Review and Measurements of Allo Katana and ApplePi Raspberry Pi DACs

[March Audio]

The amazon listing and my website both link to clear install instructions as well as stating that the header and standoffs are not included.
The install instructions can be found here:
https://drive.google.com/file/d/1b6Y6fvdK50TgxC3CAdWk2_2lO_vVovy9/view

Can you please let me know what instrument you are using to perform the measurements? I have verified the 128dB of SNR (A-Weighted) with both an Audio Precision APx515 and ATS-2. Also the 128dB would be for the balanced output, the RCA output will be 125dB.

Here is 24-bit silince A-Weighted (Mini-XLR Ouput 5Vrms is 0dB):
View attachment 14970

Here is SNR A-Weighted
View attachment 14968

Hi

Amir uses an APx555, but you are not comparing the same measurements here.

FOR DIGITAL CONVERTER MEASUREMENTS…
We recommend using this AES17 dynamic range measurement. It is intended specifically for ADC (analog-to-digital converter) and DAC (digital-to-analog converter) dynamic range and “noise in the presence of signal” measurements, as described in Section 9.3 of AES17. A similar method is defined in IEC61606.

This method differs from standard signal-to-noise and dynamic range measurements in that it uses a –60 dBFS stimulus during the noise measurement. This method is used for two reasons:

· In both ADCs and DACs, “idle tones” can be produced within the converter in the absence of applied signal. In the method here, a low-level tone is applied to the converter to avoid production of idle channel noise. The low-level tone is removed by a notch filter before measurement.

· In some DACs, the output of the device is switched off when there is no signal, providing an unrealistically quiet measurement. The low-level tone (again, notched out before measurement) defeats this muting mechanism.

At –60 dBFS, the tone is so low that any distortion products created are below the noise floor.


The 1794 does indeed support zero mute, is it active in your implementation? If so your measurement indicates the noise floor of the analyser.

Zero Detect When the PCM1794 detects that the audio input data in the L-channel and the R-channel is continuously zero for 1024 LRCKs in the PCM mode or that the audio input data is continuously zero for 1024 WDCKs in the external filter mode, the PCM1794 sets ZERO (pin 13) to HIGH.

The data sheet does use the AES17 method and shows this plot

Its a 32k FFT so FFT process gain is about 42 dB if I have done the maths right (highly unlikely) . So that would make Amirs numbers about right.

FFT gain = 10*log(M/2) where M= number of FFT points.

 

[orchardaudio]

Hi

Amir uses an APx555, but you are not comparing the same measurements here.

FOR DIGITAL CONVERTER MEASUREMENTS…
We recommend using this AES17 dynamic range measurement. It is intended specifically for ADC (analog-to-digital converter) and DAC (digital-to-analog converter) dynamic range and “noise in the presence of signal” measurements, as described in Section 9.3 of AES17. A similar method is defined in IEC61606.

This method differs from standard signal-to-noise and dynamic range measurements in that it uses a –60 dBFS stimulus during the noise measurement. This method is used for two reasons:

· In both ADCs and DACs, “idle tones” can be produced within the converter in the absence of applied signal. In the method here, a low-level tone is applied to the converter to avoid production of idle channel noise. The low-level tone is removed by a notch filter before measurement.

· In some DACs, the output of the device is switched off when there is no signal, providing an unrealistically quiet measurement. The low-level tone (again, notched out before measurement) defeats this muting mechanism.

At –60 dBFS, the tone is so low that any distortion products created are below the noise floor.


The 1794 does indeed support zero mute, is it active in your implementation? If so your measurement indicates the noise floor of the analyser.

Zero Detect When the PCM1794 detects that the audio input data in the L-channel and the R-channel is continuously zero for 1024 LRCKs in the PCM mode or that the audio input data is continuously zero for 1024 WDCKs in the external filter mode, the PCM1794 sets ZERO (pin 13) to HIGH.

The data sheet does use the AES17 method and shows this plot

View attachment 14972

Its a 32k FFT so FFT process gain is about 42 dB if I have done the maths right (highly unlikely) . So that would make Amirs numbers about right.

FFT gain = 10*log(M/2) where M= number of FFT points.

I agree, the measurements that I did and the measurements from this review are not the same measurements.

I do however stand by my 128dB SNR measurement. This was performed on the balanced output with a 5Vrms (0dBFS) reference, with A-Weighting.

