AP wins! 
I recall I had to do a trick to get the graph posted above and maybe there the mistake hides... If I set the DAC as the output device in ARTA then it would follow the sampling rate of the input device and since this was set at 192kHz to allow a view at higher frequencies, then NOS was looking like oversampling. To force the DAC operate at 44,1kHz I fed it from audacity.
I recall I had to do a trick to get the graph posted above and maybe there the mistake hides... If I set the DAC as the output device in ARTA then it would follow the sampling rate of the input device and since this was set at 192kHz to allow a view at higher frequencies, then NOS was looking like oversampling. To force the DAC operate at 44,1kHz I fed it from audacity.
I am well served.
Thank you for the clear description.
Enjoy the aural outcome.
George
I have the pleasure to see, touch, test and listen to this DAC for a little while
Here is the behaviour in freq. domain of the three modes of operation when DAC is fed with white noise at six different sampling rates.
"Passive" means PCM1794 NOS output in series with a 9th order Chebyshev LC filter (Abraxalito's Celibidache DAC passive filter).
"Off" means PCM1794 NOS output
"Digital" means PCM1794 oversampling digital filter output.
The DAC USB input is fed from a laptop PC and the DAC analog output is connected to an ADC which feeds another PC.
Both PCs are running REW, each set to the specific ASIO driver for the DAC and the ADC respectively.
ADC is set to run at 192kHz 24bit.
DAC is set to different sampling rates at 16bit
George
Here is the behaviour in freq. domain of the three modes of operation when DAC is fed with white noise at six different sampling rates.
"Passive" means PCM1794 NOS output in series with a 9th order Chebyshev LC filter (Abraxalito's Celibidache DAC passive filter).
"Off" means PCM1794 NOS output
"Digital" means PCM1794 oversampling digital filter output.
The DAC USB input is fed from a laptop PC and the DAC analog output is connected to an ADC which feeds another PC.
Both PCs are running REW, each set to the specific ASIO driver for the DAC and the ADC respectively.
ADC is set to run at 192kHz 24bit.
DAC is set to different sampling rates at 16bit
George
Attachments
Last edited:
George, thank you very much.
This is much better than I could do with one PC and Audacity to the rescue. @abraxalito 's filters are very consistent. "Passive" output is my default now. The biggest surprise is the behavior of the digital filter with rising sampling rate. I wasn't able to see that with my "method". Once again, great work!
This is much better than I could do with one PC and Audacity to the rescue. @abraxalito 's filters are very consistent. "Passive" output is my default now. The biggest surprise is the behavior of the digital filter with rising sampling rate. I wasn't able to see that with my "method". Once again, great work!
You are welcome Hayk and Kostas.
Now, using the same test setup, only reducing the test signal (white noise) to -100dBFS, I captured the noise skirt of the DAC output, up to ~100kHz.
The effect of the ADC's noise shaping (see last attachment) over all test FFT screenshots becomes obvious.
The spikes at 75Hz and it's harmonics are there, regardless of sampling rate change on both DAC and ADC, even use of one or two PCs or change of drivers.
George
Now, using the same test setup, only reducing the test signal (white noise) to -100dBFS, I captured the noise skirt of the DAC output, up to ~100kHz.
The effect of the ADC's noise shaping (see last attachment) over all test FFT screenshots becomes obvious.
The spikes at 75Hz and it's harmonics are there, regardless of sampling rate change on both DAC and ADC, even use of one or two PCs or change of drivers.
George
Attachments
-
(10) ADC_open input (192k_24bit).jpg -
(9)Test Settings_-100dBFS.jpg -
(8) Preferences DAC side.jpg -
(7) Preferences ADC side.jpg -
(6) 192k_White noise-100dBFS .jpg -
(5) 176k4_White noise-100dBFS .jpg -
(4) 96k_White noise-100dBFS .jpg -
(3) 88k2_White noise-100dBFS .jpg -
(2) 48k_White noise-100dBFS .jpg -
(1) 44k1_White noise-100dBFS .jpg
The DAC is sigma-delta and it shows... Maybe it sounds so too, although I have not a remarkable back round with multi-bit.
The 75Hz spikes is a mystery. I'm happy to see they appear at this frequency in your tests. In my humble attempts they are at 100Hz and up like rectifiers' noise but in a completely shunt regulated psu.
So, now I know it's something else.
The 75Hz spikes is a mystery. I'm happy to see they appear at this frequency in your tests. In my humble attempts they are at 100Hz and up like rectifiers' noise but in a completely shunt regulated psu.
So, now I know it's something else.
I repeated the tests of post #47 (noise skirt of the DAC output, up to 100kHz), this time using an SDR insted the ADC of the soundcard.
Now the horizontal scale is linear.
The 4kHz apart spikes btb 8kHz and 48kHz are part of the SDR noise floor as seen in att (3)
Now the horizontal scale is linear.
The 4kHz apart spikes btb 8kHz and 48kHz are part of the SDR noise floor as seen in att (3)
Attachments
Now, after using the SDR dongle to test DAC in idle, I fed the DAC with 19kHz+20kHz 1:1
Here 0-100kHz BW FFTs
Here 0-100kHz BW FFTs
Attachments
Something odd about the PCM1794 digital filter and also about the SDRPlay measurements.
Here is my AK4493 at 44100Hz with sharp filter and CCIF -3dBFS. ADC is ES9822PRO. (Dismiss the distortion numbers as DAC and ADC used separate REW instances, so the numbers are for THD, not IMD).
Here is my AK4493 at 44100Hz with sharp filter and CCIF -3dBFS. ADC is ES9822PRO. (Dismiss the distortion numbers as DAC and ADC used separate REW instances, so the numbers are for THD, not IMD).
Attachments
Last edited:
What I see Kostas is that the passive LC filter does a fine job in clearing the ultrasonic spectrum, better than the digital filter.
Not to speak about the unfiltered (NOS) clutter.
OK, speakers can't reproduce these, our ears can't hear these. But the electronic gear downstream the DAC may find it difficult to process the audio signal without intermodulation or even show signs of instability.
If we look at the 100kHz BW FFTs at post #52, with the unfiltered signal (Off) the area below 20kHz is full of high level spikes.
This is a hint that my SDR is a victim of intermodulation due to the unfiltered ultrasonic spectrum.
It is surprising but I have no other explanation.
What I could do was to repeat the test with my soundcard in place of the SDR
The attached soundcard FFTs do not show that severe IMD that the FFTs of the SDR show.
This is an (unintended) case of 'proof by example'
George
Not to speak about the unfiltered (NOS) clutter.
OK, speakers can't reproduce these, our ears can't hear these. But the electronic gear downstream the DAC may find it difficult to process the audio signal without intermodulation or even show signs of instability.
If we look at the 100kHz BW FFTs at post #52, with the unfiltered signal (Off) the area below 20kHz is full of high level spikes.
This is a hint that my SDR is a victim of intermodulation due to the unfiltered ultrasonic spectrum.
It is surprising but I have no other explanation.
What I could do was to repeat the test with my soundcard in place of the SDR
The attached soundcard FFTs do not show that severe IMD that the FFTs of the SDR show.
This is an (unintended) case of 'proof by example'
George
Attachments
I'm not sure that it is about the signals. In SDR 44k1 noise graph the digital filter output has a strange step rise in the noise floor although according to your previous ADC measurement the noise floor is quite flat upto 100kHz.