The way I see it, to lower an amp's output impedance you need:
Multiple pairs at the output, as they are in a series/parallel connection;
Low value emitter resistors (implying larger bias currents, and given the low impedance of your speakers, you're likely to have that anyway);
As wide as you can manage open loop babndwidth, to keep it as even as you can across the audio band without too much global NFB (whatever that means);
Global NFB, paradoxically as much as possible, in clear contradiction with the above;
A low, low, low impedance power supply, which again, given the low impedance of your peakers, you're likely to need anyway if you're to get the prodigious amperes of current which you need, and
Obviously, a high quality, whopping big power transformer(s).
No big deal.
Might I suggest a direct link to the Hoover dam?
Last edited:
Multiple caps or Just large values to lower PSU impedance ....
My goal ..
http://apogee-acoustics.de/en/Amplifier.html
My goal ..
http://apogee-acoustics.de/en/Amplifier.html
Well, I think it depends on the capacitors. Sometimes, a say 22,000 uF cap has a lower output impedance of two of the 10,000 uF caps in parallel.
In principle, I like a "capacitor cascade", meaning a say 22,000 uF cap in parallel with 10,000 uF in parallel with 4,700 uF. Theoretically, assuming they are of the same make and model, larger caps filter better but are slower to charge up, smaller ones don't filter as well, but are naturally faster. When you add it all up, you have 36,700 uF per supply line, or 73,400 uF per channel.
What comes after that cascade you know from the most worthwhile work of our German colleague right here, just look back (100uF // 3.3 uF foil // 100 nF foil).
You know, many, many years ago I read a text in an audio magazine (no Internet way back then), about a US company which built its power supplies with numerous 680 uF caps.
I remember it only because the text author asked the engineers why 680 uF and why so many?
They said they had no idea of how and why, but found by trial and error that this particular cap was the best compromise between size and sound quality.
They used, approximately from a photo, around 24 or so of these caps in parallel per supply line.
Not having tried it myself, I cannot comment.
I remember it only because the text author asked the engineers why 680 uF and why so many?
They said they had no idea of how and why, but found by trial and error that this particular cap was the best compromise between size and sound quality.
They used, approximately from a photo, around 24 or so of these caps in parallel per supply line.
Not having tried it myself, I cannot comment.
First loudness control was found in German radio receivers, to EQ speaker response. Then Americans second-guessed it and decided that it was based on F-M curves for EQ of ear - response, as the result the public now needs an EQ of brain-response, even on this forum.
lend me yours ....
It needs to be fine-tuned, I am working on my own curve adjustment. So far so good, but still could be better.
You need to have an array of many caps, on a two-sided PCB with unbroken planes of copper on both sides (drill holes only, with denuded hole edges on one side of pcb, for cap leads). You'll have one of those as each power rail.
With a 10x10 array of 1000 uF electrolytics, and a 1mm thick FR4 board, you should be able to get the total inductance seen at the load down to less than 1 nH; possibly 0.5 nH.
The illustrious Terry Given, electromagneticist extraordinaire, presented a prototype, with actual network analyzer measurements, and design guidelines, right here on diyaudio.
I created a set of links directly to his posts about it, so people wouldn't have to search for them individually, and wouldn't miss any, which I put at the bottom of the post at:
http://www.diyaudio.com/forums/chip-amps/224914-lm3886-component-selection-3.html#post3282640
NOTE that in the first link, Terry meant to say RADIAL instead of AXIAL.
Enjoy,
Tom
Network analyzer showed no resonances in the prototype, up to over 200 MHz (the analyzer's limit).
I suddenly realized that I wasn't even sure if his analyzer even went that low. But it does. It goes down to 100 Hz! It's an HP 3577A.
Here is the post with the Network Analyzer plot:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-103.html#post3155293
For the record, his prototype was on thicker FR4 than he wanted (1.6mm instead of 1 mm) and therefore had a slightly higher inductance than predicted, around 1.7 nH. He predicted < 0.5 nH if 1mm-thick pcb was used.
P.S. Well of course it has a resonance in the audio range. Resonance is where the impedance is LOWEST. You WANT resonance at the frequencies at which currents will be demanded, from capacitors.
Even Henry W. Ott mixes-up the terminology, in his EMC Engineering book. But it's IMPEDANCE peaks that are bad, not resonances, when you're talking about decoupling or reservoir caps.
Note that using multiple DIFFERENT VALUES of caps will always make an impedance peak between each pair of resonances.
Last edited:
Just to echo Tom's comments: there's an "art" to paralleling capacitances to get as low as impedance as you like at various frequencies, and ranges of frequencies, which is well documented. If you wish to pursue "extreme" high frequency supply impedance reduction, it can be achieved, and is highly likely to have beneficial results in audio ...
Frank
In that case we really should aim for resonances where the speaker bass (or bass/mid) driver's impedance is lowest, that'd be below 400Hz normally, before the VC inductance makes it rise again. Tweeters take relatively little current and in any case there's not much energy up that high.
