I found some more suitable EDLCs
Posted 25th August 2014 at 06:09 AM by abraxalito
Updated 26th August 2014 at 07:52 AM by abraxalito
Updated 26th August 2014 at 07:52 AM by abraxalito
Many hours can be whiled away trawling through stuff on Taobao - most recently I've been browsing the selection of electrical double layer caps (aka EDLCs, supercaps).
I looked at supercaps many years ago when they first showed up in the Farnell catalogue, initially they were designed for back-up memory purposes and always had very high ESRs so weren't much good for power supplies where any kind of current is needed. That seems to have changed and now the caps are finding their way into all kinds of energy storage applications where high pulse currents are required. ESRs have dropped to where they're attractive for audio applications which call for a really stiff power supply down to low frequencies.
The current crop (in the pic) are made by Samwha (a Korean firm) and are really nice and affordable here - 4.5rmb a piece (about $0.70). That's for a 2.7V 100F size which comes in a 22mm diameter can, 45mm tall. A 7cap hexacap packs about the same energy and roughly the same ESR as my former PCB-based ultrahexacap but its much quicker to build and almost 20dB cheaper. At this price I bave a hunch its a game changer for amp power supplies.
The 2.7V operating voltage is something of a drawback - meaning cap banks need to be put in series with some kind of charge balancing circuit. I shall begin with a simple shunt reg across the caps based on a transistor-assisted TL431 and see how this does. There will also probably be the need to incorporate some kind of semiconductor switch to isolate the caps so that they don't discharge too much in between listening sessions. I guess a low RDS(on) MOSFET might well fit the bill.
First up is to fit a two-hexacap 5.4V supply to my AD815 buffers and see if I notice any difference when up against my present 800,000uF per chip supply reservoir. If not then I'll assume that I've reached PSU transparency at 0.8F and move on to try them on a poweramp, for which I'll need to buy a few more.....
Update - soldered two hexacaps (in series) in parallel with the existing decoupling. Preliminary indications are that they're not going to come out. I've spent the last 20 or so mins listening to Sviatoslav Richter playing some Beethoven sonatas and the sound, especially in the lower registers is more addictive than previously. At first I thought there was less bass, but in fact its because there's less 'bass smearing' around the mid-lower register piano notes. When there's some deeper notes being played these have more attack, more intensity and yet more richness than before. With the supercaps on the supply I'm more drawn in to the performance. So its really looking like 800,000uF isn't enough - more listening needed, not just on piano....
I looked at supercaps many years ago when they first showed up in the Farnell catalogue, initially they were designed for back-up memory purposes and always had very high ESRs so weren't much good for power supplies where any kind of current is needed. That seems to have changed and now the caps are finding their way into all kinds of energy storage applications where high pulse currents are required. ESRs have dropped to where they're attractive for audio applications which call for a really stiff power supply down to low frequencies.
The current crop (in the pic) are made by Samwha (a Korean firm) and are really nice and affordable here - 4.5rmb a piece (about $0.70). That's for a 2.7V 100F size which comes in a 22mm diameter can, 45mm tall. A 7cap hexacap packs about the same energy and roughly the same ESR as my former PCB-based ultrahexacap but its much quicker to build and almost 20dB cheaper. At this price I bave a hunch its a game changer for amp power supplies.
The 2.7V operating voltage is something of a drawback - meaning cap banks need to be put in series with some kind of charge balancing circuit. I shall begin with a simple shunt reg across the caps based on a transistor-assisted TL431 and see how this does. There will also probably be the need to incorporate some kind of semiconductor switch to isolate the caps so that they don't discharge too much in between listening sessions. I guess a low RDS(on) MOSFET might well fit the bill.
First up is to fit a two-hexacap 5.4V supply to my AD815 buffers and see if I notice any difference when up against my present 800,000uF per chip supply reservoir. If not then I'll assume that I've reached PSU transparency at 0.8F and move on to try them on a poweramp, for which I'll need to buy a few more.....
