Phase is shifting slowly between two RTL-SDRs in sync

[saltmakrell]

Hi,

I've connected two RTL-SDRs together coherently as described here http://www.rtl-sdr.com/passive-radar-du ... l-rtl-sdr/ .

The strange thing I observe is that the first time I start them I will observe a slowly increasing phase shift between the two devices. I'm correlating the signals multiple times over time. This indicates that their frequencies are not exactly the same.. If I stop the script and restart the script, this strange phase shift usually (but not always) disappears, and it works perfectly and can run for a long time without any phase shift appearing, and I can use it for measurements.

I'm setting up the devices the same way using GNU radio and osmosdr:

def setup_device(dev):
dev.set_freq_corr(0, 0)
dev.set_dc_offset_mode(0, 0)
dev.set_iq_balance_mode(0, 0)
dev.set_gain_mode(False, 0)
dev.set_gain(10, 0)
dev.set_if_gain(0, 0)
dev.set_bb_gain(0, 0)
dev.set_antenna("", 0)
dev.set_bandwidth(0, 0)
dev.set_sample_rate(samp_rate)
dev.set_center_freq(freq,0)

Can some of you suggest what I may be missing? Is there anything in the RTL2832 or the tuner chip that prevents the frequency from being exactly the same even when the reference clock is the same ? :s This is really frustrating since I can't make some sense out of it. I've confirmed that I'm not loosing any samples.

Cheers
Ruben

[rtlsdrblog]

I'm not too well versed in using the dongles this way, but from what I've seen in discussions is that the shift is partly due to 'dithering'. Keenerds drivers seem to have a way to turn this off. Also another source of drift is because it is not possible to synchronise USB samples.

Have a look at how the guy in this post solved it http://www.rtl-sdr.com/an-rtl-sdr-phase ... on-finder/

[saltmakrell]

Thanks a lot for leading me the right way. I turned off the dithering on both devices by patching with this: https://github.com/keenerd/rtl-sdr/comm ... 804a063bfd and making sure priv->disable_dither is set to 1 when loading it via the GR osmosdr interface. After doing this, I don't observe any phase drift anymore.

Cheers!

[rtlsdrblog]

Excellent, glad to hear that the dithering solved it.

Just curious, are you not seeing any issues due to the inability to sync samples from the USB packets? I think many people have issues with that and it can only be solved with an external reference signal. Perhaps your application is immune to this desync somehow though.

[saltmakrell]

I get sample sync from the two devices by initially running a cross
correlation with the devices tuned into a known FM broadcast station.
I've implemented a gnuradio-block to do this for me automatically and
then adjust the input streams accordingly afterwards. I will share
this one when it's robust enough.

I would expect the dithering to cause random fluctuations of the
phase, but in this case there is a constant drift only one way. I'm
not sure I understand why this happens. And also why does it sometimes not happen..

[rtlsdrblog]

So in the end disabling dithering didn't completely fix it? I think the guy in the link I posted found that he needed to periodically resyncs the streams with a reference signal.

Go on the ##rtlsdr channel in IRC at freenode, there are a few experts there who might be able to provide insight.

[saltmakrell]

> So in the end disabling dithering didn't completely fix it?

It DID fix it completely! I haven't been clear enough then. I just found it a little bit strange that it was necessary to turn off the dithering. I wasn't expecting the dithering to have this effect.

When I've tuned the initial sample and phase offset, it can run for a long time without any need to retune it. It works very well!

Thanks for your help!

Cheers

[evso9816]

Hi man,
It has been longtime you worked on this dither process (since 2016). I followed the link mentioned but not really helpful.
Can you review the process of disabling the dither for us? (if possible, a step by step configuration)
Thank you!
ES

> So in the end disabling dithering didn't completely fix it?

It DID fix it completely! I haven't been clear enough then. I just found it a little bit strange that it was necessary to turn off the dithering. I wasn't expecting the dithering to have this effect.

When I've tuned the initial sample and phase offset, it can run for a long time without any need to retune it. It works very well!

Thanks for your help!

Cheers

[rtlsdrblog]

If you don't want dithering you can just use the KerberosSDR drivers now https://github.com/rtlsdrblog/rtl-sdr-kerberos

No point turning it off unless you're doing coherent stuff, or have some other advanced experiment.