Grossly parallel multibit DAC adventures

Thanks for your kind post

Regarding me not asking "why this chip", well, didn't want to spoil the party... I know you like to tinker (in the most respectable way of course) in all possible directions 🙂

Well done again, great effort eventhough "blocks" were already soldered!
 
To me asking 'why?' questions are the foundation of partying. Even though you've not asked I'll mention that I found I'd done the TDA1387 a bit of a disservice in the past through not paying enough attention to the I/V resistor I was using (just a normal thick-film one). To my surprise even using an array of thick films to spread the load out made rather a significant difference to the background.

Anyway thanks for your encouragement!

@Destroyer OS - interesting you mentioning the 'Beast of 1000 JFETs' as that was definitely part of the inspiration for this project. Now having just done a search for it I find that it was actually over 2000 JFETs so seems my DAC isn't yet worthy to join it in the hall of fame. But I do still have boards and a whole reel of chips so I'm not out of the race yet....
 
kDAC_PSU.jpg


I've now relieved the bench supply of its duty and have two dedicated hybrid supply boards powering the two channels of the DAC. The board looks a bit complex because there are two regulators - a buck, followed by a linear LDO and each stage has a dedicated LC filter. One 24VAC wallwart is sufficient to power both channels - although the total current demand is over 4A the voltage is only about 3.6V at present. To get up to the full DAC voltage I think I'm going to need two more PSU boards to spread out the heat somewhat as the LDOs are already pretty toasty as it is.
 
MKP-bal-filter.jpg


Here's the prototype implementation of the balanced filter schematic above with the various MKP caps we sourced from TB. I've used a few NP0s to bring the capacitance values up to the target values. The build seemed to take an age - the inductors were very fiddly to get right - for the next one its going to be a PCB to save some wiring time.
 
  • Like
Reactions: uptownsquash
I can't think of a reason not to make kits of various boards available. One or two of the required boards might not be amenable to turning into kits - the filter board for example which requires a lot of hand-selection of caps. But the filter's optional anyway - not everyone thinks a NOS DAC needs a steep reconstruction filter, having been trying it without the filter it still sounds highly decent - of all DACs I've listened to this one has the most 'unflappable' quality in that instruments are rock-solid in the soundstage no matter how busy the music gets.

When the second filter arrives (in PCB form) in a few days then I'll be able to see how much difference the filter does make. The filter's designed to work with a specific value for the I/V resistor which means it'll not be really suited to builds with significantly fewer boards than 14 per channel as the output voltage will be a bit low without the ability to increase the I/V resistor. Even a 50% build with 7 DAC boards per channel the I/V resistors are going to be about 12ohm which means no output buffering is required, just wire the I/V resistors across the output sockets.
 
Filter PCBs had an error - we got the physical size of the P18 inductors wrong. Got around that by mounting a few of the caps on the reverse side of the board so they didn't clash. Still quite a lengthy build process selecting all the cap values by hand.

The sound is highly addictive but I've not done any direct A/B comparison with my reference PCM58 balanced DAC yet. Its getting out of the way of the recording, that's what I'm appreciating at present.

Heat is a significant issue here - not just from the towers of DAC chips but also from the PSUs feeding them. Things are hot but not crazy-hot sitting on my bench but eventually this is going to have to live in a case where the ambient air is warmer. I've been experimenting with a LDO regulator which performs better than the LD1085 I started out with, which has a drop-out voltage in the region of 1V. I'd like to get the drop-out for the linear reg below 0.5V so that I can run at higher current and not need to find a bigger heatsink. The DAC towers I'm rearranging to have the boards vertical rather than horizontal to try to take advantage of natural convection.

Revisiting my old notebook yesterday from 2022 at that time I was experimenting with 'overclocking' TDA1387s by biassing their pin7 with a resistor from VDD. I noted at the time that power efficiency was better at lower supply voltages but at those supplies, the output current was proportionately lower too. With the 'boost' resistor though the output current at, say 3V can be increased to that at 5V, but with lower power dissipation from the chip. So I'm going to give that a try, though not on the current boards as the notion of tacking on of 1008 resistors doesn't fill me with much excitement.
 
  • Like
Reactions: uptownsquash