Precise Time Interval Measurement - ps level

Some of the best timer / counters on the market will only give a resolution down to 5 picoseconds...

Their accuracy is as good as 50 picoseconds, sp a short answer would be to get the accuracy you want i.e. better than 10 pS and to be able to resolve two edges only 10 pS apart would be extremely difficult, if not impossible with today's instruments.

I still remember an experienced engineer putting things in to perspective by telling me that light travels about 1 foot (300mm) in 1 nanosecond...

So in 1 picosecond light travels only 0.3 mm!! And you want electrons to move faster than this?

No doubt someone will correct me if this is wrong, I haven't checked its correct recently

John.

Dear John

I just was hoping to pull you in by this thread. You are a real man of measurement technology.

Actually I got into this nasty design problem which world has solved long ago.

I am working on delay line type sensor in which if you give impulse in the middle then it travel towards both ends. This wore is only 50cm long and that of copper. Now problem is to find out the time difference in reaching these signals with accuracy of say 50pS and perhaps resolution of 10pS gibing 1bit as 10ps. SRS makes such instrument but I just don't know how.

Pulses are not actually 50pS apart but anything from 0pS to say 10nS. The instrument used to measure these is in three stages. First one is CFD or constant fraction discriminator which removes walk due to peak voltage difference and gives comparator output in 100pS with 10pS jitter, which is the best possible for analyses of time.

Now people use Time to Amplitude Converter, which is like start stop pulse to charge a capacitor to a voltage using constant current source. Once capacitor is charged, you can measure the voltage in much slower time and that is what time interval was.

Now question is, how accurate can be this time to amplitude converter TAC? People say that these can be very accurate and they perhaps use ps switching gate for pumping electrons on perhaps a very small capacitor in a chip.

I am looking for an OEM board for this requirement and do not like to re-invent the wheel.

Any idea how this can be done accurately with not lots of dollars?

Hello Shyam,
I found this having no idea myself ... ;-)

Regards Uwe

A high resolution time measurement system

Yamaguchi, Y. Koyanagi, N. Katano, K.
Yokogawa Electr. Corp., Tokyo;

This paper appears in: Instrumentation and Measurement Technology Conference, 1991. IMTC-91. Conference Record., 8th IEEE
Publication Date: 14-16 May 1991
On page(s): 618-621
Meeting Date: 05/14/1991 - 05/16/1991
Location: Atlanta, GA, USA
ISBN: 0-87942-579-2
References Cited: 0
INSPEC Accession Number: 4188720
Digital Object Identifier: 10.1109/IMTC.1991.161670
Posted online: 2002-08-06 17:47:42.0

Abstract
A high-resolution, less than 1-ps, time measurement system (TMS) has been developed using a time-to-voltage (T/V) conversion interpolation technique. In order to accomplish the high-resolution measurement, the authors developed two specific techniques. One is the generation of an accurate interpolation pulse from the measured signal that is realized by the emitter-coupled logic (ECL)-gate-array. The other is the T/V converter which converts the pulse width to the corresponding voltage to thereby measure the pulse width. The T/V converter consists of current switches, and integrated capacitor and a buffer amplifier. The switched current charges the capacitor during the time that the interpolation pulse is on a high level. In this conversion, high-speed and accurate analog circuit design are being used, and therefore a 1-ps time resolution and a less than 10-ps linearity error are obtained. The system is composed of a wideband input amplifier, a 16-b microprocessor and a highly stabilized oscillator. The total RMS jitter using this system is less than 15-ps, and single-shot resolution is less than 1 ps. TMS has many applications, for example, time interval measurement, jitter analysis, measurement of frequency stability, etc

Dear Uweka,

Good. This is the type of thing I was talking about. In principle it is a very simple technique.

You have two pulses called start and stop pulse. You can also push one pulse away to known time using a delay line. This makes some good difference in minimum time interval. This was the concern of john. You can make zero time as fixed known time interval using this technique.

Now start stop pulse are used to trigger a set reset flip flop to get one pulse width equal to time interval. For large time measurement this can be used to gate say 100GHz clock to get 10ps accurate time measurement as long as clock of 100Ghz is accurate to that level.

Other method is to connet the pulse to an analog switch to leak a constant current on some small value capacitor to sample and hold charge for slow measurement of voltage equal to time interval.

Now is it easier to count 100Ghz clock or make a switch to sample the current on a capacitor is the choice of the designer. I am worried about both technique. I think the current charging method also called TAC method looks simple than handling 100GHz pulses.