On 11/18/2020 05:32 PM, Dustin Widmann via USRP-users wrote:
Hi usrp-users,
I've noticed something strange with an experiment I've been doing. The
experiment is set up like a reflectometer. I'm using an X310 with a
UBX-160 for transmit and both channels of a TwinRX for receive. In
short, I put up a tone with the transmitter by applying a DC offset,
tune to a nearby frequency with the receiver (in this case, tuned 50KHz
away) and measure the tone. The problem is that the tone isn't being
observed at the expected frequency on the receiver.
The data with which I noticed it is in the table below, where I was
only using a 2^17 point FFT (763Hz resolution)
freq target bin/freq actual bin/freq diff bin/freq dsp
freq discrepancy
60MHz 65.536 (50e3) 64 (48,828) 1.536 (763) 1160
397
61MHz 65.536 (50e3) 66 (50,354) -0.464 (-354) -61
415
62MHz 65.536 (50e3) 67 (51,117) -1.464 (-1,117) -1282
165
63MHz 65.536 (50e3) 69 (52,643) -3.464 (-2,643) -2503
140
64MHz 65.536 (50e3) 70 (53,405) -4.464 (-3,406) -3724
318
65MHz 65.536 (50e3) 72 (54,932) -6.464 (-4,932) -4945
13
66MHz 65.536 (50e3) 58 (44,250) 7.536 (5,750) 6044
294
The discrepancy looks like it could be explained by the radio's claimed
"dsp tune" not actually occuring. I ran the test over again with larger
FFTs (2^20 points) to see if the noted difference frequency would be
closer tot he dsp frequency...
freq target bin/freq actual bin/freq diff bin/freq dsp
freq discrepancy
60MHz 524.288 (50e3) 512 (48,828) 12.288 (1,172) 1160
12
61MHz 524.288 (50e3) 525 (50,068) -0.712 (-68) -61
-7
62MHz 524.288 (50e3) 538 (51,308) -13.712 (-1,308) -1282
-26
63MHz 524.288 (50e3) 551 (52,547) -26.712 (-2,547) -2503
-44
64MHz 524.288 (50e3) 563 (53,692) -38.712 (-3,692) -3724
32
65MHz 524.288 (50e3) 576 (54,932) -51.712 (-4,932) -4945
13
66MHz 524.288 (50e3) 461 (43,964) 63.288 (6,036) 6044
-8
... and it was indeed much closer. This suggests that the digital part
of the tune isn't occuring. Does anyone have any idea what might cause
this issue / is it a known problem / any alternate theories on what
might be happening?
Relevant infos:
USRP X310
* UBX-160 (1 TX channel)
* TwinRX (2 RX channel, phase sync, lo sharing)
UHD 4.0
Dustin
Are you using an external clock, or the on-board clock?
If I'm interpreting your table correctly, the residual error is under
1PPM in each case. The absolute error of the system clock will be
distributed over
all the components involved, including the DAC, the ADCs, and the LO
synthesizers in all the modules involved. I'm not sure that they'll all
perfectly
cancel each other out, given that the TX and RX synthesizers are
quite different.
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
