On 2/27/08, Richard Clarke <address@hidden> wrote:
Before I go ahead and try to implement an auto calibration routine for
the Rx frequency offset of the receive subsystem on the RFX400/USRP, I
was wondering if anyone else had already implemented something similar,
or had any pointers for me.
While what you describe is well-intentioned, it doesn't solve the real issue.
Basically what I want to do is create a
calibration routine that will determine the frequency offset between the
RFX400/USRP receive subsystem (nominally already set to equal the
frequency from a transmit source (e.g sig gen) and the transmit source,
and then nudge the tune frequency to null that offset. I find that my
RFX400 receiver after time to stabilise is around 3.2KHz in error, which
is well within it's spec I know, however I want to automatically
determine this offset and null it out.
Do you want to null it out once, based on a your signal generator
input, and store that value somewhere? How would you use this value
in the future?
I guess the actual fine frequency
adjustment will actually be occurring in the DDC, would that be correct?
If that is how you were to implement it. If all you are doing is
applying a fixed correction based on this calibration technique you've
described, then it's easier to just tune the daughterboard to the
corrected frequency. You're correct that you will see the "beat
frequency" in the I or Q channel equal to the frequency offset between
the transmitter and receiver.
However, I don't think this addresses the bigger issue, which is that
receiver frequency and time base offset will always exist, and must be
corrected in real-time if you are using a modulation technique that is
sensitive to these offsets. Even temperature differences in the
frequency reference in the USRP can cause many kilohertz drift,
especially if you are operating at high carrier frequencies, like
2-3GHz.
There are many techniques for receiver frequency and phase
synchronization. They generally involve a block to measure the
frequency/phase error, a loop filter to turn this error into an
appropriate control value, and an NCO to apply the actual fine
frequency/phase correction to the incoming signal. These would be
designed to work with the properties of the actual modulation you'd be
receiving.