ISM S-Band LFMCW Radar Prototype

An ISM (2.54GHz) Radar for automotive, industrial experimentation.

Similar projects worth following
An ISM (2.54GHz) Radar for automotive and industrial purposes and experimentation. The prototype will serve as a testbed for Doppler, Gated CW, and LFMCW applications.

Status: On Hold

This project started as a development system for roadway-entry safety. I have a blind turn out of my neighborhood with cars racing at high speeds from the traffic light one block away. I'd like to eventually setup a sensor to detect the presence and speed of vehicles approaching our driveway. A secondary application would be a miniaturized version on the flanks of my vehicle to cover my blindspots on the freeway.

The project started with the layout of two RF boards; one transmitter and one receiver with LNA and downconverter. A pair of patch antennas were also spun for experimentation. Unfortunately I'm not much of antenna designer yet so the antennas are embarrassingly bad and had to be replaced with off-the-shelf replacements.

Early Doppler experiments show promising results; the system is viable for traffic monitoring.


1. Hardware Bringup: Complete

2. Doppler Testing: Complete

3. Doppler Street Testing: Complete

4. LFMCW Range Testing: Complete

5. Antenna Redesign and Testing: Complete

6. Subsystem Consolidation and Redesign: Pending

  • May 2015 Update

    majolsurf05/24/2015 at 09:35 0 comments

    Been working quietly on this project off and on lately, time is at a premium.

    Status update:

    1. Due to high TX/RX spillover I replaced the Yagi's with high-gain cantennas. Still suffering crosstalk but less drastic. Not a fan of cans as they are not properly matched to free-space.

    2. Housed RF electronics in RF enclosures to reduce crosstalk; minor improvement. Crosstalk is in antenna subsystem and garbage cabling.

    3. Racked system in bistatic configuration on tripod mount.

    4. Moved from Baudline-based analysis and use the Velleman PCSGU250 as a modulator and for receive baseband waveform analysis. Able to get 300MHz modulation bandwidth with this test setup for about 1 meter of range resolution.

    5. Using Matlab for post-processing algorithms. Scripts are adapted for GNU Octave, however I continue to run them in Matlab as the animation time is faster.

    6. Experiments are limited to high-clutter environments (alley behind my house) and without baseband filter. The latter is only needed for sensitivity and anti-aliasing which I am not concerned with currently. The former is situational.

    A. I was able to observe myself walking in the field of view of the radar. A chain-link fence is visible in the RTI. Using MTI I am able to resolve myself walking in the fence resolution-cell.

    B. I have observed basic Doppler measurements of nearby traffic.

    7. Conclusions so far:

    A. MTI processing improves detection, however I need to move the system to a cold sky environment.

    B. Audio-baseband processing is great for 1st-order proof-of-concept, however if measuring multiple high-velocity targets, fast-chirping is necessary to mitigate the doppler/range ambiguity function of FMCW systems. This requires wider baseband bandwidths (2MHz) thus outside the audio spectrum.

    8. Next steps:

    A. Move system to open field for further land-based observations.

    B. Consolidate RX/TX boards into one design and include baseband filtration and an embedded modulator/demodulator. I would also like to include a pulse-doppler option on this board.

    C. Use doppler mode to detect long-distance targets of opportunity.

    D. Post videos of previous observations.

  • ARGtek Booster Are a Bust

    majolsurf07/29/2014 at 05:13 0 comments

    I typically don’t like publicly decrying hobbyist products, however I have recently encountered one that I feel we should all form a circle, point, and collectively boo.


    I purchased a pair of 2.4GHz T/R modules, touted as WiFi boosters, from 3D Robotics. While the receiver specs are less than stellar, the transmit side offers 30dBm of output power. No external T/R switching required, the unit detects TX signalling. The units were to be used in the ISM radar prototype.

    After unboxing, I tested the amps with a signal generator and spectrum analyzer. I attempted to normalize the output from 30dBm to 0 using a dial attenuator.

    The amplifier with attenuator.


    The output of the test setup. The spectrum analyzer displays -14.22dbm whereas it should display 0 (attenuator set to 30 minus a few dB to remove setup error).


    Additional sweeps tests were performed which are not worth mentioning. I also attempted to trigger the T/R switch by using a bias tee to inject CMOS/TTL DC signalling to trigger the switch to no avail.

    Receive testing was worse, the output waveform exhibited loss. Moreover the CW tone had significant (high) AM sidebands. I suspected the T/R switch was causing this however switching typically manifests self as the sinx/x (sinc) function.

    If I had only purchased one, I would not consider this a problem. However both units exhibit the behavior. Moreover the vendor, 3D Robotics, has not responded to me in a week since I submitted my test data. Thus ends another personal lesson in not buying overseas under-priced products.

  • Moved to Dual Yagi's and a DS2000 Mount

    majolsurf03/12/2014 at 00:47 0 comments
  • Early Doppler Experiments

    majolsurf02/20/2014 at 19:49 0 comments

    Video Demonstration

    In summary the following video demonstrates the use of a continuous-wave transmission, the reception of the received Doppler-shifted echo, and its mixing with a copy of the transmitted waveform, known as the correlation L.O.

    This type of Doppler radar can detect the presence and velocity of a body in motion. Distance, however, is not measurable with traditional pulsed radar techniques due to the short distances involved. Target motion can be measured at baseband, close to D.C., the frequency of which indicative of the target velocity.

    Demonstrated in the video is basic motion detection as well as an example of Doppler signature, in this case a house fan both stationary and oscillating.