Update - soldered two hexacaps (in series) in parallel with the existing decoupling. Preliminary indications are that they're not going to come out. I've spent the last 20 or so mins listening to Sviatoslav Richter playing some Beethoven sonatas and the sound, especially in the lower registers is more addictive than previously. At first I thought there was less bass, but in fact its because there's less 'bass smearing' around the mid-lower register piano notes. When there's some deeper notes being played these have more attack, more intensity and yet more richness than before. With the supercaps on the supply I'm more drawn in to the performance. So its really looking like 800,000uF isn't enough - more listening needed, not just on piano....
Total Comments 6
Comments
-
It's not the size, it's how you use it, ;) ...
This is getting a bit ridiculous, Richard, :p ... you might have to fill up your living space with power supply, at the rate you're going ...
This is all saying to me that the "cleanness" of the power supply to the chips is absolutely crucial - so, personally, I would try to nail a bit more the noise frequencies on the supply lines that are not allowing the chips to work at their best, and then have a more "efficient" way for ensuring that the chips don't see those nasties in their environment.
But otherwise, excellent that the sound is still improving, in the ways you describe, :) ...Posted 27th August 2014 at 01:00 AM by fas42
-
Yeah its not about size, (or numbers) but effectiveness and that can't be reduced to a quantity no matter what Lord Kelvin might think.
Living space is already dominated by power supply Frank, in the sense that bags of caps litter my floor, I can barely shift the chair at my desk without shifting a bag or two around to make space. Still now I won't be acquiring normal type caps as supercaps are so much more, ummm, 'super' at getting the sound I'm addicted to. Its really rather surprising that so far I've not uncovered a single audio application in a poweramp of supercaps. I'm not even sure Gilbert Yeung is using supercaps in his preamp (even though he has 300F+ on his supply).
Yes cleanliness of the PSU seems to be crucial, and my gut tells me its in the lower frequencies where the supercaps are making the difference. To get to 1mohm impedance at 20Hz we need 8F on the rails. So 350F as here is overkill but its needed to get the ESR not to dominate the impedance.
I'm open to suggestions towards getting LF noise down on the rails - more efficient would be great. I don't know yet if the noise is self-generated (from the ICs themselves) or from the feeding supply. My working hypothesis is its the former because I tend to agree with you that static noise isn't an issue - normal hiss doesn't seem to interfere with how the sound is perceived. But correlate the noise with the signal and then the brain's decoding processes get interfered with.
Come to think of it, a few months ago, when I was still using AD605 I came up with a circuit intended to nullify (by cancellation) the varying supply current with signal. I might resurrect that schematic and apply it to the AD815. The idea was simple - use a second chip as a 'working model' amplifier and measure in real-time its supply current and use that measurement to correct for the chip doing the real amplification. I had a series of current mirrors to do the fancy footwork. But first it might be instructive to see what variations of supply current are internal to the chip, once the output's loaded with CCSs.Posted 27th August 2014 at 03:14 AM by abraxalito
Updated 27th August 2014 at 03:54 AM by abraxalito -
Have you tried scoping or sampling what the voltages are at the power supply interface to the chips, right at the junction of the metal leg with the the plastic of the body, while the circuit is running? Could be tricky to monitor with sufficient precision, both in amplitude and frequency, to see whether there is any voltage glitching there, for any reason - and I agree it is most likely self-generated - but it may be instructive. Of course, the probing itself may upset the operation to some degree - but you should be able to get a handle on this by listening with, and without, the probe.Posted 27th August 2014 at 05:03 AM by fas42
-
No, not yet tried that - too much fun listening to the music right now to get into that. Think I'll do a bit of simulation first to get a handle on what the variations might look like in practice.Posted 27th August 2014 at 05:24 AM by abraxalito
-
Heh, heh ... know exactly what you mean, ;). When the beast is singing you think, "Measure, why do I want to measure??! - I [I]know[/I] it's workin' right!"
Something for when you get off the 'high' ... :DPosted 27th August 2014 at 05:40 AM by fas42
-
Yes I've noticed a change in my modus operandii when the kit sound's 'there'. No longer do I switch music to see how each piece is 'sounding' now, rather I choose music for itself, for the enjoyment of the performance. So its a good sign if fewer of the 'test pieces' get played and I get all adventurous with listening to new stuff....Posted 27th August 2014 at 07:26 AM by abraxalito
