• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

Pushing the limits of TDA1543 NOS DAC

While my commercial offering of TDA1543 NOS DAC has been quite successful (close to 400 kits and DACs sold) I wasn't really eager to modify or change anything. The DAC had couple reviews on 6moons, most recent one here: 6moons audio reviews: Audio Zone System

For good few years it was my reference DAC and I preferred it over Benchmark and Levinson 360S. Later I found out about Bidat and this become my new reference, pushing NOS DAC to a secondary system.

Few years ago I was contacted by another forum member (Bobken) who suggested some mods to my original design. I tried those mods and they improved the performance. We started email correspondence and what seemed like a trivial thing initially, turned later into 4 years vendetta of countless trials and experiments. While Bobken is still experimenting with his DAC, I'm pretty satisfied with my version and in this thread I will post all the details.

The DAC has been voiced and finalized in my friend's reference system which is highly optimized vinyl rig with 3 way active speakers. In a recent test, the system owner had hard time distinguishing vinyl from digital; the analog front end consisted of modified Versa Dynamics with double Forsell platter, custom air compressors, ZYX Universe silver cartridge and AR PH7/AS phono stages. The digital sources were either modified TL1 and TL0, plus this NOS DAC. His previous AR CD8 was later replaced by TL1/ NOS DAC combo (there was simply no doubts what sounded better).

To me, combined with TL0 and custom digital cable, this is the best DAC I tried so far, and because of it's superior performance I'm presently not really researching any new DAC designs which some of you might have noticed ;)
 
Last edited:
While Bobken's DAC is unbalanced, I decided to go balanced as in my tests it sounded superior. Feeding the output into S&B TX102 line stage allows me also get rid of any coupling caps. With careful chips selection the differential offset is below 10mV.

Here are the picture of the most recent version; the separate PS with dual transformers and special umbilical cable are important part of overall performance. Since I didn't produce any boards for that design yet, everything is hard wired.

I will post schematics tomorrow.
 

Attachments

  • d1.jpg
    d1.jpg
    187.7 KB · Views: 14,021
  • D2.jpg
    D2.jpg
    196.4 KB · Views: 12,979
  • d3.jpg
    d3.jpg
    196.8 KB · Views: 12,723
  • d4.jpg
    d4.jpg
    196.1 KB · Views: 12,415
  • d5.jpg
    d5.jpg
    194.6 KB · Views: 12,199
  • d6.jpg
    d6.jpg
    188.7 KB · Views: 4,394
  • d7.jpg
    d7.jpg
    194.8 KB · Views: 5,584
  • d8.jpg
    d8.jpg
    193.3 KB · Views: 5,310
Last edited:
Here are the regulators. I'm using two separate 5V regs for receiver's analog and digital sections and 8.3V for the DAC chips. If using a single DAC chip (unbalanced topology) the current setting resistor should be 3R8.

Except for 3 highlighted locations all the parts are the same.

For best results, the parts choices are critical.

The shunt regulator is based on Paul Hynes Z7808 Series1 module and was suggested and later optimized by Bobken
 

Attachments

  • Reg.jpg
    Reg.jpg
    155.6 KB · Views: 6,952
This is receiver and phase splitter circuits.

The three BG NX 0.1uF caps are mounted directly on the chip pins. The 10uF BG NX are directly on regulators daughter boards.The 5V analog supply is connected only to pin 22 of CS8412, the 5V digital supply is shared between receiver and phase splitter chips.

The PLL filter components are very critical. Countless hours have been spent here and initially different values were used. As the voicing and tuning of DAC circuit progressed, different values have been selected here as well. Most of the input for this section comes from Bobken, I was mostly checking out what he suggested. There are two 0.1uF caps in parallel. When I was checking my BG NX stock, it seemed like those caps were usually on a higher side and it's quite possible to find one that is already 0.2uF in value.

Nude Vishay works much better for input shunt than Caddock in this particular circuit, also nude Vishay was selected for PLL section, although I also tried S102 and it worked fine.

The data signal is converted through a logic chip into positive and negative signals and fed to a separate DAC chips. If balanced output is not required, the logic circuitry can be omitted or it can be still used to control phase of the DAC as reversing inputs at CS8412 will not reverse polarity and the passive I/V DAC will reverse polarity.

I tried quite a few digital transformers at receiver's input and it's my conclusion that all digital transformers were actually degrading sound with no benefits observed at all. The worse were Newava, while old stock SC and Lundahl were actually quite good, but still introducing their own signature.

I decided not to use any coupling caps as sound purity was my main objective. Two of my listening associates actually preferred the caps and for their DACs I installed them. It seems like TFCu V-Cap works best here (to them it was sound improvement) with higher value preferred (0.1 vs 0.022). To me the coupling caps, while smoothing the sound, will also affect immediacy and micro detail.

Also, you may notice that I'm not using 75R BNC connector. The listening tests I've done, indicated that connector type influence was rather subtle, if noticable at all, and not important in a bigger picture. The digital cable type was OTOH very influential and since all my cables were RCA terminated I went with RCA input connector as well. I also compared so called 75R digital RCA from Canare with stock Furutech, and again, we didn't observe anything unusual.
 

Attachments

  • rec2.jpg
    rec2.jpg
    87.7 KB · Views: 7,945
Last edited:
And here's the last piece of the puzzle, the DAC circuit.

