[USRP-users] LO leakage on receiver when tuning the transmitter

Stefano Speretta s.speretta at isispace.nl
Thu Dec 11 10:15:20 EST 2014


Based on what Ian said, I had a look at the code used to tune the AD9361 
and probably found the cause of my problem.

in host/lib/usrp/common/ad9361_driver/ad9361_device.cpp, the function 
ad9361_device_t::tune is used to tune the RF front-end
of both the transmitter and the receiver chains (which are selected with 
one of the parameters) but between lines 1816 and 1821
I can see that the following lines are executed when tuning either the 
receiver or the transmitter chain:

/* Update the gain settings. */
_reprogram_gains();

/* Run the calibration algorithms. */
_calibrate_tx_quadrature();
_calibrate_rx_quadrature();

_reprogram_gains() sets again the gain of both the chains (RX1, RX2, 
TX1, TX2) and also both calibration routines are executed every time a 
tuning is performed on any of the 4 chains.

I think that this could be the cause of my problem and I think that 
somebody may also experience a glitch in transmission when re-tuning a 
receiver for the same reason I am seeing it on the receiver while tuning 
the transmitter.

Regards,
Stefano


Il 04/12/2014 17:58, Ian Buckley ha scritto:
> Stefano,
> A very interesting discovery. Given that you have narrowed this down to incidences where the CORDIC is retuned rather than the RF LO, that narrows the possibilities tremendously.
>
> Am I correct in stating that when you say "or exactly at the LO frequency (seeing it as an RF signal)" that you have assumed the glitch is present at the LO RF rather than actually directly observing it before downconversion?
>
> I can't but help wonder if what you are seeing is the radio being forced to run an RX DC-offset recalibration…at least thats the first thing I can think of that would causes a transient DC spike in the RX spectrum given that (at least in theory) we are not altering any of the analog RF signal path to retune. There's no reason I can think of no reason that we should need to re-run the IQ imbalance functions in the analog radio for a DSP only retune event, but it could be that UHD is doing that. Of course we might legitimately need to do that when you retune the RF LO,but then you don;t see this effect which is again puzzling. I suspect someone in R&D is going to need to look at what explicit operations are precipitated by the DSP retune event.
>
> -Ian
>
>
>
> On Dec 4, 2014, at 8:23 AM, Stefano Speretta via USRP-users <usrp-users at lists.ettus.com> wrote:
>
>> What surprises me a bit is that the glitch is exactly in DC (if we look at the baseband signal) or exactly at the LO frequency (seeing it as an RF signal) no matter which frequency you tune to on the tx side. I would expect spurs of the transmitter frequency to appear in case of a leakage from the transmitter to the receiver (and maybe more than one) or, at least, to move when the transmitter frequency changes but this never happens.
>>
>> I tried changing the size of the frequency jumps and I noticed that, when the jump is more than 1 MHz, the RF front-end is retuned and in that case I do not see the glitch. If the jump is smaller (10-100 kHz) the tuning is done using the internal DSP and in that case I see the glitch.
>>
>> Regards,
>> Stefano
>>
>> Il 04/12/2014 16:26, Marcus D. Leech via USRP-users ha scritto:
>>> On 12/04/2014 04:14 AM, Stefano Speretta via USRP-users wrote:
>>>>
>>>> and I clearly see  glitch appear every time I tune the transmitter. The amplitude of the glitch, according to the FFT, is between -55 and -65. The same file was run with a USRP B100 and a USRP1 using a WBX daughter-board but the glitch never appears. In the patch I also added a message handler to stop displaying messages because sometimes using a USRP B100 there were some messages which were disrupting the FFT plot.
>>>>
>>>> Best regards,
>>>> Stefano
>>>>
>>> What you're seeing is the VFO on the TX-side as it sweeps into "lock" when you change frequencies.  That action will create spectral components that
>>>   will briefly appear (sometimes, not always) in your RX passband.  There's no way, given that both LOs are inside the same TINY chip that you can get
>>>   perfect isolation between TX and RX.  We live in a world where receivers are very, very, sensitive--like at or below 3dB, so even with pains taken
>>>   to try to isolate things (as much as can be done when you have an LO in the same chip as the RX gain and mixer chain), you'll still see these sorts
>>>   of things.
>>>
>>> The only thing I can suggest is to logically mute your RX chain when you're tuning.  And also, how do you cope with "glitches" that originate from
>>>   the outside world?  Because there will always be some--the outside world isn't generally kind in this regard, it's a useful exercise to consider how
>>>   to deal with them.
>>>
>>>
>>>
>>>
>>
>> -- 
>> Stefano Speretta
>> ISIS - Innovative Solutions In Space BV
>> Molengraaffsingel 12-14
>> 2629 JD Delft
>> The Netherlands
>>
>> Phone:   +31(0)15 256 9018
>> Fax:     +31(0)15 257 3969
>> E-mail:  s.speretta at isispace.nl
>> Web:     www.isispace.nl
>>
>>
>> _______________________________________________
>> USRP-users mailing list
>> USRP-users at lists.ettus.com
>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com




More information about the USRP-users mailing list