The Best DAC is no DAC

Update on my NO DAC DSD experiments

1. Sound from JLS relay SE board with a 90ohm/22nf RC filter never sounded close to musically right. The power Vregs used are a 5V Belleson for external USB power and a 5V ADM7150 for the isolated side.

2. Driving a 390ohm/10nf RC filter with 9 parallel logic gate outputs did nothing to improve the sound.

3. Putting a Potato Semi logic gate chip in the digital signal path did not clean up the digital signal to the op amp/DC blocking circuit which had a 390/10nf RC filter.

4. A JLS interface board of my own design using 74ALVCxx logic family chips sounded much better as the feed to the op amp/DC blocking circuit with the 390/10nf RC filter. Without this partial success I would have shelved the NO DAC DSD project.

5. The idea or using a 16-bit sine wave weighted moving average pre-filter kept gnawing away at me so in a masochistic moment I ordered the parts and built one on a perf board using 74VHC595 8-bit shift registers and Dale RN55 resisters. See picture. I inserted it between my JLS interface board and the op amp/DC blocking circuit with the 390/10nf RC filter. The first shock came when sound actually came through all this. The treble was congested a bit as if there was too much filtering. I changed the op amp RC filter to 250ohm/10nf which is a 60KHz LPF and the music became surprising good. Lots of detail and rich instrumental timbre. I am using Grado GS1000 headphones for listening tests. There is some “new parts sound” but this is diminishing with 24/7 play time. The 16-bit sine filter also seems to cancel most of the noise.

Next Steps:
1. I am encouraged enough by the SQ to design a PCB for the 16-bit sine pre-filter which can also be improved.

2. The JLS interface board layout needs to be changed to put the logic input gates right at the DSD pins from the JLSounds board to cut down on the “antenna noise” effect.

3. The RC filter values are nowhere near optimized. I will build an LDR variable resistor to make this much easier. The LDR is also an excellent sounding resistor.
 

Attachments

  • 16-bit filter.JPG
    16-bit filter.JPG
    558.2 KB · Views: 1,188
This might be of interest to some here.

Over on the TPA forum a couple of us have been experimenting with using miero's Beaglebone/Botic combo as the basis for an HQPlayer NAA solution with some good results;

http://www.diyaudio.com/forums/twisted-pear/258254-support-botic-linux-driver-177.html#post4768839

I should say that the experimentation I'm doing is using a Buffalo IIIse DAC, not a nodac, but changing that would be simple. A separate isolator/reclocker is necessary, something like the Acko SO3 or TPA Cronus. Here's a link to a description of the hardware I'm currently using;

http://www.diyaudio.com/forums/group-buys/227502-amanero-isolator-reclocker-gb-172.html#post4206993

Anyway, with the BBB/Botic/NAA combo there is no need for any sort of USB board, just plug it into a home network. It has been working perfectly at DSD256/PCM192 and almost flawlessly at DSD512/PCM384 (still a work in progress).

In an ideal world I would have liked to be running something like a boticized NAA version of Blackhole's Audiolinux but that isn't available for ARM processors so the build I'm using (which is simple and documented on the Botic thread) is based on a standard Debian Stretch release.

Ray
 
My new noDAC 0.5V ready for assembly. Next week some testing with audiolinux OS.
SAC Thailand TVC is external and used as preamplifier.
USB isolation is made with Gefen ethernet extender and a CAT 7 cable (100 m max), tested working fine up to DSD512. You can see last version here:
Gefen, LLC - USB 2.0 Extender

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.

How did the testing go?
 
Today I have connected my new noDAC. Results are not as expected.
As source I am using HQPlayer with minringFIR and DSD7 256+
I can play up to DSD256 in DoP and up to DSD512 native. The DIYINHK card could arrive to DSD1024 native, but linux alsa modules in that case should be patched. I have tested 44.1 kHz, 96 kHz, DSD64 sources.

In native mode I have a lot of noise, that almost disappears in DoP. This is very strange, maybe a firmware problem?

In DoP and native I have also some high frequency noise/whistle. With my other noDAC with Amanero and the same filter (but without high pass) and a simple single point grounding I don't have this. The whistle (not very high in volume) is present also when the volume control of my SAC Thailand TVC is at zero. I guess this is a ground problem, maybe.

The sound is congested, a little distorted and is missing some detail, the other noDAC sound more natural and unstressed.

I am sure something is wrong. I will think about this.
 
Hi, I´m following this thread silently, very nice idea, indeed.

Maybe it already came up, but this guy had a very similar idea:
6moons audio reviews: PS Audio DirectStream PerfectWave DAC

He explains his method of lowpass filtering and powersupply-concept a little bit...

Ted Smith, the guy who invented this Direct Stream DAC upsamples the frequency of DSD ten times (~3.84 MHz). Which means that he is buffering the stream and calculating interpolations in between. I guess it's easier to use a Low Pass Filter on much higher frequencies than 384 kHz.

Here is a technical talk by Ted which might be inspiring for you guys --> Youtube
 
Update on my NO DAC DSD experiments - cont

Putting a 74ALVC04 logic gate right at the JLSounds pins for DSD signals greatly reduced the noise and antenna effect.

Instead of the 16-bit sine wave weighted averaging board smearing musical details it actually increases the amount of detail. 16 bit averaging at 2xDSD acts like a 375Khz LPF which is updated continuously at the DSD clock rate.

I am hearing a mixture of amazingly good music with a small but still unacceptably high amount of digital and new parts sound. The bass is deep, tight and full of impact. Instrument and voice resolution compares with the best I have ever heard. Lots of musical energy.

I will be doing many tests with PCB's and parts that will be arriving shortly. The sine weighting resistor values accuracy will be improved by a factor of 10. The digital signal paths will be shorter. The power supply for the DC blocking/op amp/RC filter board will be upgraded.

The entire DSD signal streams including logic gate noise, over shooting, non-symmetrical signal rise and fall rates, etc are consumed and filtered into an analog like signal. I will attempt to make the digital 1's and 0's more symmetrical before conversion.

I should have enough results to report on this thread in 3-4 weeks.
 
Putting a 74ALVC04 logic gate right at the JLSounds pins for DSD signals greatly reduced the noise and antenna effect.

Instead of the 16-bit sine wave weighted averaging board smearing musical details it actually increases the amount of detail. 16 bit averaging at 2xDSD acts like a 375Khz LPF which is updated continuously at the DSD clock rate.

I am hearing a mixture of amazingly good music with a small but still unacceptably high amount of digital and new parts sound. The bass is deep, tight and full of impact. Instrument and voice resolution compares with the best I have ever heard. Lots of musical energy.

I will be doing many tests with PCB's and parts that will be arriving shortly. The sine weighting resistor values accuracy will be improved by a factor of 10. The digital signal paths will be shorter. The power supply for the DC blocking/op amp/RC filter board will be upgraded.

The resistor weights are not so important. This is not an R2R ladder DAC.

I would concentrate more on proper digtal design principles like decoupling
for the chips and very low jitter. Ground plane etc.

The entire DSD signal streams including logic gate noise, over shooting, non-symmetrical signal rise and fall rates, etc are consumed and filtered into an analog like signal. I will attempt to make the digital 1's and 0's more symmetrical before conversion.

The thing with overshoot and jitter is your low pass filter may well eat it all up but the energy is not removed but transformed and inserted into your
audio signal. It ends up as various forms of distortion.

Doing this circuit right is not trivial.

I should have enough results to report on this thread in 3-4 weeks.