|
|
N-set
Warsaw, Poland
Posts 650
Joined on 01-07-2006
|
Post #:
|
32
|
|
Post ID:
|
29828
|
|
Reply to:
|
29827
|
|
|
|
|
Just completed my DIY DAC. It was a long and painful ass-ault, which started back in pandemia and the last push lasted from dec last year till the last week, slowed down by the birth of my son (I was working at nights when he slept). The design is deliberately the old-school with a bit of a modern twist: a parallel TDA1541A with a resistor I/V stage, coupled via top Jensen SUT's to 801A Bartola-clone stage, which I described earlier here and which has been working great as a preamp (thanks Anthony for pointing to this design years ago). The digital receiver is an old-school, textbook CS8414. I deliberately wanted to take the Audio Note path - minimal signal processing on the digital side. So no simultaneous mode, no L/R signal split, no USB (I do not stream). The modern twist is the 50Hz DEM (ecsdesign idea) and the I2S line suppressors. The power supplies were the most complicated and ChatGPT was absolutely instrumental in their optimization. It was veruy smartly compiling the decoupling startegey and at each stage I was asking it explicitly to check what real life builders did and what were the sonic impressions. Each PS line is separate, there are 9 individual lines in total and 3 transformers: 3 PS lines per 1541A chip + 2 for 84141 + 1 bias to remove the DC before the SUT's. 7 critical lines have a double stabilization: LM317 pre-reg and TentLabs shunts close to the supply pins. I did my best: It's a 2 box design: PSU for both analog and digital and the signal box with the DAC and the 801A stage. I designed PCB's myself, 2 times: 1st design was crap, then redesigned using a 4 layer PCB (8414 PCB is still crap but I ignored it), I used prime parts, again checking their technical and sonic influence with ChatGPT (asking each time to compile real-life sonic feedback) and with Guido Tent. There is all the usual porn: Nichicon KZ, FG, KW, old OSCONs and toroidy.pl, Cerafines, Z-foils as the I/V and on the 801 grids and al that audiophile blalalala. It was the fucking longest and most exhausting project so far, I made probably all possible mistakes along the way, fell into numerous pitfalls, and had to tear apart that beast several times, which ate few years of my life. One time when I thought it was all done a strange brum occured making me swear like an Irish sailor. Turned out there was an interference withe the PurePower supply and I had to reference the previously floating GND to the case via a resistor (again ChatGPT was helpful in debugging this).
Here is the signal part:
Signal + 801A:
801 supply:
And the DAC PSU, mounted below the 801A supply (slightly outdated, there is a quality Toroidy.pl trafo for the bias and the 1541A trafo are horizontally placed to minimize interference with the 801A supply ):
The fine tuning listening was done using Stax Lambda Pro + SRM717, which is a resolving combo and Philips CD304MkII as the transport. I did not have the nerve to optimize by ear each step, so I: 1) listened to various chips (3 sets on hands), picked strange 1541 non-A but R1 but behaving like an -A version (found out later from 1541 gurus it's a rare and generally very good variant); 2) changed the load, so the I/V resistance; to my surprise it did not behave like the cart load and the sound was quite insensitive to it, modulo the level; 3) removed bypassing after TentLabs shunts which was interesting as the bypasses where giving a very deep 3D presentation but too deep - sounded a bit like from a tube.
Now, after all that DIY orgy, was it worth it? I'm not sure to be honest. I was not blown away. The difference to the stock 304MkII (on 1541 too) and my Sony 557ESD with TentLabs clock, which worked for years as my CDP and now as the transport, is not big. It is a bit more relaxed and less edgy/screamy. Perhaps I need to listen more. Perhaps the tube stage is not that transparent and/or the Dannoys and mask the differences. But withe the vinyls I'm not feeling any deficiencies and all the differences between e.g. vinyl masters are very obvious. Or perhaps DAC's are not that important after all? This is a bold statement fro someone with close to none digital experience, so would be nice to hear from others how (in)sensitive the digital chain is to DAC's. In any case, since I've invested already that much into digital, I want to try another transport. TL0 is out of the question, but I wanted to try TL1X, modifying it following Lampizator.
Cheers,
Jarek
|
|
|