How to measure two signals with a delay of less than 1 ns?

[pichuang]

Hi everyone,

I am wondering if anyone has tried to measure two signals with a delay difference of less than 1ns? I am working on a high speed digital circuit right now and the delay from the rise of CLK to the rise of output is less than 400ps.

Ideally, I would like to see this on an oscilloscope. However as I checked the best oscilloscope I can get has a sampling rate of 20GSa/s, which is 50ps per point. I think in order to construct a meaningful graph, it will require at least couple hundred samples(dots) on the oscilloscope to show the two waveforms?

Has anyone encountered similar problem before? How do you solve/toggle this problem? Any tips and suggestion is very appreciated.

Thanks,

Pierce

 

[FvM]

For a repeated signal, the measurement would be achieved with considerable accuracy by any middle-class DSO by equivalent-time sampling. For single events, special measurement circuits as time-to-voltage can do the job.

 

[pichuang]

Re: How to measure two signals with a delay of less than 1 n

Hi,

Can you please elaborate more on the time to voltage for single event measurement? Are you referring to the time to digital converter?

Pierce

 

[FvM]

Re: How to measure two signals with a delay of less than 1 n

It may be a building block (or the heart) of time-to-digital. Circuits can be found in literature, particular about nuclear physics instrumentation, I don't have a reference at hand. I've seen the circuit also in industrial instrumentation as US doppler flow measurement.

Some variants exist. Generally, you have a capacitor, and a switched current source, e. g. a differential pair. The current source is switched to load the capacitor for the intervall between the two pulse edges. The voltage at the capacitor is sampled then and processed, e. g. to a digital value.

The most critical design part typically isn't the time-to-voltage conversion but to generate a precise trigger from real-world signals. Special circuist as constant fraction triggers haven been designed for this purpose.

 

[ASIC]

I understand the theory, but how do you switch the current source that fast?

 

[FvM]

Re: How to measure two signals with a delay of less than 1 n

It works with unsaturated transistors as basically any high speed digital and analog circuit. Please consider, that the timing resolution isn't determined by the switching speed but the switching time uncertainty. Thus a differential pair with e. g. 1 ns rise time in output current can achieve a time resolution down to some ps in time to amplitude conversion.