Save a thread? Like using Favorites/Bookmark in your browser? Or you could go post there and then you can use Quick Links --> Subscribed Threads.
You "could" think of it like "star grounding but also for the power rail", because we are wanting to not let the currents mix, so that there isn't mutual inductance and the total inductance will then reduce fully by paralleling (as does the total ESR of the caps and the interconnects). But it's really just paralleling of everything, so that the total impedance is reduced, like when resistors are paralleled.
We could add even more parallel paths to reduce the impedance even more.
We could parallel just conductors, with no caps involved, just to take the place of a single conductor, just in order to reduce the total parasitic inductance and resistance.
ACTUALLY, if we used n conductors, each 1/n times the size of the original conductor that they are replacing, then the inductance of each smaller conductor would be lower than that of the larger original (correct?), and the inductance-lowering effect would be even greater than I showed. But I guess the resistance of each of the smaller conductors would be greater than that of the original, so that part would be worse than I thought, in that case.
Cheers,
Tom
Last edited:
Do you mean 3 rails like here?
It depends how you have them connected. But that looks like a SERIOUS bank of capacitors!
You can also go to a thread - under the "Tread Tools" thing there's a "Subscribe to thread" option that you can use without posting to the thread.
You can also use "Save as..." in your browser to save the current webpage to your own computer for reading offlone, but the current webpage for a long thread only contains some of the posts. I've got this thing set to display 50 posts per page. And before you ask it's at UserCP->Edit Options->Thread Display options. You're welcome.
Yes, it was my experiment with ultimate class A amp. 3 rails because I used augmented source followers.
I am still ignorant of many or probably most things audio, except for some of what I have just happened to attempt to delve into, over the last three or four years. Source follower implies FET buffers? JFETS, maybe? I can't even quite remember exactly how a source follower circuit is configured (Yes, that's embarrassing.). So what does the "augmented" part mean? And why three rails?
Anyway, how well did it work?
Tom,
I'm not follow you here , 3 caps 3 rails ? Is there a diagram to view or could you offer a little more details for dummies err , like myself and tvr ...
At the link below, in post # 341, is a pretty-good summary of the "problem statement" for what I have been wanting to be able to do, with all of the jazz that I have been blathering-on about:
http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-35.html#post2863973
It did work as expected. Augmented means source follower is bootstrapped by one more source follower that keeps variations of voltage between drain and source minimal.
3 rails because I used 3 identical transformers I had because positive rail needed twice of voltage of a negative one that supplied modulated source follower. Modulated source follower kept minimal variations of current through source follower. Such feed-forward approach helped to minimize distortions without additional negative feedback keeping their order also as low as possible.
Last edited:
Dare I even ask why you would want to minimize negative feedback?
Before you answer that, or maybe preferably instead of answering that <grin>, I have "a real question". I have read a lot on diyaudio and I do know that you are one of the few people who gets almost everything right and knows almost everything needed. I actually hesitate to ask you to spend any time answering a question from someone who is so far below your level. But if you would be so kind, even a very short answer could be very helpful.
How important is it to try to avoid variations in voltage-rail voltages, due to transient current demands, in audio power amplifiers? I am not even sure that it matters much, for certain types of amps. But for the ones for which it does matter a lot, how would someone be able to know or estimate what percent variation of rail voltage should be allowable, so that they could try to size the decoupling capacitors? Or should we always just try for the lowest variation possible? Or are there rules of thumb for certain topologies? Or...?
I do understand that the rail-voltage variations could cause a sneak feedback path that could affect stability (typically quelled by bypass caps). But I am more interested in any effects that might more-directly affect the accuracy of the reproduction of the sound, and what allowable percent-variation of rail voltages would be a good target.
Before you answer that, or maybe preferably instead of answering that <grin>, I have "a real question". I have read a lot on diyaudio and I do know that you are one of the few people who gets almost everything right and knows almost everything needed. I actually hesitate to ask you to spend any time answering a question from someone who is so far below your level. But if you would be so kind, even a very short answer could be very helpful.
How important is it to try to avoid variations in voltage-rail voltages, due to transient current demands, in audio power amplifiers? I am not even sure that it matters much, for certain types of amps. But for the ones for which it does matter a lot, how would someone be able to know or estimate what percent variation of rail voltage should be allowable, so that they could try to size the decoupling capacitors? Or should we always just try for the lowest variation possible? Or are there rules of thumb for certain topologies? Or...?
I do understand that the rail-voltage variations could cause a sneak feedback path that could affect stability (typically quelled by bypass caps). But I am more interested in any effects that might more-directly affect the accuracy of the reproduction of the sound, and what allowable percent-variation of rail voltages would be a good target.
