3
$\begingroup$

I am developing a continuous wave Doppler radar to detect human movement at a speed of 0.2-6m/s. For a radiation frequency of 24.15 GHz, the Doppler shift during longitudinal motion will be equal to ~32.2-966 Hz. In transverse motion, the Doppler shift will depend on the angle at which I cross the detection zone. I created a Doppler shift table for the velocities and intercept angles I needed. From this I concluded that the signal I needed would be in the range of 10-990Hz.

enter image description here

I use the InnoSent IVS-162 microwave module. I apply a constant voltage to the varactor tuning voltage input to achieve a central radiation frequency of 24.15 GHz. Next, from the signal I output, I give the signal to the amplifier and analog filters. Filters: High pass filter 4th order ~5Hz; LPF 4 order ~1500Hz. Next, I give the amplified and filtered signal to the ADC of the microcontroller. Sampling frequency 8 kHz. Next comes digital filter 8 of the order of 1100Hz. Next, I decimate the signal to a frequency of 2 kHz. Next come 5 2nd order bandpass filters for frequencies 10-300; 100-500; 300-700; 500-900; 700-990 Hz respectively. Next, I detect the signal and build the signal envelope using low-pass filter 2 of the order of 1 Hz for each frequency range (10-300; 100-500; 300-700; 500-900; 700-990).

After conducting the experiments, I saw that the lateral movements in a normal step, as expected, are in the range of up to ~100Hz and I see a signal in the frequency band 10-300Hz. With longitudinal movements at a normal pace, I see a signal already in the range of 10-300 and 100-500. But at the same time, its amplitude is approximately two times less than during transverse movement. I don't understand why this happens. After all, with longitudinal movement, the person’s area is larger.

Now I have a some questions:

  1. Why do I receive a signal amplitude that is approximately half as large when moving longitudinally as when moving transversely at the same distance?
  2. I divided the signal frequency band into 5 smaller frequency bands in order to get less noise in each frequency band and a correspondingly higher signal to noise ratio. Are there any other ways to increase the signal-to-noise ratio?
  3. At close distances from the microwave module (1-2m), with any type of movement, the amplitude of the signal is simply huge and, accordingly, any movement, for example, of a bird or an insect, can cause a false alarm. Is there any way to prevent this effect?

P.S.

  • I understand that for this microwave module you can use FMCW modulation, which will solve some of my problems, but I want to deal with CW modulation specifically.
  • For processing, I use a very simple STM32G030 microcontroller and integer mathematics, since this microcontroller does not have an FPU, so any complex processing algorithms will not work for me.
$\endgroup$

1 Answer 1

2
$\begingroup$

It's a little confusing what you mean by "longitudinal" and "transversely" in this context. You stated that

In transverse motion, the Doppler shift will depend on the angle at which I cross the detection zone.

This is true in general. Instead let's take two nominal geometries:

  • The target is traveling straight towards or away relative to the antenna face.
  • The target is traveling perpendicularly relative to the face of the antenna (left-to-right or vice versa).

In general, the Doppler of a target at some boresight angle $\theta$ changes in the perceived Doppler by a factor of $cos(\theta)$. In the first case, this angle is zero so the Doppler is due to the speed of the target alone. In the second case, $\theta$ is changing and so the Doppler you measure is due both the speed of the target and its angle from boresight at the time(s) of measurement.

Without asking for more information on your processing, I will try to answer your questions:

  1. Why do I receive a signal amplitude that is approximately half as large when moving longitudinally as when moving transversely at the same distance?

If the target is moving perpendicularly (case two above) and you take a measurement off-boresight, your power returned will be less than if the target was in the direction of maximum gain of the antenna, which is usually straight-ahead relative to the antenna face. If you took a measurement when the target is at the point of maximum gain, it could be moving in any direction and your return power should be about the same. Again, this assumes everything else being equal.

  1. I divided the signal frequency band into 5 smaller frequency bands in order to get less noise in each frequency band and a correspondingly higher signal to noise ratio. Are there any other ways to increase the signal-to-noise ratio?

Yes. You can use more pulses on the target and coherently sum them. Your SNR gain will be approximately $N$, where $N$ is the number of pulses you send out. Since your case is CW you don't have pulses per se, but you can still increase SNR by processing a longer observation.

  1. At close distances from the microwave module (1-2m), with any type of movement, the amplitude of the signal is simply huge and, accordingly, any movement, for example, of a bird or an insect, can cause a false alarm. Is there any way to prevent this effect?

A classic way to fix this is to use something called sensitivity-time-control (STC). By using STC, you can artificially attenuate your signals that are close in range. Nominally you would use an attenuation curve that's proportional to $R^{4}$ so that close in returns are more heavily attenuated vs farther out, in accordance to the radar range equation (assuming two-way propagation). This is primarily done so you do not saturate your receiver by close-in returns. In addition to this, you can implement a constant-false-alarm (CFAR) detector to have better control over what returns are considered detections.

You also stated that

I understand that for this microwave module you can use FMCW modulation, which will solve some of my problems, but I want to deal with CW modulation specifically.

I'm not sure how FMCW will help you if I'm understanding your issues. CW and FMCW both usually transmit and receive at the same time to perform their function and they both have to deal with mitigating self-saturation and isolation of receivers. By modulating a CW tone to be FMCW, you now have access to be able to measure range but won't really help in that regard.

If you have more information on what you're doing, please update the question.

$\endgroup$
2
  • $\begingroup$ You say "you can still increase SNR by processing a longer observation". What does it mean? Can you tell us more about this? Now I am digitizing the signal with a frequency of 8 kHz and then working with digital filters. How do I do this “longer observation”? You also talk about “sensitivity-time-control (STC)” and “constant-false-alarm (CFAR)”. Can you tell us more about these methods? How exactly do they work and how can they be implemented? $\endgroup$
    – red15530
    Commented Feb 24 at 13:14
  • $\begingroup$ @red15530 These need to be separate questions. All of the things mentioned are significant topics on their own. Additional questions also give you the chance to get multiple answers. By the way, most of not all of these topics are probably covered between here and the EE Stack Exchange. $\endgroup$
    – Envidia
    Commented Feb 26 at 17:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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