Measurements of an ES9023 DAC

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Dave,

Hi HP,
What were you looking to see in this picture?

Dave

there are some symmetric waves: +/- ... 5Hz, 12Hz, 22Hz, 50Hz in the -120.. -130dB region.

We have also to consider the FFT size (256K), while as used in my measurement (2^25) this will lower the bin distance and the noise level / bin, but in the other hand clear modulated (symmetric around the carrier) will be visible.

Hp
 
Hi Gary,
Well, looking at the UPV manual, it states "AVG count" in the FFT section,
give it a try.
That function setting (average, expo0nential, flat...) works for the calculation measured parameter calculation (THD for example) in the firmware on the UPV I have, but does not affect the FFT monitor screen available. The post FFT function is not implemented in this firmware release, which I assume is where it would be...

I'm working on a really old firmware version with the UPV I have. Version 1.4. Updates to the current version 3.2 are possible but I have to step through one or two previous versions to get to the latest. I had been waiting to do so until I sent it in for an upgrade to add the digital interface incase there was an issue upgrading and it went down.

there are some symmetric waves: +/- ... 5Hz, 12Hz, 22Hz, 50Hz in the -120.. -130dB region.


We have also to consider the FFT size (256K), while as used in my measurement (2^25) this will lower the bin distance and the noise level / bin, but in the other hand clear modulated (symmetric around the carrier) will be visible.
I can see how the higher FFT point size and/or averaging is desirable. Each FFT screen update when zoomed at this level changes. There are no repeatable peaks as are obvious in the other captures I took. I will make the attempt to update the firmware to see if the added enhancements also include FFT features that will help with this measure.

Thanks,
Dave
 
Hi Hp, Gary,

Dave,

there are some symmetric waves: +/- ... 5Hz, 12Hz, 22Hz, 50Hz in the -120.. -130dB region.

We have also to consider the FFT size (256K), while as used in my measurement (2^25) this will lower the bin distance and the noise level / bin, but in the other hand clear modulated (symmetric around the carrier) will be visible.

Hp

Firmware update successful - Updated picture of the 44.1/16 Jtest signal with 256K points, averaging 10 measurements per and zoomed in to +/- 100Hz. The 9023 looks to have good data correlated jitter rejection in our implementation ... Still limited FFT point size with the tool I have either way.

Dave
 

Attachments

  • 441_16.JPG
    441_16.JPG
    61.4 KB · Views: 643
Looks like very, very low jitter levels that nobody needs to worry about.

Your THD and IMD measurements may be limited by the analyzer, at least assuming the DAC sticks to its -94 dB THD spec and the output stage doesn't mess that up. 2nd and 3rd at -90 dB each would be entry-level soundcard level (e.g. Xonar DG) at best.

It may be interesting to run the usual intersample-over tests, though I doubt the designers of a modern-day DAC digital filter would mess up there.
 
Hi Hp, Gary,

Firmware update successful - Updated picture of the 44.1/16 Jtest signal with 256K points, averaging 10 measurements per and zoomed in to +/- 100Hz. The 9023 looks to have good data correlated jitter rejection in our implementation ... Still limited FFT point size with the tool I have either way.

Dave

Hmm, looks like as from a different DUT. :confused:

Hp
 
Hi Sgrossklass

Your THD and IMD measurements may be limited by the analyzer, at least assuming the DAC sticks to its -94 dB THD spec and the output stage doesn't mess that up.
The THD floor is ~120db for the instrument I have.

The 9023 does not perform to -94 dB FS (2V RMS) @ 1kHz as far as I can tell, even when pretty optimally implemented. THD improves as the frequency rises and gets a little worse lower. I can verify this by measuring in 80KHz or greater bandwidth. My guess is that THD performance is a function of power in the internal opamp, or perhaps from the charge pump.

The follower buffer/LPF performs in the mid ninety dB range as implemented.

2nd and 3rd at -90 dB each would be entry-level soundcard level (e.g. Xonar DG) at best.
Agreed, but this sounds like no sound card I have used and measures better in every other way I have tested. @Juli and EMU0404 are all I have had to compare it to. Neither comes close to the detail, silence, or placement of instruments in the image while listening here. This is not to say that there is not something better...


I'm happy to defend the design as a whole though for its simplicity and performance vs. cost. Looking at the music in my library on an oscilloscope I have not found tracks with sustained levels at 2V RMS, or even 1.5V. In the 0.5 to 1V range the 9023 works very well and has the noise floor, dynamic range, and jitter performance to make it shine where most music I listen to is.
It may be interesting to run the usual intersample-over tests, though I doubt the designers of a modern-day DAC digital filter would mess up there.
Is there a standardized testing method for this?

Thanks,
Dave
 
Hi Hp,

Hmm, looks like as from a different DUT. :confused:

Hp
Nope, same one. I have two of these buit up identically and they both perform similarly. I am reasonably sure this is the same one used for the original testing as it has not left that bench and the other is in the system to listen to...

Here are the two measurements side by side again. Any peaks and troughs in the one on the left (no averaging) are random and move around with every sweep.

I did not get the grid set up the same as all of the presets went away with the upgrade. Apologies there.
Dave
 

Attachments

  • 441_16JtestZoom_100.jpg
    441_16JtestZoom_100.jpg
    134.9 KB · Views: 600
  • 441_16.JPG
    441_16.JPG
    61.4 KB · Views: 584
Hi Cersepn,

Well said Patrick...

I've spent what little time I have had for fun stuff lately working on the FX5.

I left off comparing component changes in the follower/LPF and looking forward to comparing the 9022 vs. the 9023. These are in work and I'm sure to will get back to them in the next few weeks. I also want to do some critical listening to compare the simple filter output to that of the follower/LPF.

We'll post more details as we get there and I'm happy to answer any questions, as I know Patrick is as well.

Dave
 
First of all, great job Dave!

nice and solid implementation you have realized.. I do agree ess9023 is sounding very good, indeed. wide stage, good sonics.

my setup is very simple using separate regulator for ess9023 and batteries for local clock (running a 60Mhz standard oscillator). i2s feed using a modified hiface.

I'm testing different caps and found your recommendation post back in June http://www.diyaudio.com/forums/digi...ng-new-ess-vout-dac-es9022-4.html#post2616741

did you change anything since then?


If it proves to work as well as it seems to, I'll share some details.
Dave

interested in your PSU setup and filtering section.. I see TO220 fixed to the bottom case in the middle section of the board, are those pre-regs?

kind regards
Valeriano
 
Valeriano,

Thank you for the compliment. I enjoy the DAC very much. It is easy to build when using the recommendations from that link.The only challenge is the size of the SMT components!

Not much has changed in the implementation from the post you referenced. Right now (as time allows anyway) I am listening to and measuring different imnplementations of the output filter/buffer and am going to compare the 9022 to the 9023 next to see if there are differences that can be measured or seen. I am also collaborating with Patrick (EUVL) to see if there are differences we can detect between all MLCC and all film bypass construction.

The line PSU is pretty straight forward. It uses conventional toroids with additional HF noise filtering on the secondary side before and after the rectifers. The TO220 devices are mosfets used as you suspected, pre-regulators that also substantially reduce the ripple and improve rejection in the audible range and above.

I am probably going to change to a different output filter/buffer and go to one transformer. Using 2SK170s for the follower reduces the voltage necessary to achieve acceptable distortion from the buffer, eliminating the need for the higher voltage supply and the second transformer...

I'll post some more when I have it.
Dave
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.