibilities.
There's at least two commercial companies that have added this approach to their designs. I have no idea how they might compare in overall performance to identical circuits powered by more traditional supplies or batteries, for that matter.
Batteries certainly get around any common mode problems, don't they?
There is an experiment for anyone who wishes to see, I mean hear, for themselves. Take some piece of your audio system. Disconnect the mains powered power supply from the circuitry it feeds and temporarily replace that supply with a suitable set of batteries. That means the same voltage as well as current capacity. Now listen and see if there is any difference at all. If there is, just what is the cause of this? BTW - I presume that anybody who tries this will do it with some DIY circuit they have built and understand well. No blowing up valuable commercial equipment and blaming it on some fool who posts blather on a web based forum. Like me, for example.
Here is what Lynn Olson had to say about it.
The distortion meter is fooled but the ear is not. Not science I know..but the way most SS amps are built it seems to me is exactly what you would build if you wanted to make a hell of a racket.
hermanv said:
Possibly. But my input came as a reaction to someone who blamed electrolytics for bad sound. The statement was something like that no audiophile would want electrlytics in the signal path, and that power supply electrolytics are also in the signal path.
Now, that's a gross and wrong generalization.There is a big difference in what an electrolytic does as a coupling cap or as a power supply cap.
I believe that if you don't make that difference, but keep the illusion that all electrolytics do the same thing to the signal wherever they are in the circuit, you are misleading the same newbies you apparently wish to educate.
But, I'll shut up. No problem.
Edit: I forgot. Gerrit Boers has done some interesting research on power supply capacitors. The app was DC heaters for tubes, but the concepts are the same. Worth a read: http://gboers.xs4all.nl/daisy/home/g3/139/g1/monoblocks/dc_heater_supply.html . There's 3 more parts, you should read until the end. Last part titled: The revenge of the caps 😉
Jan Didden
Jan, you are disappointing me!
For an power amp output cap the AC situation is exactly the same as for PSU caps (note that the brigde and Xformer are "non-existent" for most of the AC mains cycle). Both are passive bridge legs, and are effectively in series with the load. When there is no feedback loop, then any series effects of the caps (and wiring etc) are fully effective. With FB, it depends on how well it is implemented (incl. from where the reference point is taken etc) to what amount the caps are still in the signal path.
- Klaus
For an power amp output cap the AC situation is exactly the same as for PSU caps (note that the brigde and Xformer are "non-existent" for most of the AC mains cycle). Both are passive bridge legs, and are effectively in series with the load. When there is no feedback loop, then any series effects of the caps (and wiring etc) are fully effective. With FB, it depends on how well it is implemented (incl. from where the reference point is taken etc) to what amount the caps are still in the signal path.
- Klaus
Thanks for reporting your listening impressions, this confirms what I have learnt so far from the few sources that are available.
And sorry to be nitpicking - the choke input will have .9 x V RMS. Regarding the voltage surge under no load, you need caps rated at 1,5 times the nominal voltage which I think this is not a big deal.
Jan, thanks for providing the latest link in caps.
PMA, I still can't open your link on my MAC. I still have program problems with my computer. If it is what I suspect, however, I am missing out.
I haven't followed EVERY twist and turn of using caps in the power supply, but one important factor is that they are pre-biased with significant DC voltage, and I think that this makes a difference, sonically. If the same caps were input and output caps, like they used to be, then we might be less happy with the sonic result.
PMA, I still can't open your link on my MAC. I still have program problems with my computer. If it is what I suspect, however, I am missing out.
I haven't followed EVERY twist and turn of using caps in the power supply, but one important factor is that they are pre-biased with significant DC voltage, and I think that this makes a difference, sonically. If the same caps were input and output caps, like they used to be, then we might be less happy with the sonic result.
MRupp said:
Not big for the caps but for the rest of the circuitry is can be serious. And I use a separate supply so the voltages are potentially exposed on the connector. And the supply in the system was 50V regulated and as high as 120V open. Still worth the effort.
- Status
- Not open for further replies.