TDA8932 class D amp
One channel is completed - barring the off-board hexacap shunt - second channel under construction. Then for a comparison with the TDA1521s 
Update1 - both channels running now, 160,000uF on each. Bass is tuneful but LF colourations noticeable on piano (Ashkenazy Diabelli Variations). Solution to this is simple - more caps
Update2 - an extra ring of caps around the original hexacaps now, so we're around 250,000uF. Bass is so much better than from the TPA3116 but the HF is more ragged, less transparency overall than the TPA3116. So I'll next try an outrageously decoupled implementation just to check I'm not limiting the chip with poor PSU hygiene....
Update3 - I have over 30 1206 10uF caps per side now as local decoupling - HF problems are slight. Makes me wonder if I use it only on the bass/mid and a TDA1521 for the tweeter duty we might reach audio nirvana.... Looking at the DS its easy to see why the HF's a bit of a mess - the PSRR is about 20dB at 20kHz, even worse than an LM3886. It'd be a real pity to give up on this chip as the bass is just amazing, even without totally crazy amounts of supply caps. TDA1521 needs a lot more to get its bass uncoloured.
Update4 - my first attempt at extreme decoupling of a TDA8932 - 1206 10uFs close in (top and bottom of the package), further out are hexacaps to be added.
Update5 - in addition to the 1206s (now around 40 per phase of the output stage, 20 on each side of the board) I have some low ESR lytics, 680uF. There are six on the bottom, six on the top. Audio nirvana is being approached - this crazily decoupled amp is the liveliest amp I've ever had in my system. I am a total classD devotee now. Gotta do a PCB of this amp but how to get all the caps within range of the IC......
Update6 - added a snap of my most insanely decoupled TDA8932 yet - soldered between two X-wings of hexacaps, supported by the supply pipes. I got rather lazy with the output chokes - in this case 9 paralleled 220uH unscreened 'cheapies'.
Update7 - last night as I was casually browsing Taobao I came across this little amp board. All the specs scream that this is using TDA8932 but they don't announce the chip anywhere and the markings aren't visible on the pic. So if anyone's interested, here's the Aliexpress link : https://www.aliexpress.com/item/35w-m...156114449.html. I shall be ordering some from Taobao to check them out and suggest mods to decoupling and layout.

Update1 - both channels running now, 160,000uF on each. Bass is tuneful but LF colourations noticeable on piano (Ashkenazy Diabelli Variations). Solution to this is simple - more caps

Update2 - an extra ring of caps around the original hexacaps now, so we're around 250,000uF. Bass is so much better than from the TPA3116 but the HF is more ragged, less transparency overall than the TPA3116. So I'll next try an outrageously decoupled implementation just to check I'm not limiting the chip with poor PSU hygiene....
Update3 - I have over 30 1206 10uF caps per side now as local decoupling - HF problems are slight. Makes me wonder if I use it only on the bass/mid and a TDA1521 for the tweeter duty we might reach audio nirvana.... Looking at the DS its easy to see why the HF's a bit of a mess - the PSRR is about 20dB at 20kHz, even worse than an LM3886. It'd be a real pity to give up on this chip as the bass is just amazing, even without totally crazy amounts of supply caps. TDA1521 needs a lot more to get its bass uncoloured.
Update4 - my first attempt at extreme decoupling of a TDA8932 - 1206 10uFs close in (top and bottom of the package), further out are hexacaps to be added.
Update5 - in addition to the 1206s (now around 40 per phase of the output stage, 20 on each side of the board) I have some low ESR lytics, 680uF. There are six on the bottom, six on the top. Audio nirvana is being approached - this crazily decoupled amp is the liveliest amp I've ever had in my system. I am a total classD devotee now. Gotta do a PCB of this amp but how to get all the caps within range of the IC......
Update6 - added a snap of my most insanely decoupled TDA8932 yet - soldered between two X-wings of hexacaps, supported by the supply pipes. I got rather lazy with the output chokes - in this case 9 paralleled 220uH unscreened 'cheapies'.
Update7 - last night as I was casually browsing Taobao I came across this little amp board. All the specs scream that this is using TDA8932 but they don't announce the chip anywhere and the markings aren't visible on the pic. So if anyone's interested, here's the Aliexpress link : https://www.aliexpress.com/item/35w-m...156114449.html. I shall be ordering some from Taobao to check them out and suggest mods to decoupling and layout.
Total Comments 5
Comments
-
What I would be interested in, when the energy reserves roughly match in the two amplifier technologies, is how close the sound tonality is between them - ideally, they should match - a nice exercise, what you're doing here ... :)
Posted 28th May 2014 at 06:52 AM by fas42 -
Posted 28th May 2014 at 07:28 AM by abraxalito -
Yes, probably shouldn't have used that term - at least with equivalent capacitance in the game then some sort of benchmark can be taken - if a very large, equivalent number of smoothing caps are used causes the qualities to exponentially approach each other or not ...
Posted 28th May 2014 at 11:32 PM by fas42 -
It only just dawned on me when watching the bench supply current meter that classD is a good way to make better use of supply capacitance than B. Simply because its higher efficiency it means it effectively transforms the load to higher impedance meaning less ripple on the supply at any given output power.
Posted 29th May 2014 at 12:10 AM by abraxalito -
Yes, it seems to me that the quality of the decoupling is going to be crucial to these types of amps, the most subtle variations in physical parameters and layout would be audible ...
Posted 31st May 2014 at 12:55 AM by fas42