TOPIC: NanoAVR HDA Noise Floor

Hello zmaile,

thanks for your helpful reply!
We have in fact very good equipment at work (I am an analog electronics engineer), including spectrum analyzers and 4GHz MSOs, but none of them is audio-related, so "my" Keysight MSOS 404A does decode the I2C, UART and SPI but not the I2S protocol. So there is some effort to prove the HDA's behaviour at work, for example an appropriate hdmi source is missing.

On the other hand I am very interested to find the real source of these disturbances. The I2S protocol seems to be straight so it should be easy to observe as you stated.

The USB protocol includes 1ms frames or 125us microframes (full speed), so this could be a source. Otherwise the HDA should't generate these frames as it does not act as a master and the host (PC) was connected via LAN, not via USB.

Cheers, Timo

Hello,

here are some additional nanoAVR HDA measurements regarding the noise floor:
link to the measurements



It would be nice to read a comment from devteam.
- Is there a possibility to change the filter from fast to slow rolloff? Would this be helpful for lower noise?
- Do you know any other possible changes in the future to get lower noise?

The nanoAVR HDA is a very nice solution for my requirements of course, at least the DSP-part including the filters. Thank you for that development!

Cheers,
Timo

Hello,
because of requests for pictures I put some insights of the nanoAVR HDA here:
https://ibtk.de/project/amplifier/nanoAVR_HDA/Fotos/20180411_MiniDSP_nanoAVR_HDA_fotos.html
The device uses the miniDAC8 board, which is available on the miniDSPs website too. So it should be possible to replace it with a DAC/balanced output amplifier board with better SNR data in principle.
There exist some ideas, LTSpice simulation files and Eagle files using the ES9038pro and OPA1612 already but it's still a lot of work.
Is there anybody out there, interested in the realization of an alternative DAC board together with me (would need some experiences in low noise analog design and especially µC design and programming)?
Cheers, Timo

@ Timo

These measurements are very odd and suspicious.. How did you make them? We've grabbed 5 x nanoAVR HDA on the bench on the Audio Precision APX and can't get any to behave like this.. I wonder if something is causing a loop/artifact in your measurement. Removing the USB cable to the nanoAVR might break that loop as you're doing a loopback. Just a thought.

as for the nanoAVR, you can look at the miniSHARC, that's similar pin out. All this is outside our tech support structure though..

DevTeam

Hello devteam,

yes, I was surprised by these pictures too. Especially by the differences between the Fireface UC and the nanoAVR HDA measures, respectively.

A sample of the noise taken by an iphone 6S, simply placed directly in front of the short horns of the main speakers:



Watch it twice if necessary, to hear the 2Hz-"rhythm" in the noise. This behaviour is visible in the measurements also, see the pictures 3.08 to 3.10: the narrow pinches in the real time curves (0.5s repetition rate) and the 2Hz spectrum peak in 3.12.

Mesurement setup:



There is no USB-connection between the nanoAVR HDA and the PC nor to any other device. The LAN-connection is still present for parameterisation, of course. But the LAN connection should be isolated (on both ends) by definition, isn't it?
The other nanoAVR outputs are connected to the 8ch-amplifier with NC252MP by pseudo-balanced cabling (Cinch to XLR). There is no hum to be heard from the Speakers. I did not measure the spectral content though.



I repeated the measurement some days before, with a similar result unfortunately:



with a greater resolution:



The sample rate is reduced to 48kHz to avoid eventually possible aliasing effects.
Measurement are done using the latest REW revision.
00 to 10
- The nanoAVR HDA's output connected to the (Line-)input of the RME Fireface UC.
- RME's output 1 is connected to its's input 4 as a loopback (not included in the sketch).
11
- The RME's input 3 is left open
- RME's output 1 is connected to its's input 4 as a loopback.
12
- The RME's input 3 is shorted
- RME's output 1 is connected to its's input 4 as a loopback.
13
- The RME's input 3 is shorted, no loopback
14
- The RME's input 3 is shorted, no loopback
- The nanoAVR is switched off

All measures 11 to 14 lay well below 00 to 10 and show none of the artifacts.
Because the RME is fully independent from the nanoAVR, I do not see any reasons for an interaction, which could produce such distortions. The output of the nanoAVR is analog (DAC + low pass), the input of the RME is an opamp, followed by the ADC. This structure is a good decoupling against digital interaction. The RME itself does not produce these artifacts as to be seen in the measurements 11 trough 14.

As shown in the document's pictures 3.14 to 3.16, the I2S is quiet below -100dBFS.
I believe, these points are a good reason to acclaim, that the DAC of the nanoAVR HDA produces thes artifacts, resulting in a relatively high noise floor, at least in my device. As I stated in the document, it may relate to the used internal filter of the DAC.

I would like to get another example of the nanoAVR HDA as a substitute to repeat this measurement. Do you think, that this procedure would be possible? In this case you could measure my device with your AP system too.

Thank you!
Regards, Timo