    The ISM Radar Testbed hardware used is a transmitter consisting of a VCO, power divider (for the correlation L.O.), and an amplifier capable of +17dBm. The amplifier was turned off for the first experiment thus the output power was -5dBm. The receiver consists of an LNA and mixer. Baseband filtration is intended to be provided externally, however was not included in the initial bringup. I also laid out two simple patch antennas which offer about 3dB of gain each in a bistatic configuration. I plan a series of antenna respins to experiment with beam forming and gain improvement.


    Early models show that with a vehicle whose RCS is 32dBsm, at 5 meters this system is capable of outputing a receive echo at -25.1dBm. With a decent baseband filter the system noise floor is better than -85dBm thus making the system sensitive enough to detect vehicles 100 meters away, provided the following conditions are met:

    1. Multipath is managed through narrower beamwidths and sidelobe shrouding
    2. Backend processing SNR requirements are:
    a. If performed in the time domain, 25dB SNR is sufficient
    b. If an FFT or Goertzel algorithm is used, detection may go beyond the noisefloor however processing time must be minimized
    3. All spurs and interferers lie 25dB below the carrier if time-domain processing is used

    Also explored in the video are the various sources of noise in the system, both fixed frequency spurs and “wideband” noise floor bumps. The main culprit in this system is a cheap lab supply, which utilizes full-wave rectification and poor filtration which splatters the baseband spectrum with 120, 240, 360Hz, etc, spurs. A prominant 60Hz spur also exsists in the laptop sound card.

    Turning off the transmitter makes all spurs but 60Hz disappear. This makes me believe that the transmit supply is bad and is generating sidebands which make their way to the receiver via the L.O. Of course direct coupling through the ground plane is possible, however knowing the source is key to suppression.

    Proper filtration techniques shall be employed, as well as alternative power sources, such as batteries. The system is intended to be deployed on a vehicle, which one would think is safe due to battery operation, however alternators are a famous source of noise.

    Multipath cannot be discounted as a source of interference in this demonstration, as well as the fact that the transmit VCO is not disciplined with a phase lock loop. Also and most unfortunate, the wifi on the laptop was left on during this experiment. The drift in the very likely to be operating near the wifi band of operation which has two effects:

    1. The first is visible in the noise floor of baudline, where through reciprocal mixing the wifi occupies enough noise bandwidth to raise the entire FFT noise floor.

    2. In the Doppler signature section of the video, we see the laptop wifi drop out briefly as I approach the radar. I am possibly inducing another multipath mode which may very well affect the wifi receiver.

    Although wifi radios employ OFDM modulation to maintain operations in environments of high traffic, noise, and multipath, the effects of a 20dBm CW tone generated 18 inches from the wifi receive LNA cannot be discounted. No matter...

    Read more »

  • Early Bringup Activities

    majolsurf02/20/2014 at 19:44 0 comments

    Not going into too much detail tonight, just finished bringing up the transmitter subsystem prototype. +14dBm output power which is about 3dB off of the mark. However the measurement was taken with a detection diode, not very accurate.


    In the link budget calculation in the block diagram above, +14dBm is more than enough to sense vehicles over 100 feet away. Frankly the detection distance if far greater, however even 100 feet may be pushing the limit of practicality on the highway. Without high antenna directivity, multipath will be more of an issue.

    I do intend to investigate whether FMCW is a more robust detection method in multipath environments, specifically when an FFT is applied to the back end. For example, in an environment where a known path interferer will always be present, ie the road, then an interference zone (IZ) can be designated.

    More details to come, the project is very much a work in progress. FPGA tutorials are still in the works, stay tuned for the second NCO lab!

View all 5 project logs

Enjoy this project?



tomcircuit wrote 03/11/2014 at 09:54 point
The back end was a PC and MATLAB. I built the coffecan radar unit (google charvat tin can radar). I very much followed the recipe of Dr Charvat, and eventually took the week-long radar short course at MIT last summer that featured building a small radar. Lots of fun!

If you keep your chirp rates slow enough, you can use an audio-type CODEC as the digitizer. An FPGA would be a fine back end, but I have to admit it is very convenient to use MATLAB for all the processing (in batch mode). Consider prototyping with MATLAB (or similar) and then moving to FPGA once you know what to expect. The CFAR, in particular, can need a lot of tweaking.

  Are you sure? yes | no

OneShot Willie wrote 03/10/2014 at 21:37 point
Hmmm... might try one of these - they're $25 on Ebay...

  Are you sure? yes | no

majolsurf wrote 03/10/2014 at 22:10 point
Ah that makes things much easier, but less fun. :)

  Are you sure? yes | no

tomcircuit wrote 03/09/2014 at 20:42 point
Nice project! You should have lots of fun with the LFMCW, I sure did. My own homebrew 2.4GHz radar was able to discern large-ish RCS objects (e.g. trucks) at 100m. It really suffered at short distances ( < 5m) due to leakage between the antennas. So I'd recommend antennas with some directivity to avoid the problems I had. Looking forward to seeing more updates on this!

  Are you sure? yes | no

majolsurf wrote 03/10/2014 at 22:14 point
What did you use for your backend processor on the LFMCW? I'm planning on an FFT in an FPGA, but Baudline seems like it should be sufficient but is not producing results.

It may be because I'm having antenna isolation issues. My first patch antennas you see above are completely wrong. To move development forward I bought a pair of yagi's. I have yet to perform isolation measurements yet.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates