I have a simple radar module (k-band) that is capable of fm modulation and it provides both I and Q as baseband outputs. I like to extend the measurement range of this module by just signal processing options. (I.e. Without changing any hardware)
-
Would modulation help in terms of range? I can generate any signal such as triangle, sinus etc at the FM input of the module. (I know some range info could be obtained but I am after SNR, range knowledge is not material to my objective)
-
Currently I only use I to do my calculations. Could I also use Q together with I to reduce noise or improve SNR ? (I couldn't find any literature to that effect but I think there could be some gain in SNR if I use this additional signal)
Best Answer
There are very strong limits on what you can do to improve SNR. After all, your received signal consists of a low level reflected signal, and a noise signal in the same frequency range, that cannot be filtered out.
A couple of things to do
a) Don't throw away any signal that you don't need to, use both I and Q
b) Reduce the detection bandwidth as much as possible, to reduce the amount of noise around the signal, without reducing the signal. How much you can do this depends on the phase noise of your transmit and receive LOs.
As the I and Q outputs are produced by the module mixing the received signal with the transmit LO, you will be able to use synchronous averaging. As the phase noise of the oscillator (from which we could deduce coherence time) is not specified, we cannot figure out up to what range this will work well.
You will be able to make some test measurements. Your range agnostic signal processing consists entirely of DC averaging the I and Q signals, with no FM applied. Do this with no reflector in range. The result will be your DC offset. Now repeat with a reflector present, and see the result.
Unfortunately in using DC averaging, you will be reducing noise at the cost of DC drift. A little thought experiment around the setup of getting the DC offset above will illustrate this. What about transmit signal that 'leaks' straight into the receiver? Won't that appear in I and Q? Yes, anything in the environment that puts transmit signal in the receiver will affect your result. Your ability to detect remote (weak return) targets confidently rests entirely with the stability of these unwanted returns. If they are stable, you can subtract them, if not, you can't tell them from returns.
This is where the FM modulation comes in, to add range information to your returns. With range information (you said you didn't want it in the OP, but it turns out to be necessary if you want sensitivity to weak signals), you can tell the difference between a weak reflection 10m away, and a change of leakage signal at the module/antennae.
With FM, the receiver is mixing the returned reflections not with the LO that was transmitted, but with an LO whose frequency has changed since it was transmitted. If you gate your reception properly, and use triangle FM (triangle in frequency deviation, not triangle in voltage at the tune port, they are only the same for linear tuning), then a substantial part of your received signal will be at an IF frequency of +/- the distance.
Unfortunately, to use synchronous averaging, you need the offset phase to be stable at the IF. Although in theory, a stable FM will give you stable PM, in practice any slight change in FM gain and FM drive level will make the phase deviation wander around, the unwanted deviation getting worse as the time/distance increases (PM is the integral of FM after all), precisely the conditions where you want to use this technique to dig your signal out of noise.
The technique is therefore to gate your received signal, to use it at times when the FM offset is consistent. Then take an FFT, to encode reflection distance against phase. Then do it again and again, summing the FFTs, so the consistent target phase adds as voltage, and the inconsistent noise adds as power, gaining you 3dB SNR every time you double the number of readings in the sum. Then ignore reflections with a phase shift indicating they are from Tx to Rx leakage at the module, and accept ones from further out.
Obviously the FFT will be unable to distinguish between signals that have had a phase shift of \$\theta\$, and a shift of 2\$\pi\$+\$\theta\$, so you will need to choose your FM deviation and modulation rate to suit your maximum range, so that significant reflections never exhibit more than 360 degrees reception phase shift.
Obviously, a moving target will be smeared out over several range phases, just as VCO phase noise will smear a stationary target out. The latter can be used as a limit to determine how fine an FFT it's worth using.