bladeRF x40 - 1PPS tamer mode stability

[OskarM]

Hi,
I didn't really needed it but wanted to check if I can calibrate frequency on bladeRF (and on the rtl-sdr via bladerf) using external 1PPS signal. I connected a decent GPS module (Telit 873) and connected 1PPS output to bladeRF x40.

The accuracy & stability of frequency without 1PPS is very good:
https://drive.google.com/open?id=18lskb ... LxOoiv8O59

With 1PPS signals I get things like:
https://drive.google.com/open?id=17GNJp ... IZiMqdd05r
https://drive.google.com/open?id=19RPPC ... Ns_slAYeST

Now - 10k of 1Ghz is 10ppm and equals 10us of the 1PPS signal which seems to be well below expected 1PPS quality. I do have u-blox on my desk, couldn't yet get it connected to compare results, but 10us is really poor - u-blox ie. states <20ns, telit has not strict information but claims to be in 100ns range.

Anyone tested this tamer mode feeding directly from 1PPS (not GPSDO)?
thanks

[bglod]

The source of the reference (GPS, test equipment, etc.) should not matter as long as it is:

• 1.8 V CMOS, square wave
• libbladeRF is told the correct thing (1PPS or 10 MHz)

How long does that go on for? When the VCTCXO tamer is first enabled, there is a "coarse tune" phase of the VCTCXO tamer. It sets the VCTCXO to its minimum trim voltage, logs the error; sets it to the maximum voltage and again logs the error. From these data, it computes a line: DAC counts on the y-axis, error on the x-axis. The y-intercept is the DAC count that theoretically results in 0 error. Once here, the tamer enters "fine tune" where it monitors the error over the previous 1 second, 10 seconds, and 100 seconds and makes adjustments accordingly. It's a naive approach, but it was stable enough to keep the bladeRF tamed in a windowsill for a few days, while an untamed bladeRF wandered around significantly more and failed some GSM tests.

I suggest leaving the board powered on and operating for several minutes before enabling the tamer. This will allow components to come up to temperature and may result in a better tune line.

If this behavior happens for more than a couple seconds at the start of taming, then something's wrong with your setup or the reference. The only 1PPS "reference" I have is the pulse generator output on an oscilloscope. It's not very good, but I set it up just now and it appears to work as intended for me. I had the bladeRF transmit a 1 GHz CW into a spectrum analyzer. Enabling the VCTCXO tamer for a 1PPS reference shows the course tune occur for 2-3 seconds, then it reaches steady-state quite a bit away from the factory tune value. Once there, the transmitted tone stayed at that position (it did not jump around). There is a wobble of like +/- 10 Hz, but that same wobble was also evident without the VCTCXO tamer module enabled.

[OskarM]

Hi,
thanks for explanation - it didn't improve with time unfortunately. It seemed like either it cannot stabilize (algorithm is oscillating) or the PPS output is very unstable (unlikely). But there is a new hope with a new day - I couldn't reproduce this error today. The trimming stopped at a single trimdac value and is very stable. I'll come back with the issue if it happens again.

thanks, b.r.

[Daccus]

Did you manage to get it completely stable, OskarM? I'd love to know how exactly.

[OskarM]

I didn't check it exactly. If you ask about stability in terms of "allan deviation" or such then I have no tools to measure that. In terms of "bad/good" - very stable. The signal tends to randomly float a bit (some tens of Hz) but this is small analog drift and when you blow on the bladeRF it goes crazy

In terms of precision - I have only RTL-SDR v3 at hand so no help here - I would have to have a 10PPB class device to meaure that I guess. If you're fine with <1kHz @ 1Ghz (1PPM) then that should be already satisfied without 1PPS tamer. The only gain here is long term temperature & anti-aging stability.

Some older picture - bladeRF 1GHz into RTL-SDRv3.
https://drive.google.com/open?id=18lskb ... LxOoiv8O59

b.r.