Another Rainbow
Have a look at the size, price and availability of a 100uf MKT (metallized polyester) film cap. (the smallest, cheapest film type).
A 100uF polyester film cap is $32-1,000 from Mouser and non-stock. Here is a Kemet data sheet for cheap MKT types but to save you the read, the size for 100uF 63V is 42x40x20mm - big! http://au.mouser.com/pdfdocs/KEMETR60Datasheet.pdf.
100uF Polypropylene film types, generally best for audio, are not available from Mouser, at least. The sizes and prices however, would necessarily be staggering.
Are you trying to build a museum piece of high-end extremism, a practical amplifier or just tell us about your dreams of perfect, monster and unaffordium parts?
Have a look at the size, price and availability of a 100uf MKT (metallized polyester) film cap. (the smallest, cheapest film type).
A 100uF polyester film cap is $32-1,000 from Mouser and non-stock. Here is a Kemet data sheet for cheap MKT types but to save you the read, the size for 100uF 63V is 42x40x20mm - big! http://au.mouser.com/pdfdocs/KEMETR60Datasheet.pdf.
100uF Polypropylene film types, generally best for audio, are not available from Mouser, at least. The sizes and prices however, would necessarily be staggering.
Are you trying to build a museum piece of high-end extremism, a practical amplifier or just tell us about your dreams of perfect, monster and unaffordium parts?
Last edited:
Such an approach is impractical and of little benefit in my opinion.
What are you decoupling, anyway? Line level or power amp?
You can get the same result that you are attempting to achieve (for much less money and much less board space) by paralleling appropriate capacitors. I have recently done some informal tests with this and have achieved subtle subjective differences in sound. My tests are incomplete and inconclusive, but promising.
For line level, you can bypass with 22 or 47 uF electrolytics ( which have to be close to the opamp but not right on top of it) in parallel with maybe a 1 uF poly (closer) and a 0.1 uf ceramic (soldered directly under the opamp right on the supply pins). You can try smaller electrolytics and bigger polys and see how it works.
Your results may vary.
For headphone amps and power amps, the caps will obviously have to be scaled. You're on your own here but I intend to do some research myself on this topic. I did experiment a little with capacitor banks consisting of multiple electrolytics in parallel for headphone amps, and the results were very promising. Since smaller electrolytics have better high frequency response than larger ones, you might obtain excellent results with this approach without using polys (but still use the ceramic bypass!) Do some experimenting on your own. Parallel capacitor banks have lower esr than equivalent large capacitors, so you get an added advantage.
For the record, I used rat shack metallized polyprop caps of 1 and 2.2 uF that I have had sitting around new in the package for years. I will be ordering a lot of new parts in the future, but right now I'm experiment with parts I have laying around.
Don't be sucked in by the woo of boutique capacitors. Excellent results can be obtained with ordinary parts if you employ appropriate design. Some of the ideas I have suggested are not "industry standard" by any means but all have been discussed here in some capacity.
What are you decoupling, anyway? Line level or power amp?
You can get the same result that you are attempting to achieve (for much less money and much less board space) by paralleling appropriate capacitors. I have recently done some informal tests with this and have achieved subtle subjective differences in sound. My tests are incomplete and inconclusive, but promising.
For line level, you can bypass with 22 or 47 uF electrolytics ( which have to be close to the opamp but not right on top of it) in parallel with maybe a 1 uF poly (closer) and a 0.1 uf ceramic (soldered directly under the opamp right on the supply pins). You can try smaller electrolytics and bigger polys and see how it works.
Your results may vary.
For headphone amps and power amps, the caps will obviously have to be scaled. You're on your own here but I intend to do some research myself on this topic. I did experiment a little with capacitor banks consisting of multiple electrolytics in parallel for headphone amps, and the results were very promising. Since smaller electrolytics have better high frequency response than larger ones, you might obtain excellent results with this approach without using polys (but still use the ceramic bypass!) Do some experimenting on your own. Parallel capacitor banks have lower esr than equivalent large capacitors, so you get an added advantage.
For the record, I used rat shack metallized polyprop caps of 1 and 2.2 uF that I have had sitting around new in the package for years. I will be ordering a lot of new parts in the future, but right now I'm experiment with parts I have laying around.
Don't be sucked in by the woo of boutique capacitors. Excellent results can be obtained with ordinary parts if you employ appropriate design. Some of the ideas I have suggested are not "industry standard" by any means but all have been discussed here in some capacity.
While the large-value PP caps are quite large, physically, and I haven't seen radial versions, they are not all that expensive, at the places that sell speaker crossover supplies, such as madisound and parts-express. Their lead length would be way too long for high-frequency decoupling, unfortunately (which is also needed, for power amps).
The answer is, I think, arrays of electrolytics, on unetched 2-sided PCBs.
A 10x10 array of 1000 uF electrolytics should give less than 1 nH of inductance!
Terry Given made one, and measured it with a network analyzer.
Here are most of the relevant links (near the bottom of the post at the following link):
http://www.diyaudio.com/forums/chip-amps/224914-lm3886-component-selection-3.html#post3282640
-----
I developed some equations for calculating required capacitances and maximum tolerable inductances, for decoupling and reservoir caps, at:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-169.html#post3320547
and
http://www.diyaudio.com/forums/chip...-rms-power-5-watt-chip-amp-5.html#post3315914
and
http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-23.html#post2806854
and the links at the bottom of the post at
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-3.html#post3097232
-----
The decoupling and reservoir caps are the main signal path! THE SOUND that comes from the speakers is directly from CURRENT that comes from the decoupling and reservoir caps, and occasionally from the rectifiers. See the image at:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-38.html#post3117390
Cheers,
Tom
The answer is, I think, arrays of electrolytics, on unetched 2-sided PCBs.
