One Farad capacitor array
This array of 127 10,000uF caps measures -j2.9mohms at 50Hz which I reckon makes it comfortably over 1F in capacitance. ESR is 1mohm @ 50Hz.
The individual caps were measuring on average 8,500uF. Although they're marked 10V (the brand is ChengX a slightly shady Chinese one) they're not good for anywhere near this voltage. I reckon they're in reality 6.3V and they seem to have decently low leakage at this working voltage. I shall be using them at 5V or below.
Update1 - seems I wasn't using enough copper in the wiring. I've now built a second one with 2.5mm2 wiring around the loops and 1.5mm diameter cross links and ESR is down to 0.5mohm @ 50Hz now, decreasing at higher freqs. I have enough caps left for a third one...
Update2 - now the first array has a basement level giving 2.1F total with ESR below 0.4mohms @ 50Hz. There are six paralleled TDA1387s atop the combined hexagons giving unprecedented LF dynamics. Amp long, long overdue for an upgrade....
Update3 - yesterday I spent a few hours building another of these but using 16V/10000uF caps with the intention of using it as the power base for my old TDA8566 chipamp which has almost 200,000uF stacked higgledy-piggledy on top of it. I ran into problems because although the caps were 16V they couldn't operate reliably anywhere near this voltage. I learned (or perhaps relearned) a lesson that cheapo caps can't be trusted to go anywhere near their rated voltage. These caps though weren't particularly cheap at 1.3rmb, 3X the price of the 10V caps I used initially. So the search now is on for caps which will operate up to 16V with some margin. Perhaps I'll go down to using 4700uFs, we shall see. Taobao has some nice looking Nichicon KT..
In the end I destructively tested the array of 61 at 17.6V to see which caps started to balloon, and estracted them. I took out 15 (so around a quarter of them) and replaced them with pre-selected caps from the bag, tested at the same voltage for low leakage. I now have an array which might go up to 16V on a cool day without some of the caps thermally running away... Leakage is about 80mA at this voltage, not too impressive but just about usable.
A quick browse around Taobao shows the sweetspot might well be my old favourite Sancon caps, 4700uF/25V are not any bigger than their 16V part although slightly dearer. I happen to have a bag of these and they look to have a decent margin on the voltage unlike elcheapo brands so I'll go with these for now. I;m envisaging a two-layer array, with 15,000uF (snap-in type) caps on the bottom and these on the top. Reason being the bigger caps have poorer ESR per uF. So whilst they're great for raw uF per buck, lower ESR is wanted in the mid-band. Tge Sancons show about 20mohm ESR @ 1k so I'm aiming to do under 0.5mohm from an array of 61.
The individual caps were measuring on average 8,500uF. Although they're marked 10V (the brand is ChengX a slightly shady Chinese one) they're not good for anywhere near this voltage. I reckon they're in reality 6.3V and they seem to have decently low leakage at this working voltage. I shall be using them at 5V or below.
Update1 - seems I wasn't using enough copper in the wiring. I've now built a second one with 2.5mm2 wiring around the loops and 1.5mm diameter cross links and ESR is down to 0.5mohm @ 50Hz now, decreasing at higher freqs. I have enough caps left for a third one...
Update2 - now the first array has a basement level giving 2.1F total with ESR below 0.4mohms @ 50Hz. There are six paralleled TDA1387s atop the combined hexagons giving unprecedented LF dynamics. Amp long, long overdue for an upgrade....
Update3 - yesterday I spent a few hours building another of these but using 16V/10000uF caps with the intention of using it as the power base for my old TDA8566 chipamp which has almost 200,000uF stacked higgledy-piggledy on top of it. I ran into problems because although the caps were 16V they couldn't operate reliably anywhere near this voltage. I learned (or perhaps relearned) a lesson that cheapo caps can't be trusted to go anywhere near their rated voltage. These caps though weren't particularly cheap at 1.3rmb, 3X the price of the 10V caps I used initially. So the search now is on for caps which will operate up to 16V with some margin. Perhaps I'll go down to using 4700uFs, we shall see. Taobao has some nice looking Nichicon KT..
