IQ imbalance caused by echotek config

march 31,2006

plots:
30mar06 I,Q voltages and spectra before echotek config fixed (.ps) (pdf):
31mar06 I,Q voltages and spectra before echotek config fixed (.ps) (pdf):

The setup:

Upshifted and downshifted plasma line data was taken with the echotek card on 30mar06. The data came from the carriage house. The setup was:

The ipp was set to 10 milliseconds.
The upshifted plasma line was centered at 435.8 while the downshifted plasma line band was centered at 424.2.
The rf of 420 to 440 Mhz was mixed to 30 Mhz IF, and then split. The downshifted line was now centered at 24.2 Mhz while the upshifted band was at 35.8.
The tx samples were mixed to (24.2 Mhz) and added to the downshifted IF.
The two data streams were then input to two echotek samplers each running at 80 Mhz. The echotek card mixed one stream from 35.8 Mhz down to dc while the second stream (with the tx samples) was mixed from 24.2 Mhz down to dc.
We wanted a 5 Mhz band around each line. There are 4 channels on the echotek that can handle 2.5 Mhz bw each. To get a 5 Mhz bw, two of the channels are combined by shifting one of them by 90 deg.

When the tx samples were mixed to 24.2 Mhz, there was a frequency offset in the mixing of about .5 hz. This caused the tx samples to drift slowly in time.

The data:

On 30mar06 we looked at the data and saw that the total power of the tx samples were changing with time. This turned out to be a problem in the configuration of the echotek cards causing the IQ samples to not be 90 degrees apart. Since the tx Samples were offset from dc by about .5 hz, the total power vI^2 + vQ^2 did not remain constant. If the tx samples had been mixed exactly to DC we would not have noticed this.

The plots show the problem (on 30mar06) and after the fix (31mar06).

30mar06 I,Q voltages and spectra before echotek config fixed (.ps) (pdf):
31mar06 I,Q voltages and spectra before echotek config fixed (.ps) (pdf):

The top plots shows the I,Q voltages tx sample 86 over 500 ipps (5 seconds). A sine wave was fit to the I (red) and Q(blue) voltages separately and then over plotted (dashed blank lines) on the data. The fits shows the phase offset between the I and the Q was:

30mar06 phaseI: 209.86 phaseQ:228.75 delta: 18.89 degrees
31mar06 phaseI: 23.80 phaseQ:293.80 delta: 90.00 degrees

The bottom plots are the spectra for the 500 ipps using the I and Q values:

30mar06 there is a peak at .763 hz and an image at -.763 hz. This is being caused by the I,Q voltages not having a 90 degree phase offset
31mar06 there is now only 1 peak at .763 hz that corresponds to the frequency offset in the mixing of the tx samples to 24.2 Mhz.

On both days there are birdies 40 db down around +/-40 hz. On 30mar06 they are double on 31mar06 they are single (since the IQ problem was fixed.). These birdies are caused by 60 Hz aliasing back into the band. They are offset from +/-40 hz by .763 Hz. The offset (which matches the mixer offset) tells us that the 60 Hz is coming from the Tx sample signal (rather than being added at baseband from the digitizer circuitry).

processing: x101/060330/iq_30mar06.pro, iq_31mar06.pro

home_~phil