I'm using two chips that are sorted for low differential offset. The offset will drift and will depend on temperature, but after proper selection, it's pretty stable and usually less than 10mV. I don't use any coupling caps here only because the output is fed directly into TVC. Elimination of coupling caps brings another level of dynamics and transparency to TDA1543 chips and if you thought they were pretty good, you might get nice surprise here.

Coupling caps will be required in any other non transformer applications and IME, Teflon V-Caps are the best choice. You may also consider MIT RTX, BG N or whatever else you feel comfortable with.

I'm placing the chips very close together to minimize traces length. There is no additional bypass on the chips, the 10uF output cap from a regulator should be placed directly on voltage and ground rails.

For I/V and Vref I'm choosing Caddock TF020 and in testing this circuit, they were preferred over nude Vishays.

The BCK and WS signals are connected in parallel to both chips. The Data+ and Data- are fed to chips separately and you may noticed that to correct for phase reversal caused by passive I/V, the positive data chip is connected to pin 3 of XLR output and negative data chip to pin 2.

This concludes circuit description. I will later post some pictures of earlier prototypes and modification of stock DAC board.
 

Attachments

  • DAC2.jpg
    DAC2.jpg
    177 KB · Views: 7,578
Last edited:
I have a suggestion. Try the following clocking scheme:

-Tent Labs 2.8224 Mhz (64*fs), connected directly to the tda1543.
-Use format 3 of cs8412 (fsync and sck are inputs).
-Use one 74hc590 synchronous counter to obtain fsync at 44.1KHz.

This is super high quality clocking. Low jitter oscillator providing the conversion clock directly to the dac chip. It sounds amazing, I could never go back to the cs8412 recovered clock.

In my system, one single sample is dropped each 5 seconds or so, but I can't hear this. I am able to eliminate the drop by synchronizing the source and it makes no audible difference.

Considering 5 seconds between slips, there are 220499 samples reproduced and finally you lose one single sample. The benefit of having the low jitter clock is much bigger than losing one of 220500 samples.

Try it and see if you can hear the dropped sample. If you can, it is not difficult to cure it, just feed the clock to an ISO isolator, multiply it, and clock your source with it.

Best,
Alex
 
Last edited:
Whenever I tried reclocking in highly optimized system, it never sounded natural; the sound had always more digital artifacts.

Whenever I tried Tent clock, it never sounded better than stock crystal that came with a transport (CD-94, Z-1000, TL0, Shigaclone). The Tent clock sounded more 'digital' than a stock crystal.

So presently, I am really reluctant to any clock tricks ;)

This DAC, when combined with TVC and high quality transport, is intended to sound like analog front end, in some ways it actually may sound better.
 

Attachments

  • TVC.jpg
    TVC.jpg
    195.3 KB · Views: 6,590
wow Peter! your work always amazes me! I love your proto board and point to point work!

A little off topic, but I keep seeing a computer in your recent pictures. I know you are using a hiface for USB->spdif, but I was wondering if you had any more info on the machine itself posted somewhere I could Look at!

Thanks again for all the great info! I sure enjoy following your posts!
 
I turned mini ITX recently and I'm pretty excited about the format. You can source nice case and pico PSU from here: Mini Box M350 Embedded Fanless Mini-ITX CaseMini Box M350 Embedded Fanless Mini-ITX Case

I'm using latest Gigabyte mobo, i5 2500K, SSD and custom 12V supply.

This system is presently used for movies only, but one day I might try it with music too. Add to this Dell touch monitor and it beats any laptop (except for 11" macbook air ;)

http://accessories.us.dell.com/sna/...CJ&cid=24471&lid=566643&acd=10550055-1727683-
 

Attachments

  • c1.jpg
    c1.jpg
    194 KB · Views: 3,965
  • c2.jpg
    c2.jpg
    194.1 KB · Views: 3,363
  • c3.jpg
    c3.jpg
    193.8 KB · Views: 3,413
Can you tell more about the case material? I really like the understated simplicity.

Is this going to be offered as a kit or are we on our own with this? After I finish the several other projects I'm working on this has moved up the list. I've been looking for a high quality balanced dac and this looks to be a reasonably simple design. Thanks for posting all this.
 
I wanted to make the cases out of 3 x 6" aluminum extrusion but it was hard to source it locally. So instead I used 4 x 6 extrusion, cutting it in half to reduce the height. It actually came out better this way, with a ventilation slot and less resonances.

4 angles and 12 screws hold the whole structure together, it's all very rigid.

I cut everything myself and finished with bead blasting (aluminum oxide, 80 grit). Later it was sprayed with Engine Finish which produces quite durable and basically invisible shield.

I'd rather see it as a guideline for improving existing DACs, than a kit offer. Some ideas shown here can be also implemented in other designs.
 

Attachments

  • 1.jpg
    1.jpg
    194.2 KB · Views: 5,426
  • 2.jpg
    2.jpg
    194.7 KB · Views: 5,127
  • 3.jpg
    3.jpg
    191.3 KB · Views: 4,961
Disabled Account
Joined 2002
Whenever I tried reclocking in highly optimized system, it never sounded natural; the sound had always more digital artifacts.

Whenever I tried Tent clock, it never sounded better than stock crystal that came with a transport (CD-94, Z-1000, TL0, Shigaclone). The Tent clock sounded more 'digital' than a stock crystal.

Probably because the source IS a digital source.