In the end I destructively tested the array of 61 at 17.6V to see which caps started to balloon, and estracted them. I took out 15 (so around a quarter of them) and replaced them with pre-selected caps from the bag, tested at the same voltage for low leakage. I now have an array which might go up to 16V on a cool day without some of the caps thermally running away... Leakage is about 80mA at this voltage, not too impressive but just about usable.
A quick browse around Taobao shows the sweetspot might well be my old favourite Sancon caps, 4700uF/25V are not any bigger than their 16V part although slightly dearer. I happen to have a bag of these and they look to have a decent margin on the voltage unlike elcheapo brands so I'll go with these for now. I;m envisaging a two-layer array, with 15,000uF (snap-in type) caps on the bottom and these on the top. Reason being the bigger caps have poorer ESR per uF. So whilst they're great for raw uF per buck, lower ESR is wanted in the mid-band. Tge Sancons show about 20mohm ESR @ 1k so I'm aiming to do under 0.5mohm from an array of 61.
Total Comments 10
Comments
-
... and very artistically photographed, too, ;) ...Posted 3rd April 2014 at 03:19 AM by fas42
-
Thanks Frank - I was discussing with my friends here how to get some of these made as it took me a few hours to wire up. Labour should be fairly cheap locally to get a dozen or so built up so I could have a 10F cap array on the rails of my TDA1545......
Posted 3rd April 2014 at 03:27 AM by abraxalito
-
Now, this is getting interesting ... quick sim, an isolated 10F charged up to 5.5V, feeding 4mA supply current to the DAC chip, will last 100 minutes before dropping to 3V, lowest spec'ed V for the chip. So, you could easily do a whole album on 'pure battery power' - would be a fascinating exercise ...Posted 3rd April 2014 at 07:00 AM by fas42
-
Yes the DAC chip could, except I'm also using a pair of output buffers (AD815) which are quite a bit more power hungry. They will have their own caps but not as large a value as the DAC chip. Battery poiwer could always be applied from an SLA if the purpose is just to test for isolation-related issues. Alternatively (and more compact) a handful of supercaps (3F and up) could be used as power source. I don't use these directly because their ESR sucks, but they could easily feed the cap array.Posted 3rd April 2014 at 10:31 AM by abraxalito
Updated 3rd April 2014 at 10:33 AM by abraxalito -
Posted 3rd April 2014 at 03:18 PM by abraxalito
-
Why not just do 2 in series, for 50F/5.4V?
Edit: man, why not go for broke - Illinois Capacitor DCN Series: 3,500F; 2.7V; 2,300amps; ESR - 0.24mR; 0.6Kg ... :D:DPosted 3rd April 2014 at 11:03 PM by fas42
Updated 3rd April 2014 at 11:18 PM by fas42 -
Yeah except charge balancing is a bit of a pig with two in series. LT makes a nice chip to do it, but its easier to begin with to go for a single one and see how it pans out. I know there are major sized ones available, but charging time becomes an issue and those hell-for-leather ones have terminals at opposite ends, not well suited to DAC chips...
Posted 4th April 2014 at 12:05 AM by abraxalito
-
Just to check things out, no need to go to 5.4V charge - at the worst the actual values may be say 2F and 3F, just charge up to 4V, hence a safety margin ...Posted 4th April 2014 at 10:25 PM by fas42
Updated 4th April 2014 at 11:48 PM by fas42 -
Good point - but still charging time is an issue if I wanted to turn this into a reproducable design. My customers aren't going to want to wait hours for sound to come out, powering from cold.
I'll go step by step and see what supercaps bring to the party, if anything. I don't yet know for example how low in frequency the low impedance needs to go. For sure supercaps are major overkill in terms of going several octaves below 50Hz with their impedance.Posted 4th April 2014 at 11:31 PM by abraxalito
-
Yes, once in the single digit or so milliohm range, everything starts to matter - how much copper, for what distance, is being used to pass current really starts to be relevant ...Posted 8th April 2014 at 01:48 AM by fas42
