TDA1387 octal DAC looks like a good modding base

Thanks to Matthieu (Malefoda) for the heads up on this. It looks distinctly like a DAC inspired by the DAC-AH - note not only the 8 DAC chips but also the PCB layout closely follows that design's. The price on Taobao is about half the AH (368rmb) so I shall be ordering up one or more to have a play. Matthieu found it on another website with prices in USD, here - https://www.cart100.com/Product/42124644877/.
Note there are two options here (blue, black) - the cheaper one is a kit of parts I believe, the more expensive ($84) the finished unit. Going on the translation of the Taobao page, the kit does not include the transformers, mains switch or the case.

Update : here's the summary of the mods so far, for details go to Malefoda's thread linked in the comments.

1) Re-route the output ground so the opamp filtering caps aren't subject to injected CM noise from the destination component (amp or pre).

2) Reduce the supply voltage to the output opamps so X7R caps can be used between the rails to good effect. Lower voltage on X7Rs means higher capacitance so really there should only be enough supply rail volts to achieve the desired output level (2VRMS). Increase the resistors feeding the shunts (add 100R in series with existing 150R).

3) Fit 10uF/25V 1206 X7Rs across both opamp supply pins (4 per chip). Remove the 4 film caps doing decoupling. Change 10k output grounding resistors to 1M.

4) Swap out the NE5534s (if yours has these) for something a bit more RF-resistant and unity gain stable. So far AD845s sound excellent but I'm going to try others.

5) Remove 8 ceramics on TDA1387 pin7s and replace with 8 * 1000uF/6.3V lytics.

6) Remove 2.2nF caps in feedback network of opamps.

7) Include a 2 pole LC output filter with series L and shunt RC arms to attenuate the OOB image freqs and boost the HF, overcoming 'NOS droop'.

8) Experiment with classA loading the opamp output stages with resistors (or current sources) to the -ve rail.

9) Add more capacitance to the DAC chips' supply, depending on how much bass clarity you're after.

Update - here's a pic of how the modded one is looking Eagle-eyed viewers might notice a mod or two which I've forgotten to mention above.....

Update2 - matt has a Wiki about this DAC now which references this blog, so there's going to be some duplication between here, the thread and the Wiki. No bad thing. Here are the more recent mods I've been doing, many of which are mentioned in more detail (with supporting schematics) on the thread:

10) Remove the two opamps entirely and go over to single transistor I/V in differential mode. This entails splitting the 8 DACs into two camps, + and -, 4 in each and this means 4 I/V transistors. A digital inverter feeds the negative polarity DACs. Then fit LC filters (I'm currently using 5th order quasi-elliptic) between the outputs of the I/V stage and some classA discrete buffers. Finally feed the balanced outputs to a trafo to convert back to SE. The discrete buffers I've powered with +/- 7.75V supplies by reducing the voltage on the shunts that used to feed the opamps. This choice of rails allows me to use some low ESR 16V rated caps.

Subjectively the result is purer HF - noticeable on violin especially. I take it the improved treble sweetness stems from the fact that there are now no classAB stages to generate rail noise - all the active circuits are pure classA once this is implemented. The active I/V stage makes the DAC sound like it has far more capacitance on the supply rails. This I surmise is because the DAC's performance in passive I/V is limited by the biassing (pull-up) internal CCS. Presenting a much lower impedance to the DAC output means the CCS's PSRR has less of an effect.

11) Improve the PSRR of the shunts by replacing the feed resistors with LM317Ts configured as CCSs. Include ballast resistors in series with the LM317s as they have a nasty tendency to go unstable when they drop too many volts (say more than 5V). Bass performance seems to improve (fairly subtly) from this mod.

Update3 - pic of the underside posted showing the pin7 caps added (two per DAC chip). Also visible is the 74HC04 used to create the inverted phase data - 5 out of 6 inverters not used so their inputs are tied to the power rail. The PCB bottom right is a passive LC filter on the DAC's power rail.

Attached Thumbnails