New year, new DAC

Here's the 'Ozone desktop pagoda' DAC for 2015.

A single TDA1387 feeds a 3 inductor quasi-elliptic filter followed by AD8017s as buffer-amps. The large ferrite cores in the base do the bal-SE conversion. The power supply is 4 * AA NiMH cells which should in theory last for a whole day's music.

The design is really a 'MkII' version of the Ozone portable where the AD815 buffers have been replaced to allow a more compact construction and lower power draw without the constraint of being able to drive IEMs directly. Whereas the portable used a stack of 1387s due to the choice of 7mm TDK inductors, this one's using pot-core chokes giving a much higher working impedance and hence higher output levels from just the one DAC chip. The desktop footprint is about that of a CD.

Update : I found some bargain Jamicon caps on Taobao which give the tower more elegant proportions, as well as improving the supply impedance to the buffer-amps and giving more breathing room to the PCBs. So here's how it's evolved....

Update2 : producing sound now, though one channel has a little digital hash at low level on it, I suspect a wiring error in the filter. Initially the bass dynamics are improved but HF (orchestral strings for example) is a bit harsher. I shall give the caps more time to bed in before adjusting the decoupling.

Update3 : did find one poor solder joint on a group of paralleled NP0 caps that turned out not to be totally in parallel. Low level digital hash still present though, needs more investigation. The caps take a few hours to break in (I probably have about a day's worth of listening on them now) but already the single DAC design wins out on LF dynamics. Currently listening to a Mercury Living Presence recording (1957 vintage) and getting blown away by the LF impact and acoustic. The superiority of this design leads me to the belief that as far as the TDA1387 goes, its the self-generated noise which is the limiting factor for LF dynamics. Hence one DAC chip per hexacap seems to be the way to go - although overall I have slightly less capacitance with this design, the Ozone portable has a stack of 4 chips sharing the cap bank. I'm not interested in further improvements to my DAC now until I've upgraded the amp to keep up - its clearly the limiting factor now and I have a couple of ideas for how to get improvements there...

Update4 - having got a considerable improvement in the amp's transparency I noticed that there was very little to choose between the tweaked Taobao sourced DAC and this one, despite the Pagoda having far superior power supplies and industrial strength filtering. My suspicion fell on the AD8017 buffer so I've now swapped this out in favour of an OP260 (thanks to Esperado for alerting me to this part). The OP260 has a diamond buffer output stage and its PSRR figures look much more realistic because the graph's specifying a feedback resistor, meaning a closed loop measurement. The package is bigger though (SO16-W). Installing this with a tweak to the output filter has brought about a small but worthwhile gain in dynamics and bass impact to nudge the Pagoda into the lead again.

Update5 - after more tweaks to the Taobao DAC pushed it ahead once again, I looked harder at this. An improvement was found going to single transistor I/V - in the same way as I did on the Taobao DAC. However the biggest improvement (primarily to the bass) was gained when I realized I'd skimped on the turns in the output transformers so they were presenting too low an inductance to the output chips. Substituting some other trafos did wonders for the bass - the new trafos have around 40H of inductance, the old ones were around 100X lower, definitely too hard on the PSU.

Attached Thumbnails