FMCW sweep duration vs FFT spectrum/distance resolution
Let's assume that we have simple FMCW distance measurement system with linear frequency sweep from F1 to F2. Sweep duration is T.
ADC frequency is enough to digitize 128 points for each sweep duration T. Then we do some windowing and perform an FFT. For example 1024pt FFT (yes, i know about 128 points and interpolation, but it is from real application).
Here is my question:
If ADC frequency is in hardware and cannot be changed, but VCO allows to make wider frequency sweeps... Can we improve distance resolution of this system in this way?:
1) double sweep depth (before: F1 to F2, after: from F1 to F3, where F3=F2+F2-F1)
2) double sweep duration time t=2*T
3) now we get 256 points for each sweep from ADC (ADC frequency is fixed in hardware)
Then we put this 256 points to 1024p FFT, and something tells me, that resolution must be improved. In 128pt we had more zeroes data = more interpolation and blurry spectrum.
As you can see, frequency sweep is the same, it is just wider two times. Also we must get the same beat frequencies (dF/dT is the same, because 2*dF/2*dT=dF/dT).
And what if we make modulation even wider, so ADC gives 1024 points? I think we get best spectrum.
Thanks for your attention!
I altered sweep table for VCO and resolution improved as expected. Sweep band now is near limits of VCO. ADC and DAC sampling rate is at their limits too.
How resolution can be improved further? I thought about making more rapid sweep (double speed of df/dt), it gives less maximum range and better resolution, but sampling time also reduces to 128 points, so it is more blurry. I can't just take 8 sweeps data and put it in one 1024 point FFT. Maybe perform some spectrum accumulation, for example take 8 spectrums and make an rectangular or gaussian average of 8 spectrums?
p.s. I read some papers on FMCW, and their spectrum looks very sharp and clean. But sweep band and ADC is much better than mine. As i can't improve them easily, i want to find some way of getting more data.
so anybody knows how it done right way? I see there is a problem with practical applications. When i analyze industry fmcw measurement system i see that data is very poor in resolution. Only few points of data, then some windowing and fft. I think it is not best possible solution.