A 10x10 array of 1000 uF electrolytics should give less than 1 nH of inductance!
Terry Given made one, and measured it with a network analyzer.
Here are most of the relevant links (near the bottom of the post at the following link):
http://www.diyaudio.com/forums/chip-amps/224914-lm3886-component-selection-3.html#post3282640
-----
I developed some equations for calculating required capacitances and maximum tolerable inductances, for decoupling and reservoir caps, at:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-169.html#post3320547
and
http://www.diyaudio.com/forums/chip...-rms-power-5-watt-chip-amp-5.html#post3315914
and
http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-23.html#post2806854
and the links at the bottom of the post at
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-3.html#post3097232
-----
The decoupling and reservoir caps are the main signal path! THE SOUND that comes from the speakers is directly from CURRENT that comes from the decoupling and reservoir caps, and occasionally from the rectifiers. See the image at:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-38.html#post3117390
Cheers,
Tom
Last edited:
This is pretty much what I'm working on! You made my work too easy. I feel like I'm cheating now.
Well, heheh, sorry, but at least it looks like you were/are definitely on the right track!
One thing that I've been trying to follow-on with, for that, is how to best mount and connect everything. It would be a shame to have 0.5 nH inductance and then ruin it by using a stupid connection scheme for the load.
For chipamps, I'm pretty-much down to thinking that one rectangular pcb will hold two cap arrays (one for each polarity), and the chipamp will be between the two arrays.
But not wanting the external components and traces for the chipamp circuit to break up the planes, I'm thinking the chipamp circuitry, minus capacitors, will have to be on a daughterboard, on very short standoffs, straddling the dividing line between the arrays' power planes, probably over the non-component side of the arrays' board. Then the power and ground pins and connections can all go basically straight down through the daughterboard and into the array board, to either side (power or ground), with minimum connection lengths.
Alternatively, the array pcbs could be mounted parallel to each other, with their non-componnt sides close together, with standoffs between them, and the daughterboard could be, maybe, perpendicular to them, and mounted very close to one edge, so that the chipamp pins could directly reach the array boards.
The heatsink mounting is somewhat awkward, either way. But I can't see any way to not have anything at all behind the rear plane of the chipamp. Both of the schemes above would occupy the space "below" the chipamp pin level, behind the rear plane of the chip. So the heatsink can be connected but it can't go "downward", at all, behind the chipamp.
Wait, maybe the chipamp package could end up parallel to the non-component side of the array board, with the front of the chip facing the board, and the heatsink plane would then also be parallel to the board. That is one way that the entire half-space behind the chip could be available, enabling use of a large heatsink mounting face that extended in all directions in the plane of the chip's rear surface.
"Daughterboard" is a new term to me, but that is one strategy I'm working on. Thick wires soldered to the pcb traces and short leads to the motherboard is the idea. Poly caps or a couple of small electrolytics can be added to the motherboard as local bypasses. It's all up in the air at this time.
It all started when I was mulling over an iterative parallel op amp headphone amp design (I still am). I was googling around for ideas and parts and I saw one vendor selling exactly such a "daughterboard." It was a compact board with paralleled medium value electrolytics for pretty cheap. So I did some more studying and found out that there were a lot of advantages to paralleling electrolytics vs using one big can. A little quality time with the breadboards and I found out that it does indeed make a subjective difference. And the whole concept dovetails nicely with iterative design. So it's win-win on every front: excellent performance from cheap common parts and simple but clever designs. Why didn't they teach me this stuff in engineering school?
Yeah, an EE degree is like an exquisite set of tools that you don't really know how to use.
I am very grateful to have been provided with the tools. But it would have been nice if there had been some course in "practical theory".
A friend of mine at Purdue got an EET degree at the same time and I think that he DID learn a lot of practical stuff, too. But he only got to sleep about 2-3 hours a night.
I am very grateful to have been provided with the tools. But it would have been nice if there had been some course in "practical theory".
A friend of mine at Purdue got an EET degree at the same time and I think that he DID learn a lot of practical stuff, too. But he only got to sleep about 2-3 hours a night.
Well, I started sudying Engineering in February 1969 .... and had been making Electronic (and not Electronic too) stuff for a few years.
And didn't keep it as a "hobby" but started commercially producing and selling stuff that same month, so .... either had to solve all practical (and *economics*) problems right away, or wouldn't have lasted long.
So far, it worked.
In fact, even some Professors asked me on some practical tips.
Not dissing them, University is the *big* base, but everything else has its value, of course.
And didn't keep it as a "hobby" but started commercially producing and selling stuff that same month, so .... either had to solve all practical (and *economics*) problems right away, or wouldn't have lasted long.
So far, it worked.
In fact, even some Professors asked me on some practical tips.
Not dissing them, University is the *big* base, but everything else has its value, of course.
Just wanted to clarify that by "daughterboars", I was referring to a small pcb with the chipamp on it, plus the input, feedback, muting, and Zobel components.
If the chipamp daughterboard was mounted against the cap-array pcbs, very close to them, so that the chip pins could easily reach either side of the array board, then the only caps needed on the chipamp board would some tiny high-frequency bypass caps for the chip's power pins. It's possible that not even those would be needed.
OK.
That's a great idea. I always keep my leads as short possible and use as big a gauge wire as practical (16 gauge unless you want to use a torch to solder). You can parallel the leads for less inductance.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.