DIY Audio Analyzer with AK5397/AK5394A and AK4490

Member
Joined 2009
Paid Member
I have made a measurement of the 1 and 10 kHz distortion in loop-back.
The generator level is -10dBFS for all measurements. I have attenuated the input by 10dB. Otherwise the distortion from the ADC would dominate the picture.

There is only a small variation in the distortion when using different sample rates.

But to my surprise the distortion at 10kHz is lower than the distortion at 1 kHz :confused:
At 48kHz sampling there is of course only one harmonic distortion component, but that is not the whole explanation.
 

Attachments

  • 1 and 10 kHz distortion at -10dBFS signal output.pdf
    145 KB · Views: 446
Member
Joined 2009
Paid Member
I just made a new set of measurements at 10 kHz. Previously my analyzer had been set up to only include 10Hz - 20kHz in the calculations. In the new measurements the upper limit has been increased to 100kHz. This does change the numbers, but they are still better at 10kHz than at 1kHz.
 

Attachments

  • 1 and 10 kHz distortion at -10dBFS signal output_2.pdf
    146 KB · Views: 334
Member
Joined 2009
Paid Member
As requested, here are some measurements of the THD vs frequency and THD vs level.

The THD vs frequency was measured at an output level of -10dBFS, equivalent to around 3.5Vrms. The input attenuator was set at 10dB attenuation. Thereby the ADC also got around -10dBFS. Levels around -10dBFS generally work well with the AKM converters I have used in this design (AK5397 and AK4490) and the converters I have used in a previous design (AK5394A and AK4399).

The distortion seems to be around 0.0002% up to around 15 kHz. Above that there is some increase in the distortion.

I have included measurements of the THD vs level for 48, 96 and 192 kHz sampling. For some reason the 48 kHz curve does not get down to the same minimum as the 96 and 192 kHz curves. At the moment I have not investigated why.
 

Attachments

  • Distortion_-10dBFS_left channel_192k sampling_input-att10dB_150813_4.png
    Distortion_-10dBFS_left channel_192k sampling_input-att10dB_150813_4.png
    17.6 KB · Views: 1,836
  • Distortion vs. amplitude_left channel_48kHz sampling_150813.png
    Distortion vs. amplitude_left channel_48kHz sampling_150813.png
    9.6 KB · Views: 1,829
  • Distortion vs. amplitude_left channel_96kHz sampling_150813.png
    Distortion vs. amplitude_left channel_96kHz sampling_150813.png
    9.7 KB · Views: 1,822
  • Distortion vs. amplitude_left channel_192kHz sampling_150813.png
    Distortion vs. amplitude_left channel_192kHz sampling_150813.png
    9.8 KB · Views: 1,802
Last edited:
Dear JensH,


"For some reason the 48 kHz curve does not get down to the same minimum as the 96 and 192 kHz curves. At the moment I have not investigated why."

Just a hint, though I'm sure you have arranged for it OK:
The problem with my AK4490 board was that the dac has 'manual mode' setting for the sample rates as a factory default. One has to force the ACKS pin "High" for "auto" sample rate recognition. If by chance it would be left floating than it is internally pulled down 'low', manual. In such a case the actual rate setting is defined by dif0;dif1;dif2 pins. It could be that they point at 'double mode' sampling in your case, and so resulting (slight) distortions at 'normal mode', like 48k.. but OK at 96, 192k.

Just a hint, again.

Ciao, George
 
Member
Joined 2009
Paid Member
Hi Sonny,

No, I haven't seen that. Where can I find it?

On the outputs of my design I have DC offsets of 6.4mV on one channel and 16.8mV on the other channel. This is after the post-DAC filter and output stages. Signal level is 10Vrms balanced at 0dBFS.
 
Just short...

http://www.diyaudio.com/forums/digital-source/278507-some-dc-measurements-ak4490-outputs.html

It is measured directly on the outputs of the DAC. Before filter. I wanted to split it up and remove the suggested DC blocking CAP of 100uF. The minimum load impedance to ground should be more than 2K. My first thought was to add an MFB network with 0dB gain. For this in theory i needed an virtual ground for the opamps at 1/4 the reference voltage.

But that is not the case... Not with 155mV offset.

- Sonny
 
Yes i would suspect that you uses the filter type from figure 40 or 42. I wanted to go fully balanced right after the DAC.

Meaning double filter of figure 40. But then we get below 1K AC impedance that they suggest. If the filter gets changed to 2x resistance values the noise level rises.

That is why i was investigating the other solution.

I have around 4 - 6mV DC offset with Figure 40 & 42 filter but i also uses a JFET type opamp,,,
 
Member
Joined 2004
Paid Member
Jens has done a remarkable job on this. Especially getting such good performance on first pass.

I have issues using opamps for the initial low pass/reconstruction filter for a DAC. These all have a lot of high frequency energy in the output. Usually full scale for several multiples of the sample rate. I tried a passive L/C filter which has two benefits. First, no HF overload. Second, it unloads the DAC at high frequencies with an inductive input. That both reduces the current requirements inside the DAC and the demands + supply and ground noises to support the DAC. In my testing it worked great. I can dig out the details if anyone is interested.
 
Member
Joined 2009
Paid Member
I think that the high frequency noise is less of a problem with the AK4490 than with some other DAC's. I assume that this is due to the switched capacitor filters used in the DAC. I have seen the same on the AK4399 and the Cirrus Logic CS4398. In contrast to this, I have seen a lot of high frequency energy on the output of the TI PCM1794.

I can check this tonight when I get home.
 
I think that the high frequency noise is less of a problem with the AK4490 than with some other DAC's. I assume that this is due to the switched capacitor filters used in the DAC. I have seen the same on the AK4399 and the Cirrus Logic CS4398. In contrast to this, I have seen a lot of high frequency energy on the output of the TI PCM1794.



I can check this tonight when I get home.


I have used the pcm1794a too... I Can only agree with you.


Sent from my iPhone using Tapatalk
 
Member
Joined 2004
Paid Member
This is the output spectrum on an unfiltered AK4399 DAC 0-5 MHz, 10 dB/div. The stuff continues up. The lower trace is whats left after the passive filter shown. It should be pretty low distortion (at least from my measurements) but its another round of new stuff.
 

Attachments

  • AK4399 spectrum 0-5MHz.jpg
    AK4399 spectrum 0-5MHz.jpg
    724.8 KB · Views: 1,784
  • DAC filter.PNG
    DAC filter.PNG
    39.5 KB · Views: 1,720
  • DACoutput full diff.asc
    3.9 KB · Views: 164