 

[March Audio]

I agree, the measurements that I did and the measurements from this review are not the same measurements.

I do however stand by my 128dB SNR measurement. This was performed on the balanced output with a 5Vrms (0dBFS) reference, with A-Weighting.

All measurements need to be qualified with information about the method otherwise you may not be comparing apples with apples.

BTW Is the mute active?

 

[orchardaudio]

If so your measurement indicates the noise floor of the analyser

The noise floor of the analyzer is much better than 128dB. With the input shorted the noise measured was 870nV, that is equivalent to 20*log(0.000000870/5) = 135.2dB.

With the ApplePi DAC the noise measured was 2uV this equates to 128dB.

If you subtract the noise from the analyzer, the actual SNR is even better.

 

[March Audio]

About 6dB, much better? But yes you are right, even if the mute is active the I to V / D to SE op amps will still be operational creating some noise.

Have you established if the mute operates in your design?

I'm not sure what you mean by subtract that analyser noise. The dac noise exceeds it.

Don't get me wrong, this is a great performing dac, just establishing the conditions.

 

[orchardaudio]

I'm not sure what you mean by subtract that analyser noise

When the analyzer measures noise it measures it's own noise with the noise of the device. It's own noise is 870nV. The noise measured with the DAC is 2uV. Since noise is not fully additive you have to use RSS method.

If you subtract the 870nV from 2uV using RSS method you get 1.8uV or 129dB.

Have you established if the mute operates in your design?

Mute is implemented.

 

[March Audio]

So the analyser noise isn't really a significant issue, which was really my point.

So with mute implemented the measurement isn't really valid. Try applying a very low signal level, maybe - 138dB 1kHz to keep the mute off and see what the noise floor/DR is like.

 

[helloworld]

With the ApplePi Volume-Clocker this jitter should be completely eliminated.
View attachment 14971

With this clocker, does the thd+n performs better? Did you test that?

 

[orchardaudio]

With this clocker, does the thd+n performs better? Did you test that?

THD+N stays the same, the Volume-Clocker helps with jitter and adds hardware volume control.

 

[bennetng]

THD+N stays the same, the Volume-Clocker helps with jitter and adds hardware volume control.

How about send it to Amir for measurement?

 

[orchardaudio]

So the analyser noise isn't really a significant issue, which was really my point.

So with mute implemented the measurement isn't really valid. Try applying a very low signal level, maybe - 138dB 1kHz to keep the mute off and see what the noise floor/DR is like.

I no longer have access to an analyzer but you can see the change in noise from these two images below:

Silince

-60dBFS signal

Between these two you can see the noise floor only slightly goes up, hence the mute function has very little effect.

 

[orchardaudio]

How about send it to Amir for measurement?

I can send for measurement.

 

[orchardaudio]

@amirm send me a message if you would like the ApplePi Volume-Clocker for evaluation.

 

[orchardaudio]

I can send for measurement.

ApplePi Volume-Clocker is on the way.

 

[dc655321]

Here is another way to interpret your chart:
View attachment 14982

I think you're missing some items on the left side of your chart.

 

[March Audio]

I no longer have access to an analyzer but you can see the change in noise from these two images below:

Silince
View attachment 14975

-60dBFS signal
View attachment 14976

Between these two you can see the noise floor only slightly goes up, hence the mute function has very little effect.

Thanks for posting these plots. That's a good result. What is the Fft resolution?

It looks like it ties up with the datasheet which means Amir's measurements are about right; about 42 dB Fft process gain assuming a 32k fft

 

[orchardaudio]

@BE718

It looks like it ties up with the datasheet which means Amir's measurements are about right; about 42 dB Fft process gain assuming a 32k fft

I believe that it is 32k but I don't remember as that was done quite some time ago, it would not have been less than 32k.

I did find another plot with -90dBFS input:

 

[derp1n]

With the ApplePi Volume-Clocker this jitter should be completely eliminated.
View attachment 14971

Why not just make the DAC have lower jitter instead of upselling extra parts?

 

[derp1n]

Here is another way to interpret your chart:
View attachment 14982

By that intellectually dishonest metric, the $8 CM108 Bare Board USB DAC destroys you at 8.9dB per dollar.

 

[orchardaudio]

Why not just make the DAC have lower jitter instead of upselling extra parts?

The ApplePi DAC is hat compliant, that means it has to be that size. It is not possible to add the reclocking and HW volume control to the board and keep it the same size.