I am building an FMCW radar that has a bandwidth of 1 GHz. The IF frequency is up to 1 MHz. Would the ADC requirements simply be that it has to b 2 times the IF frequency of 1 MHz? If that's the case can I simply put an ADC of 5 MHz and not worry about anything else? What are some things a should look at for DSP requirements? I'm not so clear on how range, velocity, and resolution affect the requirements. Thanks for any help. I expect editing this question as I am sure I left some parts unclear. Thanks

  • \$\begingroup\$ If it has a BW of 1 GHz (unlikely of course) then it can't have an IF of anything less than 2 GHz off the top of my head. \$\endgroup\$ – Andy aka Aug 15 '16 at 22:13
  • \$\begingroup\$ The ramp is from 24 to 25 GHz and the IF is 1 MHz. \$\endgroup\$ – Daniel Aug 15 '16 at 22:16
  • 1
    \$\begingroup\$ The carrier sweep range is not the same as the bandwidth. The bandwidth is a function of the rate of change of the carrier frequency and the rates at which targets will be moving. \$\endgroup\$ – Dave Tweed Aug 15 '16 at 22:18
  • \$\begingroup\$ If signal of interest is right at the Nyquist aliasing limit 1/2 sampling rate, that probably isn't really good enough. With only 2 samples per cycle, the amplitude error could be as high as 100% (e.g. if sample is at just the wrong moment, at the zero crossings.) Usually around 10-20 samples per cycle there is enough information to recinstruct the input signal well enough. YMMV. You may also look into undersampling techniques, since you mention 1Mhz is the IF and might not be the actual signal bandwidth. \$\endgroup\$ – MarkU Aug 15 '16 at 22:27
  • 1
    \$\begingroup\$ It is only necessary to sample at twice the information bandwidth to meet the Nyquist criterion- although as stated by others this is a mathematical limit. In practice you should sample a little higher. If you wish to extract IQ rather than magnitude only, you will have to double the sample rate again. \$\endgroup\$ – N.G. near Aug 16 '16 at 1:37

FMCW radar is a bit tricky, and your "IF frequency is up to 1 MHz" is a bit vague.

In an FMCW system, a target returns a signal that is delayed based on its range. The frequency difference between the transmitted signal and the received signal is a function of that delay and how fast the transmit signal frequency is changing. This frequency difference swings positive and negative, but has a zero mean if the target is stationary. If the target is moving toward or away from the radar, then the mean IF frequency shifts negative or positive, respectively.

The key point is that in order to extract the maximum information from the signal (range rate as well as range), you need to be able to distinguish positive and negative frequencies in the IF, which means that you really need to cover a range of -1 MHz to +1 MHz, or a total bandwidth of 2 MHz.

Nyquist requires that your sampling rate be at least 2× your bandwidth, so a 5 Msps ADC would probably be considered "barely adequate". 8-10 Msps would probably make your life a lot easier.

| improve this answer | |
  • 1
    \$\begingroup\$ Except that to cover -1 to 1 MHz and disambiguate positive from negative components, you need an IQ system, which means two ADCs. But they can be half as fast. So actually, two 2.5 MSPS ADCs would be "barely adequate". The single 5 MSPS ADC only works if the system has an IF centered at 2.5 MHz (or 7.5 MHz etc) rather than 0. \$\endgroup\$ – Chris Stratton Aug 15 '16 at 23:00
  • \$\begingroup\$ @chris not quite sure the requirements really are IQ ADC here. You generate the beat frequency by multiplying TX with RX with a single e.g. Gilbert cell. Of course you'll always get a positive difference frequency that way - if you don't do IQ, it's your job to interpret the results accordingly \$\endgroup\$ – Marcus Müller Aug 15 '16 at 23:17
  • \$\begingroup\$ @ChrisStratton: Yes, I'm guilty of a bit of hand-waving there. But the point is, one way or another, you need an absolute minimum of a total of 4 million data points per second. \$\endgroup\$ – Dave Tweed Aug 15 '16 at 23:39

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.