High-Precision Timer

If such high precision timer is not easily available in the market, is there anyway I can solve the problem through other means?

Look at it on a fast oscilloscope and measure it. I'd have thought the hard part was catching the pulse, once you've caught it the measurement is relatively easy.
(Bit like mice).
You don't say if this is a continuous process or a single event you are measuring, and why, all of which are relevant.
Del

http://www.directindustry.com/prod/spectracom/time-interval-and-frequency-counter-34040-198441.html

Try that one!

It's possible, barely, to do this digitally (using painful expensive unobtainium parts), but most designers solve this problem with a combination of digital and analog techniques, using switched integrators to make interpolation measurements between clocks. We discuss the details on pages 1022 to 1024 of our book, The Art of Electronics, 2nd edition (image captured from Google). The design task isn't trivial, but it quickly allows much better resolution than you're seeking.

If you ask the question on the usenet group, sci.electronics.design, you may get some good answers. The easiest way to access this group is via Google groups, http://groups.google.com/group/sci.electronics.design/topics You can also Google old s.e.d. threads discussing this and closely-related topics.

Scott Nelson et al

Over 50 years ago I and others were searching for a means to measure the VSWR, reflection coefiicient, Return Loss etc. in waveguides, essentially we were seeking the VSWR of individual pairs of Waveguide flanges.

We built the fastest pulse generator known to man at the time which was a coaxial transmision line fired by a Mercury switch, the rise time was a few nano seconds and the pulse width could be varied by changing the length of the open circuited coaxial cable.

We then modulated a carrier wave from a klystron with this and produced a very fast short pulse of RF energy which was then transmitted along the wavegiude to be measured. Apropriate directional couplers, detectors and oscilloscopes (as per Dell above) could then measure the VSWR of each individual flange and we could then tell if changes had to be made in the flange configuration. The limitation of the technique was the dispersion introduced by the waveguide which effectively smeared the pulses, the longer the waveguide the more the pulse was smeared.

We then extended this technique to measure a large range of conditions; the oobjective was to produce better and better transmission lines for the carriage of highly compley FM FDM sysytems - this was before the invention of waveguide which did not require flanges every few feet a it was produced in long sections with a flange at either end.

We were fortunate in that we had flanges at measured intervals so that we had the ability to say there should be pulse here, one here and so on. Sure enough there were. We then just had to calibrate the effect. Not difficult.

I had previusly spent many months measuring by standard means, the VSWR of individual flanges in Waveguide and seeing the effect of mis-alignment horizontally, vertically, diagonally; putting space betwen flanges, using different materials, shims of different materials and anything else that we could think of - we were seeking higher precision than had been obtained before!

But then Elliptical waveguide was invented and perfected to the point where the problem that we were trying to solve went away!

Sleepy

Hi Win,

Thanks a lot for your inputs! I will find your book to read. But before that, let me explain the problem in more details. The timing measurement unit (TMU) I am designing has to measure width of arbitrary pulse shape, transition time of a digital signal (not pulse). The resolution is 500ps. To do that, I plan to input the signal to two analog comparators with pre-set start and stop threholds which trigger and stop the counter to measure time. I cannot afford to design a chip for that purpose and all the devices to be used must be cheaply available in the market. Will your scheme work in this case?

Hi Win,

will the HP 5370B Time Interval Counter work in my case? according to the spec, it can achieve a resolution of 20ps.

Thanks!

It can, and it may work for you. You can get 5370B counters on eBay, from time to time; I have one, paid in the $250 to 500 range, IIRC.

The Stanford Research SR620 is a modern replacement, much more expensive, not quite as good in some respects, actually.

I was going to ask how many pulses, what repetition rate, how long, etc. For example, for measuring the width of individual short pulses to say 0.1%, pulses no longer than 500ns (0.5ns accuracy), you could use a single fast comparator and a fast integrator, along with a few calibrations. It's amazing how well today's fast op-amps make textbook integrators.

Oh, My! The 5370B counter is so big! I only have a small place for the time measurement unit on my board. I was expecting something like several chips or capacitors, etc. So this probably would not work for me.

The spec does not say about the repetition rate of the pulse. But usually we would apply a single step (ideally) input signal to our target circuit, and measure the rise/fall time of the circuit response. So I guess the signal may not repeat at all.

