MarkT:
One approach is to use very short sample gate time with a high speed sample/hold circuit before the ADC.
You constantly retrigger pulses and offset the sample time w.r.t the pulse time a little different each
time.
This enables a low performance ADC to see into very short timespans - the narrow sampling gate
pulse provides a window, and you slide that window a little for each reading.
Mark,
You hit the Hard Part(TM) head on. They sell ADCs into the $10,000s that are lightning fast. I'm not trying to reinvent the TDR, though. If I had an affordable ADC, one of the work arounds would be necessary, right? Making the pulses is comparative child's play compared to immediately being able to read the reflection's voltage on the line.
So let's say I go the ADC route. Thre are $60 ADCs at retail places that can do 500 MSPS at 8 bits. Let's look at both parameters:
**500 MSPS **
- Yields 2' (approximate) precision. This is because 1ns = 1/1x109. So if I sample at 1GHz (1 GSPS), I would theoretically sample voltage once every nanosecond. So 1/2 the sample rate = 2x the distance, or 2 x 1' = 2'. I guess you could call 2' precision 2' *resolution * in theis case.
- One foot precision kicks *ss! But the jump in price is significant between 500 MSPS and 1000 MSPS (1 GSPS).. Again, I'm trying to make something I (the average tech) could build & use. There's just something about using test equipment you built yourself... Someone should call it the Heathkit Effect...
8 bit
- At first, this sounds really low, doesn't it? But then you imagine a 128x64 LED disoplay, and you realize your vertical range is 8 bit, or 64 LEDs.
- If you have ever sent pixels, or even small text to a 128x64, or even a 64x64 LED, you realize that you could easily see an impedance mismatch on it-at least one significant enough to warrant a signal failure.
- The target, reluctantly for the Geek™ in me, is not high accuracy or even high resolution. I'll simply rescale the output by a multiplier defined by the highest & lowest signal, and plot the 128 for display.
Thoughts? Is this a good enough logic & spec for 'build or buy'?
Since this is just overall project discussion, and I want this to be an open/for everyone design, a quick note on the horizontal display. If you imagine that you are sampling every 2' of line, you can see how 128 horizontal LEDs will yield 256' of display (yes, there is a dead zone). We know that you often have to do a full roll, or an insanely long run, that is more than this, so I'm thinking in my head (a vaccuum) that 2k' is a reasonable stop. But its it? With a horizontal zoom/scroll you could (again assumes 500 MSPS) either zoom into a 256' portion, or zoom out as far as you like. I'd have to add a pot for the adjustment on an initial test, though I hate the idea of knobs sticking out on field equipment (fix all that late, or leave it up to the builder's choice). So the display once zoomed would display the center 'distance' in small font in a corner of the LED and one could scroll a window through the range, reading a single result at leisure.
But back to the distance/max range for a moment. As these are samples taken either manually or at a slow rate, the limit here is voltage. Because I'm not limiting testing to coaxial or UTP/STP, adjusting the voltage of the pulse would be handy. In my work with coax I have seen opens that could be bridged by a high enough voltage spike, and I imaging the same for twisted pair is possible. I've seen damage by birds & arrows, but the hardest to find was sabotage. Don't continue if you aren't interested in real-world stories (this should avoid complaints).
I include the following as an example of the kind of coax damage that may or may not have been a good example of using a variable voltage spike to catch. As you will see if you read on, a high enough voltage spike may have indicated the kind of open, and at which end. I welcome your opinion.
queue gratuitous flashback music & wavy lines
We had a stretch of CATV .750 that would go down when it was cold up here in CT. I mean really cold. When it was warm, daylight, all was well. I know I spolied it by saying in advance it was sabotage, but any [line] techs have any ideas yet?
I didn't warrant a bucket yet, and had just hooks. Every night, I'd go out, and sure enough, the trunk amp on one stretch would work, and the next one downstream had no voltage (pulsed DC, or square wave). I went up & down those 2 poles a lot that winter-on hooks.
So I tore off the heat shrink on both ends, unscrewed the connectors, and sure enough, you could see the teeth marks on the copper clad aluminum (connector was secure). I even checked connectivity between the connector and the trunk amp's motherboard connector inside. All good. In the parlance of the militay: "WTF?"
The only time I could troubleshoot was when it was happenning, and that was late at night, when it was windiest & coldest-on hooks. So I went up & TDR'd (please forgive the expression/verb) from both ends. From the upstream end, the distance was correct, and so was the impedaance (MB removed downstream = open), and the distance was right (best TDR you could get at the time. I know because it felt like 5,000 lbs). From the downstream end, it was open-no return at all. [Any techs have a guess yet?]
I was about to put an adapter and an RG-6 extension for the dead zone on the downstream amp's input, so I took off the spacers (hard to explain here, but they & strap cover the cable) to bend the .750 out of the way to make room, and the coax UNDER the spacer split in 1/2. WFT?!
Someone on a ladder (no way to rach on hooks) had gone up the pole, and about 4' upstream from the trunk amp had cut the .750 clean with a hack saw under the spacer & strap, then covered it up with the spacer, re-wrapping the strap under it. Devious? On a scale of 1-10, an 11.
end of gratuitous 'only tdr story'; queue flashback return music & visuals
If you read this far, you'OK in my book! -pat
Thanks to all for contributing so far.