I did not want to minimize it, actually. I just wanted to get low distortions without it, because I believe that low order distortions of the same magnitude are perceived quite differently from high order distortions of the same magnitude.
You can open datasheet for typical device and see for yourself how it's parameters, like gm and capacitances change with current and voltage. That changes generate distortions. The less they change, the less are distortions.
In my view and experience, a quality voltage/current regulator is very important. It can, and often does, make an effectively modest power amp sound much larger than it really is. A good example would be the British Naim integrated amp from the mid-80ies, rated at just 30 wpc, which consistently managed to sound bigger and larger tham most Japanese factory offerings at 100 wpc, even though objectively, both measured well for rated power.
Where many fail is because of manufacturer skimping. Their regulator output stages are usually very sparse, I suppose the cost becomes prohibitive after a certain point. Marantz had such a series in the early 80ies and it passed us by unseen, left no impression at all, nobody remembers it any more.
Someone, somewhere, at some point seems to have set a rule which says your one regulator output stage transistor can safely feed two of your output stage transistors, assuming they are of the same rating or very similar. My experience seems to show that this is quite simply not so, this is good enough to get by, but for true quality it has to ba one on one proposition. Let your output stage "breathe" properly and you'll be surprised at what it can do.
The other factor where many fail is that for some reason, it is assumed that if you use electronic regulation, you can get by with smaller to much smaller filter caps. Like saying if you own a Ferrari, you can get by with diesel fuel because you have a V12 under the bonnet - ridiculous. Using electronic regulation still obliges you to use hefty filter capacitors.
Hi Tom,
I believe you should make a distinction between slow variations in supply voltage, and fast, transient-driven variations which typically manifest them selves as signal-related ripple.
Slow variations below the audio range (say a volt of variation over a couple of seconds) should be completely unaudible. Why?
The amp will be designed for a target supply voltage, say 30V. All things that nmatter, like output stage bias or input stage tail current will be designed to be immune to supply variations. So the amp will sound absolutely identical whether it is powered by 28V, 30V or 32V. Except of course that the lower the supply, the earlier it clips and that may in itself give a audible difference between, for instance, early morning with low mains or mid day with high mains. We know that amps clip much more than we think so this definitely is a factor to be aware of.
Transient-driven (within the audio range) supply ripple could in principle end up into the signal chain through capacitive coupling. This is less straightforward as it depends a lot on how competent the designer has taken care of this issue of Power Supply Rejection (PSR). So, given that there ARE transients on the supply line, whether they will be audible depends to a large extend on the design.
If you want to decrease any transient ripple, large supply decoupling caps don't work for transients as they have too large equivalent series R and equivalent series L and are too far away from a stage, in general. You need smaller capacitors, that have low ESR and ESI, and placed as close as possible to the stage you want to decouple. In effect, that cap will then provide the transient current demand for the stage, as the main reservoir cap will be 'isolated' from the stage through some wiring that has ESI and has itself (the cap) ESI and ESR.
But you can't give a general statement in this, it all depends on the amplifier design.
jan
Very much like language, most people are stuck in high-school level math (that includes many engineers - sadly). High-school level language is OK to communicate, but high-school or even 1st year university math is often not enough for state of the art physics or engineering work.
Every scientist who knows enough to understand his/her own limitations has at some point realized that it all boils down to his/her grasp of mathematics (which is The Science - or the foundation for all sciences, if you prefer).
As Einstein had written in a letter to Tullio Levi-Civita:
"I admire the elegance of your method of computation; it must be nice to ride through these fields upon the horse of true mathematics while the like of us have to make our way laboriously on foot"
Fortunately (for poets), poetry is extremely forgiving in terms of the use of language (in fact you can even make up your own words in the name of art).
Math is not forgiving at all. Science, after all, is not a Democracy. Masuka's description is more accurate, IMHO.
We agree on BlueJean and Belden cable, by the way.
Last edited:
I'll give you some of that. Not quite all. I only made it through second semester calc, which is not enough for engineering. Higher math is only one of many tools an engineer needs. My argument is that math is the language of nature, the language of science. It can be developed, studied, improved and quite elegant. "We" invent new maths to describe ever more complex observations of science. Geometry, Calculus, Chaos. It is not the science. "We" excludes me and the math I invent. As a dyslexic, is not very useful.
BTW, my boss is a mathematician, my sister a mathematician, about half of my customers are mathematicians. None of them are engineers. Most need supervision around sharp objects.
BTW, my boss is a mathematician, my sister a mathematician, about half of my customers are mathematicians. None of them are engineers. Most need supervision around sharp objects.
Californian room heater.
No one could ever accuse you of wasting chassis material ......
- Status
- Not open for further replies.