Signalyst DSC1

Speed switching...

Transformers are connected directly to the resistors.

Bruno Putzeys Mola Mola DAC, while quite different from the DSC 1, uses shift registers as the final final (acting as FIR filter I think) stage, on a single bit input at very high rate. This DAC, if you believe Mola Mola's published measurements, has a superbly low noise floor. Any thoughts on this? Would not limited speed of the shift registers matter here as well?

BTW, if I am remembering correctly, the NXP chips have better claimed performance (on the data sheet) than the TI chips, are you using the TI chips supplied with the Chinese DSC 1 Balanced kit?
 
That's measured with -0db level without AK4137. The DSD128 dsf file is converted from WAV by the saracon program.
Found an error in his measurements. The EMU was not properly calibrated.
The reality was not so beautiful. :(
  • -1db THD 0.0037%
  • -5db THD 0.0014%
  • -10db THD 0.00028%
  • -20db THD 0.0007%

BTW, if I am remembering correctly, the NXP chips have better claimed performance (on the data sheet) than the TI chips, are you using the TI chips supplied with the Chinese DSC 1 Balanced kit?
I used the ones that were in the сhinese kit - TI 74ahct595.
I tried in the unbalanced version of Fairchild 74VHC595M. Same for speed but more noise immunity (on the data sheet). The noise from Fairchild increased.
 
-1db THD 0.0037%
  • -5db THD 0.0014%
  • -10db THD 0.00028%
  • -20db THD 0.0007%
Doesn't look too bad to me. This of course includes also the ADC's THD+N.

As usual with many DACs/ADCs, best performance in this setup seems to be at about -10 dB point. 0.00028% THD+N gives -111.06 dB (equivalent of about 18.5 bits) level which is probably closest to the background noise. Enough for most recordings...

PSU noise level is critical. One could also experiment with something like 2.2 µF OSCON + 100 nF X7R/C0G in parallel on each 595's PSU pin.

Not sure how much a transformer will let the ultrasonic noise through, would be good to check.
 
[/LIST]
Doesn't look too bad to me. This of course includes also the ADC's THD+N.

As usual with many DACs/ADCs, best performance in this setup seems to be at about -10 dB point. 0.00028% THD+N gives -111.06 dB (equivalent of about 18.5 bits) level which is probably closest to the background noise. Enough for most recordings...

PSU noise level is critical. One could also experiment with something like 2.2 µF OSCON + 100 nF X7R/C0G in parallel on each 595's PSU pin.

Not sure how much a transformer will let the ultrasonic noise through, would be good to check.

It's worth noting that any OP voltage swing on the resistor network will cause
distortion. If you use a conventional virtual gnd I-V linearity will improve.

Having said that, it may not sound better! :)

Also to get a really accurate measurement, it's best to use a notch filter to attenuate fundamental.
ADC for measurement will then be much less stressed.
 
Last edited:
PSU noise level is critical. One could also experiment with something like 2.2 µF OSCON + 100 nF X7R/C0G in parallel on each 595's PSU pin.
For every 595 installed ferrite beads+tantalum 10µF+0.1 X7R. PSU - Two modules TPS7A4700 on main PCB.
This project looks like a lot of work. I wonder if the end result will pay off?
Certainly! DSC1 is now my main DAС. And I do not want to change it to something else. I like its analog/intelligent sound very much.

Many thanks to Signalist for this wonderful project!:cheers:
 
For every 595 installed ferrite beads+tantalum 10µF+0.1 X7R. PSU - Two modules TPS7A4700 on main PCB.

Certainly! DSC1 is now my main DAС. And I do not want to change it to something else. I like its analog/intelligent sound very much.

Many thanks to Signalist for this wonderful project!:cheers:

I would love to compare it with my DSD DAC. I decided to offer our commercial Purestream DSD DAC as a kit for DIYers to build. I'll be starting a thread on it soon. There will be all sorts of ways of customizing the unit for different sound and budgets. It's specifically designed to use software SRC/SDM just like the DSC-1, only uses a different strategy. All SMT components will be soldered on the boards already. It will be a very simple kit to build for anyone at all skill levels.
 
Is there anything to share yet in that regard please? Been looking to connect DSC1 to network storage via ethernet for a while :scratch:

Can't wait to see your design somewhere on ebay or alibaba!

In the short(er) term I've had terrific success using the solid-run ClearFog Base as an NAA accepting Ethernet (including fiberoptic Ethernet) input and output via USB to the DAC. Its cheap.

My own project now awaits my completing a Linux device driver which is not my forte but coming along. It is very similar to the Merging Zman board, employing a hybrid ARM/FPGA (zynq). Won't appear on ebay ;)
 
Assembled a variant of differential DSC1 with two different outputs for a quick comparison of SQ.
I like the sound more with transformers.
 

Attachments

  • DSCF2857.jpg
    DSCF2857.jpg
    199.3 KB · Views: 1,079
  • 1000Hz_-10db.png
    1000Hz_-10db.png
    92.5 KB · Views: 1,025
Sweet

This set of three boards did as a test. I wanted to compare the sound quality with different outputs.
The final version will be on one board with transformers.

If you are planning to offer for sale I would be interested. Perhaps make the transformer mounting positions include holes for various applicable parts, and high quality trannies like Jensen, etc. Most of these guys have PCB mount options.
I seem to damage shift register chips when soldering (do not know why, never had a problem with SMD chips of any other kind...) maybe offer the shift registers mounted for (apparent) less skilled solderers... I have no prob with resisters etc...