0
$\begingroup$

Spectra when everything is stable

Spectra when Tx channel changes

Another example of spectra when TX channel changes

The image shows the frequency spectra of a signal recorded using a SDR. The SDR is of DirectDownSampling type and has some compensation in the signal chain to compensate for the DC spike which would appear at 370MHz the center frequency. But unfortunately this has caused a dip at the centre frequency. Now since the Tx channel is varying the signal power fluctuates which is shown in image 2 and image 3. The null at 370MHz is at about -22dB but this null should be at about -17dB with the other nulls. I want to be able to dynamically compensate for the 370MHz null during varying channel conditions bringing it up to -17db and also compensate for the slope on the flat tops adjacent to the 370MHz null. I have been trying of fit a line to the top of the flat tops with little success.

$\endgroup$
  • $\begingroup$ Thank you very much :) $\endgroup$ – Marcus Müller Jun 24 at 11:21
1
$\begingroup$

has some compensation in the signal chain to compensate for the DC spike which would appear at 370MHz the center frequency.

Yes, a typical problem of IQ demodulators is the LO leakage/DC spike. You really can't do much about that. You don't mention the SDR you're using, but many can be used with offset tuning, completely moving the LO out of your band of interest and thus completely solving that problem

But unfortunately this has caused a dip at the centre frequency.

Why is that a problem? Looking at your spectra, I'd clearly say you've got four subcarriers working there, and the centre frequency happens to end in the gap between the center two. There's really not much you'd need to fix.

Now since the Tx channel is varying the signal power fluctuates which is shown in image 2 and image 3. The null at 370MHz is at about -22dB but this null should be at about -17dB with the other nulls.

Why? It's a null. There's no data here. Simply leave it alone!

I want to be able to dynamically compensate for the 370MHz null during varying channel conditions bringing it up to -17db

That's a bad idea. You'll break more than you fix when trying to equalize based on a cancelled component. You get overproportional noise amplification that way!

Instead, you'd want to base your equalization on something that comes through well.

This is exactly the discussion you probably read when you first were introduced to frequency-domain equalization: Zero-Forcing looks easy, but will amplify the noise on channel-suppressed frequencies. You'd be doing even worse: you'd be selectively taking the noise on a suppressed frequency and use that to equalize. That means you'll multiply a lot of noise onto your whole spectrum. Bad, bad idea.

also compensate for the slope on the flat tops adjacent to the 370MHz null. I have been trying of fit a line to the top of the flat tops with little success.

Why should there even be a flat top here, or a straight line? This is clearly a multipath channel, and a curve fit isn't what you'd use to equalize that.

From the look of your signal, I'd say this is probably one larger or four low-subcarrier count OFDM transmissions; if that is the case:

The whole point of OFDM equalization is done by the OFDM receiver on each subcarrier individually, and is not something that you should even do before.

All in all, the frequency selectivity of this channel looks pretty benign.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.