Loudness contour
http://www.trustinme.de/images/G1-1.jpg
This is the Rational Audio amp ( the turntable was interesting ,the arm too ambitious to be reliable ) . If memory is correct only 9 transistors per channel in the power amp . A massive transformer looking like Naims Holden and Fischer . Simple feed-forward correction and a loudness contour which seemed only to enhance it with the Linn Isobariks . I wish I still had one . Like a fool I lost the circuit diagram . The volume control could be pulled forward to disconnect the preamp section . It was a DIN plug and socket inside the amp . Cheap and worked well . I played it to the Linn Rep ( Tony S ) . He showed the remarkable objectivity of all such reps and disliked it instantly . So sad as real music sounds like that . I told him once I disliked his companies amps as they sounded the same regardless of if the door to the toilet was closed when I listened ( I was being serious , first generation designs ) . Another friend said in reaction to that " Bright and harsh is one thing , dull and harsh , how does one do that "? I should point out I only reserve such extreme views for people who treat hi fi as politics . The saddest thing is the friend who said it sold a few . I have lost many sales by trying to be truthful ( it is subjective so it is only my truth ) . I will say one thing . That Linn amp allowed me to quickly know how much I liked something . It didn't make me want to listen repeatedly to the same music . An odd but interesting quirk . Predigested would be an analogy . The timing quality might be bandwidth limiting . It might make rhythm the key quality ?
I am taking this loudness thing seriously . I think I can see a way to do it which has minimal effect on the forward signal path . If I can work out how to adjust it like grey scale of a tube TV I will be happy .
DF 96 thank you for your views , especially how the speaker interacts . As said the Aura amp and Linn worked , I suspect massive serendipity . I understand there is an app for iPhones to use as a sound meter . Has anyone used it ? It doesn't have to be B&K , it just has to give relative measurements .
Off to try to do some work now , forgive any typos etc .
http://www.trustinme.de/images/G1-1.jpg
This is the Rational Audio amp ( the turntable was interesting ,the arm too ambitious to be reliable ) . If memory is correct only 9 transistors per channel in the power amp . A massive transformer looking like Naims Holden and Fischer . Simple feed-forward correction and a loudness contour which seemed only to enhance it with the Linn Isobariks . I wish I still had one . Like a fool I lost the circuit diagram . The volume control could be pulled forward to disconnect the preamp section . It was a DIN plug and socket inside the amp . Cheap and worked well . I played it to the Linn Rep ( Tony S ) . He showed the remarkable objectivity of all such reps and disliked it instantly . So sad as real music sounds like that . I told him once I disliked his companies amps as they sounded the same regardless of if the door to the toilet was closed when I listened ( I was being serious , first generation designs ) . Another friend said in reaction to that " Bright and harsh is one thing , dull and harsh , how does one do that "? I should point out I only reserve such extreme views for people who treat hi fi as politics . The saddest thing is the friend who said it sold a few . I have lost many sales by trying to be truthful ( it is subjective so it is only my truth ) . I will say one thing . That Linn amp allowed me to quickly know how much I liked something . It didn't make me want to listen repeatedly to the same music . An odd but interesting quirk . Predigested would be an analogy . The timing quality might be bandwidth limiting . It might make rhythm the key quality ?
I am taking this loudness thing seriously . I think I can see a way to do it which has minimal effect on the forward signal path . If I can work out how to adjust it like grey scale of a tube TV I will be happy .
DF 96 thank you for your views , especially how the speaker interacts . As said the Aura amp and Linn worked , I suspect massive serendipity . I understand there is an app for iPhones to use as a sound meter . Has anyone used it ? It doesn't have to be B&K , it just has to give relative measurements .
Off to try to do some work now , forgive any typos etc .
Last edited:
Sounds like a good idea! So we add capacitors, which should also lower the inductance a little. We could also use a thinner pcb, and/or a better hf material than fr4.
I will probably try that. I might have to order some pcb blanks if I don't have any suitable 1mm or thinner on hand. I do have "several" lots of large-ish electrolytics lying around, but will have to test them for suitability, at this point. And I do have access to high-end Agilent network analyzers. I'll probably have to solder an RF connector to the board, to be able to connect the analyzer, but I'll first do a little research to see how that should be considered.
I'm not so sure about dismissing the tweeters' requirements, although I suspect they'll be met well-enough. I am still harboring the suspicion that tiny inaccuracies in the transient response, including the tweeters, could harm the imaging. But having worked out the differential equations for what minimum capacitance is required and what maximum inductance (often connection length) is allowed, in order to provide the current necessary to reproduce an arbitrary linear or sinusoidal transient event, with a given accuracy in both time and amplitude, it seems clear that the lowest-possible inductance is desirable. So we are at least heading in the right direction. But it would probably be extremely challenging to not ruin the low impedance in the way the capacitor bank gets connected across the power output devices.
Last edited:
Well how's the art , i plan on using caps close to the outputs seperate and additional to the mains .
What ratio ...?
- Status
- Not open for further replies.