DIY 800..1600 MHZ FMCW GPR – in progress...

Work in progress FMCW ground penetrating radar. Sharing measurements, build logs, and asking for feedback on sensitivity and antenna design

Description

I have a background in electronics and have long been fascinated by geophysics, especially ground-penetrating radar. That interest may have been sparked by watching too many Time Team episodes (John Gater).

I don’t have a practical application for it. The fun for me is in the electronics and building the device. I’d be happy if my design could detect stones or bricks buried up to 75 cm deep in relatively dry, sandy loam soil. Then I can test it in my garden, that has some buried remains in it, like a filled in pool.

After reading everything I could find on the internet I decided to try a FMCW ground radar, sweeping from 800MHz to 1600MHz, as that seemed the least challenging.

Details

STARTING

Since RF design can be tricky, I began by building some small RF test circuits (filters, microstrips, amplifiers, etc.) to make sure reality matches theory.

I etch the PCBs myself using 1.6 mm FR4 as the substrate, with a full ground plane on the bottom layer. For simulation I use QucsStudio, and I have access to a small but decent 4GHz VNA and a test receiver.

THE PROJECT

Over the past year, I built something that “works,” but not quite good enough. It gives repeatable results when used on soil, but when I dig, there’s nothing there. However, when I place the device on the concrete floor of the first story of my house, it clearly detects people in the room below.

So I think it’s time to step back and get some advice.

WHAT COULD BE WRONG?

The issue could be in several areas:

The RF front end
The antennas (reflections, isolation, ground effects, or distance from the ground—too close or too far)
The operating frequency (I chose 800-1600MHz to keep the device portable and because of the AD8347’s minimum frequency, but it is probably too high)
The data analysis stage

APPROACH TO IMPROVE / LEARN

My first step is to confirm whether the electronics work as expected. I plan to replace the antennas with a coax cable and attenuators, then share the schematics along with measurement results to get feedback. From there, I’ll decide on the next steps.

Files

Schematic_FMCW.pdf

EasyEDA schematic

Adobe Portable Document Format - 124.36 kB - 08/20/2025 at 20:00

Preview

Project Logs

Collapse

Question : does the frontend perform good enough ?
Fieldman • 08/22/2025 at 12:46 • 0 comments
SUMMARY OF THE PROBLEM (For more details about the hardware see my first log)

I build a (DIY) prototype of a 800-1600MHz FMCW GPR and I am not that happy with the results yet.

There could be many causes but I would like to start by getting your opinion on the performance of the front-end. If that is not ok, there is no point at continuing with more complicated stuff like antenna's.

THE MEASUREMENT

I attached a good coax cable + a series of attenuators between antenna input and output.

This way I should get an idea of the sensitivity of the device and if it is working.

Coax properties (measured with calibrated VNA) : 800MHz : -0.5 dB , 1600Mhz : -1.1 dB, 20ns delay (TDR measurement) , length 5 meter.

Attenuators I used 10dB and 20dB inline attenuators ( see image). I increased the attenuation from 60dB to 110dB and the measured voltages behave exactly as expected ( 20dB attenuator -> voltage 10x lower). So confirms there are no measurement errors or stray signals leaking in.

A scan measurements with 80dB (4x20dB) attenuation between TX and RX input

A scan ( with hardware high pass / differentiator enabled).

The frequency is about what I would expect :
with a 20ns delay and a 800MHz/1ms sweep I expect a beat frequency of 20ns * 800MHz / 1ms = 16KHz. And there are about 16 cycles. Note : the FFT uses zero padding that is why the FFT peaks at 4x higher value.
A scan measurements with 110dB (4x20dB+2x10dB) attenuation between TX and RX input
Amplifier gain is set higher to maximize signal.

What I would like to learn from experienced builders :
- Is this front-end performance OK ? Do you need extra measurements to say ?
- Would you expect a higher sensitivity ? Or a cleaner sine wave ?
- Any other test you could suggest to learn more ?
system components
Fieldman • 08/20/2025 at 19:38 • 0 comments
System overview

Block diagram RF FRONT-END

Complete schematic

General info
- VCO is swept linearly in 1ms from 800MHz to 1600MHz. VCO non linearites are compensated.
- Microcontroller averages 16 sweeps, does analyses (FFT) and sends results 30 times/s to smartphone app through a Bluetooth serial link.
- Large VCO harmonics are reduced with a tunable low pass filter that is swept synchronously with the VCO. Resulting in clean constant power output, with second harmonic at least 35dB lower then main frequency.
- Special care is taken to isolate the input/output circuit and reduce VCO pulling.
- The AD8347 is a quadrature demodulator, but currently I am only using the Q output
- The Q output is connected to 2 LF amplifiers, both are measured by the microcontroller 12bit ADC inputs.
  - the first LF amplifiers only amplifies
  - the second LF amplifier has a high-pass filter. The idea is that the further away a reflection is the weaker it will be and the higher the frequency is. So it makes sense to amplify higher frequencies more. In software I can choose which amplifier to use.
- The AD8347's amplification is controlled by the microcontroller ( using a I2C DAC)
Smartphone app

With the Android app I can :
- set the sensitivity (amplifier gain)
- view live A scan's ( average of 16 sweeps).
- view B scan ( time - depth )
- control many settings
I will add more about the antennas & performance later, but first I would like to find out if the electronics perform good enough or not. That is my first question in a next post.