DIY UWB Pulse GPR

[Muerzi]

Hi Treasurehunters.

I almost read every thread here regarding GPR.
Some nice papers and helpful schematics were posted.

My conclusions:

-TX is not the hardest part, there are very many schematics and circuits around the internet to generate 100ps-2ns pulses.
-RX will be the most difficult part as receivewindow is only 30-40ns wide. A minimum 1GSPS osci single shot is required to read the signal.

I found a good document to sample with very high rate with low cost components.

feel free to check this out.
https://www.researchgate.net/profile...ication_detail
http://icawww1.epfl.ch/uwb4sn/slides...n_Lausanne.ppt

maybe there are more documents out there for this project.

maybe we can start a project based on this docs

[Muerzi]

Another great paper from this researchers, but in German

http://www.imst.de/itg9_1/vortraege/.../folien_01.pdf

[Davor]

This construction is perhaps not the most useful one for our purposes, as the bandwidth is very wide. But the working method remains the same.
Analysing echo signal is perhaps the most demanding part here.

[Muerzi]

What about putting this signal to a eg 600mhz trimmed dipole antennan?

[Davor]

That's precisely the idea. In the schematic you most probably noticed a stub as a differentiator. Place an antenna instead of it, and you'll obtain a pulse shaped to - the pulse response of such antenna. No further pulse shaping required.

The requirements for a GPR are slightly contradicting each other. If you want great resolution, you need very sharp pulses, but those are of very wide bandwidth, and it will be attenuated terribly. A dipole will present itself as a sort of gamatone filter with bandwidth equal to the antenna bandwidth. With narrowband antenna you get slow time response, and fuzzy images. Broadband antenna will give you sharp image. It is a matter of compromise. But playing with antenna bandwidth is fun.

[Muerzi]

So an easier project would be to make a gpr like FMCW as it has a better SNR and is waaaaaaay easier to measure

[Davor]

No.
There is a proximity problem. FMCW radars work beautifully at distance, and their simplicity is beautiful. However...
FMCW results in Rx having a sort of doppler beat, but only for targets that are far off. The beat frequency is proportional to distance.
The limitation is the repetition of a chirp, due to the limited bandwidth. It results in discontinuity at the chirp end. So move a target close, and you lose the distance information. There are but a few tricks you may pull, and ultimately not even tricks will make the close by targets any more accurately detected.
The soil is in effect a LPF that limits the meaningful chirp bandwidth to under 2GHz, and that's also a big stretch.

It could be possible to design a CDMA or chaos radar with reasonable increase in complexity. These are based on a correlator which may be shifted in time. These may be realised at a mixer level and delay may be swept in discrete steps. That way you are diminishing the ghosts in an image. However, I'm not sure if that's practical at $500 level.

[Muerzi]

So what do you think would be the best compromiss between good resolution and easy/cheap to build?

[Davor]

The best resolution is obtained using the antenna with least possible gain - a dipole, or even better a shorted dipole.
Pulse excitation directly coupled to the antenna, and separate Rx antenna(s), direct sampling at antenna with varying sample delay.
The most complicated part is echo interpretation software.

[Muerzi]

What about using a Picoscope 2204A in repetitive mode (ETS)? Should sample up to 2 GS/s and costs around 150€
When pulse is transmitted with 100Hz, should be enough to get good results

[Davor]

... provided you can control the timing well, perhaps.

If I'm a software genius, whom I'm not, I'd try my skill with holographic images obtained by a doppler radar.

[Muerzi]

Aso a great article about a diy 1GHz BW sampling head
http://electronicdesign.com/boards/1...asily-modified

[Davor]

That's a good way to put it. Sequential sampling is the easy way to get a sample. Once you have the samples and you plot them in a waterfall fashion, you end up with the familiar cone-like GPR plots.

I remember one amateur FMCW radar implementation a few years ago by prof. Matjaž Vidmar http://lea.hamradio.si/~s53mv/ but please note that its minimum distance is limited - as noted previously.

[Muerzi]

If I'm a software genius, whom I'm not, I'd try my skill with holographic images obtained by a doppler radar.

Any articles about this method?

[Cyruswiseman]

Use bowtie antenna for gpr application. It is possible to design a good one by antenna magus software