time domain reflectometer

[danishkaleem]

sir i have design a time domain reflectometer .i m getting reflections on oscilloscope.time difference between transmitted and reflected pulses is 4 microsecond.how can i measure this time difference using PIC16F877A MCU.i need ur help.
regard:
student

 

[ted]

I suppose you want to detect also partial reflections (i.e. not a full short or open)? If so, what you essentially need, is to nearly to construct a digital oscilloscope - kind of, at least. And the pulse driver, as well (which is relatively simple, anyway, in comparison)

I can't give here a cookbook solution. However, I would approach the task by developing he following circuits:
1) Pulse generator. That makes the signal-to-be reflected. The signal has to be driven to the cable/transmission line. This is the easy part!
2) Pulse receiver. I would propose a fast op-amp-based front end circuit "listening" to the line for a returning signal.
3) Fast comparator with programmable reference, giving essentially a one-bit A/D conversion from output of the receiver amplifier. The programmable reference would define analogue magnitude which you want to detect (PIC, or whatever processor you end up with, sets it).
4) A clock generator, with high frequency, and thus a small clock period (t)A clocked delay line = a shift register, length , onto which you fed the output of comparator. By adjusting the clock period (t), you would control the time-scale resolution. By giving the shift register exactly clock pulses (the length of shift register) you would capture samples of amplitudes, when the processor reads the shift register contents. Of course, your time unit would be (t) nanoseconds between each sample.

And then you need some very simple control logic and non-trivial (however, totally writable) software to run the system: Setting different reference threshold values and clock periods, reading the one-bit digitized received pulse, and then figuring out the reflection amplitude and its timing.

As you can see, I would consider the task to be so-called "tall order", if done properly. Thus, maybe even if the question fits to forum "Electronic Elementary Questions" , the answer might not fully comply to the "Elementary" part....

Good luck,
ted

 

[HMS1021]

A TDR (time domain reflectometer) generally includes measuring the relative amplitude and timing of the features in the returned signal with sufficient resolution to see and measure discontinuities in connectors which suggests a resolution of 1mm or so. This requires a pulse width on the order of 3ps.

While you have TDR in the title the request is to just measure the cable length which is different than a reflectometer since amplitude is not mentioned. Using just the PIC you can produce pulse widths of about 200ns and determine their transit time also with a resolution of 200ns by using the comparator. This gives a total resolution of 400ns or about 60m in air. With multiple measurements you can apply centroiding and possibly reduce this to 40m or so. In RG-58 coax the resolution would be ~26m with centroiding due to the lower propagation velocity in the dielectric. By resolution I mean you could measure the length of a 500m coax cable with a certainty of 26m. This assumes there is a fairly severe mismatch at the end of the cable that will have a sufficiently large refection to make it back and still be over the detection threshold.

The basis of the 200ns is the 20MHz clock (50ns) and 4 clocks per instruction. It is also the compare time for the comparator. There may be a method to use the 12.5ns capture timing of the PIC's comparator and edge triggered interrupts to significantly shorten the timing and improve the resolution but I can not see it.

The internal ADC was not considered due to its much lower sample rate.

By using a external short pulse source and a high sample rate ADC followed by memory the PIC could be use to control the measurement and at a slower rate retrieve the data from memory for processing.


Just for reference, what is the sample rate of the scope you used to measure the 2 microsecond long cable? Given this sample rate, have you calculated the uncertainty of the measurement?