Hi Will,

I think your comparator + integrator scheme sounds very interesting. Would you please elaborate it?

Thanks.

Scott

OK, tell me the maximum length of your pulses. Can you have a range switch?

Hi Will,

The max length of the pulse is 50ns and we can have a range switch. One thing that I am concerned with is the accuracy of analog implementation. The issue is not just about calibration, but also about its sensitivity to temperature etc.

Thanks!

Ah, only 50ns max, that's great! Then a 0.5ns measurement error would be only be a 1% error! That's very helpful, it just means you need really fast analog electronics, basically using the tried and true techniques that worked well in the ns territory, but with faster ps-scale parts.

It's late, and I'm off to sleep, but a quick note about one way to do the job.

Here at right we have a current source with a fast diode current-steering switch that shunts the current onto the capacitor during the ON pulse time. After the pulse you'll measure the new ramped-up capacitor voltage, and at some later time you can reset the capacitor to be ready for the next measurement.

You discharge the capacitor with a special low-capacitance switch, but it will have some charge injection, so after the reset is finished there may be some residual charge on the capacitor, as shown in the drawing. That's why you also take a measurement before the pulse, and take the difference to get your answer.

It's easy to add circuitry for say 10ns and 50ns calibration pulses, etc., but we won't get into any of that.

Thanks, Will. I will try it!

If there is a PC involved, buy a scope card like the one here:

http://www.acquitek.com/news/data-acquisition-systems.html?idpage=583&art=985

Pleanty of resolution, and only one small board. Many others are availble if you search the net long enough.

In a related area, can someone suggest the easiest way to determine whether optical pulses (either short laser pulses or longer ones modulated at gigahertz frequencies) arrive at two or more optical detectors simultaneously, to a precision of a fraction of a nanosecond? (Conceptually, like having photocells in series) When the experiment was first proposed 40 years ago, the hardware did not exist (or at least wasn't easily found). Apparently we have come a long way since then. The determination of simultaneity should not require human observation, as in looking at oscilloscope traces.

If it was for a quick graduate-student experiment, I'm sure they'd grab a common 500MHz digital scope, or a slightly faster one, and get started (I've lost several scopes from my lab that way, borrowed and then integrated into experiments). They'd trigger 32,000-point, etc., two-channel traces, and write a LabView program to grab and process (cross-correlate?) the waveforms, and create nice output displays.

The whole idea was to be automatic, with no human involved. You have a horizontally scanning pulsed laser and a vertical array of detectors (2 or more in the image plane) which view the world. If two adjacent detectors detect the pulse at the same time, they were reflected at the same time, implying the illuminated object (looking forward) is vertical. Several successive simultaneities as the laser scans hoizontally measure the width of the object. (It's wider than a tree and smaller than a church) If the pulses are reflected from the ground or from foliage, the pulses will be smeared and most likely not simultaneous.) If you are an autonomous vehicle, you can avoid hitting buildings or, if you are a missile, you can seek out buildings and ships.

Fair enough, but that's always the prob with these Qs, we often don't know the application which makes sensible replied difficult.
I'd assumed you were looking at a one off measurement problem not a realtime problem.
Del

I might add that the same hardware could be used to generate a 3-D map of the scene, with contour lines. Many things which would not be noticed because of a lack of contrast, fuzzy edges, shadows, camoflage nets, etc. will stand out in 3-D. For example, flat surfaces or cylinders will stand out with straight parallel contours, while natural objects, like trees, will have irregular or indistinct contours.

Like I said, you could get started by experimenting with a fast digital scope. But for the real thing you'd want a PC-board-mounted ADC. There are some nice fast dual-channel folding-amplifier flash converters that should be well suited. For example, National Semi's ADC08D1520, a dual 8-bit 1.5GS/s converter.

Add a pair of fast fully-differential amplifiers, like Analog Devices ADL5561, set to G=5, etc., and you're good to go. Well, you'll also need an FPGA dealing with 32-bit lvds data words at 750M words/sec. Sheesh!

These wicked-fast ADCs are expensive too, $728 each (min order 60 pieces), and the eval board costs a cool $5,600. Hey, at least it includes an FPGA. Hmm, a used 500MHz 3GS/s digital scope isn't looking so bad.

Thank you guys! I will try the approaches you suggest!