Home

Void your waranty at your own risks: I am not responsible of hardware updates that might deteriorate the GPR performances.

The ProEx GPR control unit exhibits some sampling rate drift (slides presented at IWAGPR2017) as the control unit temperature is stabilizing after switching on the unit: variations of up to 3% over a duration of a couple of hours have been observed, meaning 5 ns delay on a 170 ns echo, or one full phase rotation for a 200 MHz antenna. We have concluded that the cause of this timing drift is the fast ramp generation in the stroboscopic voltage to time converter (ref 1) used to trigger the measurements (T&H trigger of the ADC). We hence consider remplacing the analog integration of a constant voltage to create a linear voltage ramp with a digital implementation.

The operational amplifier in charge of the integration is an AD843 which gets awfully hot as the electronic board warms up after being powered. The reset signal is available on the top right pin of the MAX334 analog switch (green test point on the top figure). Hence, our objective is to get rid of the AD843 analog ramp generation, generate a digital ramp whose output feeds the signal at the output of the op-amp (pin 6 of the AD843, green dot on the top figure). The ramps that feed the comparator are indicated by two green test point dots on the top figure on pins 1 and 8 of the AD790. The clock signal for this digital electronic is provided by the 16 MHz master clock found on the image below.

The main challenge with the digital implementation is to keep the fast rise time of the ramp when the emission pulse is triggered, and yet keep the slower ramp (+/- 15 V in about 7 us) noise free, especially from the clock. 30 V in 7 us still requires a pretty fast DAC: 256 steps in 7 us is 25 ns/step or a 40 MHz DAC.

At the moment, best results were achieved with an external arbitrary waveform generator -- Tektronix AFG3102: despite a documented jitter of 500 ps on the external trigger as found in the datasheet, generating the fast ramp with this instruments yields sub-50 ps drift and jitter on the resulting timebase (300 ns separated echoes, 0.5 hour measurement). While this solution is acceptable for a lab based demonstration, an embedded solution is needed.

Reference:
ref 1: B. Johansson, Ground penetrating radar array and timing circuit, US 6496137 B1 (2002), and most interestingly this figure.