How to characterize a transceiver

[wtr]

It's been a while since I've been in the weeds with DSP. I've mostly been involved in more abstract levels of data processing on frames (networking), or doing crypto etc. in FPGA land. Therefore I'm rusty and need guidance and a refresher on DSP techniques.

I've been assigned a task to characterize a transceiver.

Here is my thought process. Please help me iron out the correct steps.

1. I have a signal generator blasting a signal in at fc=2440MHz.
2. I've setup constant bandwidth and gain etc on the transceiver - receiver.
3. I then program the transceiver center frequency to sweep across the so called passband and roll off (I'm moving Rx window, but could move Tx)
4. For each step of the sweep I generate a binary data file.
5. The binary data file can be unpacked to generate iq0 and iq1 or samples=np.stack((iq0, iq1), 1) * 2**-15
6. These samples can then be viewed in spectrum or fft or iq form.

At point 5 my thought process breaks down. I can see in point 6 that the spectrum intensity of the line (representing transmitter at 2440MHz) decreases as I move away from the center frequency.

My initial thought was to somehow compress the entire 10ms window to 1 max value for frequency at offset.

How do I do this? What steps do I take to turn samples into max hold style graph?

Regards,
Wes

 

[kaz1]

you seem to have a simple single tone test for your Rx. you can do fft on Rx data and see the line. If you want you can average multiple fft frames within each sweep (such as pwelch). But I believe one fft or so is enough as sanity check. If amplitude drops then that is what you want see.

 

[wtr]

I get the amplitude dropping by 3dBm between sweeps. Is this to do with fft bins?

 

[kaz1]

I get the amplitude dropping by 3dBm between sweeps. Is this to do with fft bins?

if you are playing same tone that appears in same bin then 3dB drop is not expected.
if your receiver is sweeping across then it will hit the filtering transition edge leading to drop. That is what want to see and measure I guess.

 

[wtr]

Y - Axis is DB
X - Axis is Freq offset from center of transceiver.

Can anyone explain why a simple fft would lead to such a discrepancy?