On Mon, 23 Sep 2019 at 16:55, <erik@...> wrote:
The nanovna has a very strong 1KHz component - probably a result of the
1KHz codec frame-rate.
I opened the nanoVNA and connected the line in of a good USB audio device
to the reference SA612. But as I did not dare to do two connections the
unbalanced signal had a lot of noise. And there were 1kHz peaks independent
from the selected output frequency, possibly coming from the PC itself as
connecting the audio input to the ground of the nanoVNA gave the same noise.
Looking with a scope showed a very clean triangular 5kHz signal, although
with some jitter that might explain some of the disturbances we see.
As disconnecting the USB resets the nanoVNA U was not able to measure with
disconnected USB as I do not know how to stop the sweep (e.g.the writing to
the SI5351) from the UI
As another way to isolate the source of the 1KHz peaks I ran a test where I
adjusted the PLL in the tlv320 so it would generate a different sampling
rate. It was a one line change to tlv320aic3204.c, line 46:
I2CWrite(AIC3204_ADDR, 0x06, 0x08); /* J=8, was 10 */.
With the correct value, the frame-rate calculation is: 8.000MHz*10.7520 =
86.016MHz, 86.016MHz/(2*7*128) = 48kHz /48 = 1KHz frame rate
With the modified value, the frame-rate calculation is: 8.000MHz*8.7520 =
70.016MHz, 70.016MHz/(2*7*128) = 39.07kHz/ 48 = 814 Hz frame rate
With the correct value, the FFT at 1.23GHz is:
[image: image.png]
With the modified value, the FFT at 1.23GHz is:
[image: image.png]
The peaks are now approx. 800Hz, which closely tallies with the modified
frame-rate.
I think this is pretty good evidence that these peaks are related to the
frame-rate, not usb or Si5351. The strongest suspect is coupling from the
CPU, which does its heavy dsp processing once each frame
Rgds,
Dave
