Here the frequency spectra at sample frequencies 48kHz, 96kHz, 192kHz.
There is no peak around 50Hz from the power line as there was in an other post. The DAM is powered with a trafo AC and is not shilded in a case.
Again the diagrams for the RME Fireface are added to compare.
Although the RME measurements look better I can't say I like its sound better.
Keep the same ration btw FS and FFT, like 48000/16132 or 96000/32264 and so on.
Set Wnd Kaiser7 and Average Linear or Exp when measures THD.
From what I see your Fireface has the same noise pattern over 35KHz like EMU0404, rise up the noise very easy...don't set the range too high. Measure btw 20-20KHz.
I think that you will have the best results to 48KHz sampling.
Have you calibrate the inputs, on ARTA menu?
If not, do it this with 1Vrms signal, I think your card is capable for this level.
What is your source, the RME Fireface via spdif.
At the least you know, that the limit factor will be the most weak point of your measuring setup.
Ok, let's say you have seen enough And you have someone you can ask
Some people automatically thought they were buying a TotalDAC for cheap not realizing that there is a lot of work to be done,obviously not soldering but fine tuning and that requires time and effort. And some knowledge.
I am competent enough to change caps and regulators(but I have my own John Swenson around too), other people can design output stages and even program FPGAs, it would be great if Soekris DAC becomes what we expect it to be - a fine engineering example with great sound.
And you have someone you can ask Some people automatically thought they were buying a TotalDAC for cheap not realizing that there is a lot of work to be done,obviously not soldering but fine tuning and that requires time and effort. And some knowledge.
I am competent enough to change caps and regulators(but I have my own John Swenson around too), other people can design output stages and even program FPGAs, it would be great if Soekris DAC becomes what we expect it to be - a fine engineering example with great sound.
Have you calibrated, your setup?
I see lower noise floor from zfe, almost 160dBFS with 44.1KHz sampling!. Your card, has THD+N, Min. Gain < -100 dB (0.001%) @ 1 kHz, -1 dBFS
There is no any point if you haven;t calibrate the signal, before the measurement. 150 dBFS noise floor is like spdif to spdif loop-back!
Chears,
Last edited:
I'll measure again with these settings ... tomorrow.
Yes I calibrated. That is, I measured no calibration was necessary, The DAM 1021 produced almost exactly the 1mV and the Fireface's in and outputs are calibrated (with switchable sensitivity) which also proved to be nearly accurate.
The source was the RME Fireface SPDIF Coax (the Toslink looses at 192kHz synchronization depending on the optical cables).
I know that with 24 bit you are at most at 144dB but is not me who drew the graphs
However the measured THD+N of the Fireface is quite close to the manufacturers specifications (DA <0.001%, AD<0.00063%)
Last edited:
Aside from this 44.1k 80s technology.... If your DAC is fed with a natively recorded super Hi-res material like DxD (352.8) would it make sense to bypass all these up sampling processes/filtering and operate in true NOS mode?
Have you calibrated, your setup?
I see lower noise floor from zfe, almost 160dBFS with 44.1KHz sampling!. Your card, has THD+N, Min. Gain < -100 dB (0.001%) @ 1 kHz, -1 dBFS
There is no any point if you haven;t calibrate the signal, before the measurement. 150 dBFS noise floor is like spdif to spdif loop-back!
Chears,
In this case no, it was a quick and dirty match up of the level of the signal to the recorded signal.
The "noise floor" is directly related to the bit depth and the number of bins used in the FFT. It is not a measure of the level of the hardware noise floor.
I'm also using "time overlap" setting of x16 in RX4. Not sure if ARTA supports overlap, but it improves resolution of the FFT.
https://cas.web.cern.ch/cas/Denmark...r on overlapping FFT signals 2009 CAS2010.pdf
I'll have to dig out some different software for a calibrated measurement - can't use ARTA as Virtual Box doesn't support pass through of firewire.
Last edited:
Until the manual is ready, there is great info here: https://hifiduino.wordpress.com/2015/01/30/building-soekris-r-2r-dac/
All shipping products so far have rev 0.8 firmware.
If feed 352K/384K then the FIR1 filter is bypassed. I have the opinion that FIR2 is best to keep on, it's a short one operating at high sample rates.
But when I release the filter tools and docs, there will be instructions how to bypass that one too....
Talking about filters.... The really critical ones are the 44.1K and 48K ones. I have been researching a little on the net, and found out:
HQPlayer seems to be the reference and their best filters is what they calls "Poly-Sinc" filters, but I don't seems to be able to find any info what they are exactly.... Anybody know something ?
There is something called "Windowed Sinc" filters, have found some references and t.ex. the ScopeFIR package can generate them.
Anybody knows of other tools that can generate them, preferable no or low cost ?
Any other ideas to what the best filters actually are ?
HQPlayer seems to be the reference and their best filters is what they calls "Poly-Sinc" filters, but I don't seems to be able to find any info what they are exactly.... Anybody know something ?
There is something called "Windowed Sinc" filters, have found some references and t.ex. the ScopeFIR package can generate them.
Anybody knows of other tools that can generate them, preferable no or low cost ?
Any other ideas to what the best filters actually are ?
Thanks Soren, this is excellent!
Personally, I have this fuzzy feeling about upsampled music, you know trying to fill in missing bits with nulls or approximations and then trying clean out with elaborate filtering that will only color the sound whatever the technology. -I think
So it is interesting to see how this DAC performs with with DxD selections notably
from 2l -Nordic
Last edited:
I have a feeling this dac will turn out to be like fine wine. Right now it has just been bottled and tastes a little rough. It's potential is evident though. Time is needed for its full potential to be realised. I think the filters will be one of the key elements. They seem to be just place holders at present until the options are better understood. Soekris might like to collaborate with someone who can really contribute in this area in an open source way...
I am sure it will be. My own expectation, if it can outperform the 9018 DAC, I for one would change my position from sell to buy (or resell)
Which version? It would be helpful if everyone giving subjective impressions mentioned which version they are using
I2S is I2S. You have three connections, plus 3.3V power and gnd.
SCLK -> BCLK
LRCK -> LRCLK
SDIN -> DAT
You'll have to workout where to pick up 3.3V and GND to power the isolators.