Here is the AK4490 dac output of 88kHz sine at 384kHz.
/Martti
Please can you run just a single shot (if you scope allows that)? But it looks just like our distorted waveforms...
I set just these two at startup:
Control1 -> 0x0f
Control4 -> 0x02
Thanks. So you have oct mode and I2S 32bit, IIUC.
I reconfigured my clock gen and I2S output to I2S 32bit (used 24bit previously), no change in the output waveform.
Series of waveforms at fs/4, from 48kHz to 384kHz (NOS filter mode selected for <=192kHz). Direct chip output.
Klaus, that looks like they use MCLK=FSx64 (=>oct/hex) for all samplerates in this mode. Otherwise I do not see how they could get those distortions on <384kHz samplerates. But that still requires 2x oversampling (32bit sample x 2 => 64bits in sigma-delta.
Control 4 (05h), Super Slow Roll-Off Filter Enable (bit 0)NOS mode - how do you switch AK4490 to that? I do not see any register setting...
I don't think this is distortion, rather it is fully expected behaviour.Klaus, that looks like they use MCLK=FSx64 (=>oct/hex) for all samplerates in this mode. Otherwise I do not see how they could get those distortions on <384kHz samplerates. But that still requires 2x oversampling (32bit sample x 2 => 64bits in sigma-delta.
I'll check MCLK pin (a bit difficult to access in the RME).
Please can you run just a single shot (if you scope allows that)? But it looks just like our distorted waveforms...
No can do. Single shot does not seem to work. I need to get a new scope.
bohrok2610 - please what is your analog filter cut-off freq?
Fc=99.2kHz. Straight from the AK4490 datasheet (figure 40). However AK4137 also has a filter so I'm running 2 filters.
Control 4 (05h), Super Slow Roll-Off Filter Enable (bit 0)
Are you sure? AK4493 manual says:
Super Slow Roll Off(Digital Filter bypass mode)
Setting/resetting that bit makes no visible change on my output waveform. IMO it just bypasses the cutoff filter, not oversampling. That would explain why I see no change on sines well below fs/2.
I don't think this is distortion, rather it is fully expected behaviour.
What would then be the point of oct vs. hex, when no oversampling occured in either case?
Control 4 (05h), Super Slow Roll-Off Filter Enable (bit 0)
I tried that but no change to output. It may well be that AK4137 filter cleans things up in my case.
It's not NOS mode literally. Still oversampling but rather than filtering, they just repeat the same sample value N times which is effectively NOS.Are you sure?
Last edited:
I tried that but no change to output.
Same with me. But your DAC output is same as mine (in oct mode), it's your analog filter which cleans the signal. My analog filter is at 440kHz because I am testing higher frequencies. I am afraid you are getting the very same signal. Please try 96kHz instead of 88kHz, that will keep your output waveform static.
It may well be that AK4137 filter cleans things up in my case.
Your resampler outputs clean 88kHz at 384kHz. Just like my signal generated directly at 384kHz by sox. The filter in AK4137 has really no effect on that, it's part of the x2 resampler.
Same with me. But your DAC output is same as mine (in oct mode), it's your analog filter which cleans the signal. My analog filter is at 440kHz because I am testing higher frequencies. I am afraid you are getting the very same signal. Please try 96kHz instead of 88kHz, that will keep your output waveform static.
You are correct. At 96kHz the dac output looks like yours. However the filter output looks like a clean sine so I'm not sure if there is any issue.
My analog filter is at 440kHz because I am testing higher frequencies.
All the datasheet filter characteristics are only up to 192kHz. Maybe there really is an issue with higher sampling rates. Have you tried lowering the cut-off of you filter to e.g. 200kHz?
It's not NOS mode literally. Still oversampling but rather than filtering, they just repeat the same sample value N times which is effectively NOS.
You are right. So let's call that setting NOS.
But I am still totally confused. What does the oversampling actually do?
NOS at 192kHz samplerate figure 4 (signal fs/4 48kHz) - does it mean the DAC does no oversampling starting from 192kHz? Because I do not see any midpoints added, only the original 192kHz samples (4 per period).
Yet with NOS off (filter slow instead of super-slow) the output is properly resampled, nice sine. What samples does the digital filter actually filter?
The output conversion runs always at 12.5MHz, regardless the MCLK. 12.5MHz/32bits = 384kHz - x2 oversampling must have occured, IIUC. But what was oversampled to what?
BTW the final-conversion rate stays at 12.5MHz even for 768kHz, where the 32bit bitrate is already 24MHz - that means the DAC must be dropping precision
Fig1 - final conversion for 192kHz, MCLK x256 49MHz
Fig2 - final conversion for 192kHz, MCLK x128 24.5MHz
Fig3 - final conversion for 768kHz, MCLK x64 49MHz
In all cases it is 12.5MHz.
Attachments
- Home
- Design & Build
- Equipment & Tools
- AK4493 at oct/hex sampling speeds - oversampling filter issue?