[time-nuts] TAC/TDC

[Bruce Griffiths]

Magnus Danielson wrote:
> On 12/15/2010 08:13 AM, Josef Kölbl wrote:
>> Magnus,
>> with regard to your questions:
>> measurement rate = 1 kHz
>> rms jitter <= 10 ps
>> linearity better 3 ps
>> resolution of 1 ps is the minimum differential measurement which can be
>> made (ie granularity).
>
> Thanks, this makes your requirements much clearer.
>
> This does not seem to be within reach of the ACAM devices and the 
> PICTIC II is not in that neightborhood either:
>
> http://www.thegleam.com/ke5fx/pictic.htm
>
> Considering the 25 ns maximum TI difference (if I recall your specs 
> correctly) it would require a 16 bit ADC besides good input.
>
> I doubt that you would find a complete solution lying around, but 
> either you revamp the PICTIC II concept or go for some other variant 
> of time interpolation.
>
> Maybe Bruce may give some suitable advice.
>
> Cheers,
> Magnus
>
The resolution and noise requirements can be met using a TAC.
The Wavecrest 2075, for example, uses a TAC plus a high resolution ADC 
to achieve sub ps resolution coupled with a noise of a few ps rms.

A 16bit ADC will suffice for <1ps resolution over a 25ns range.
The trick in achieving the required TAC performance will require 
adoption of similar techniques to those used in the Wavecrest 2075 as 
detailed in the associated patent.
Most of the details within the patent were well known decades before the 
time the patent was issued.
The Waverest 2075 makes extensive use of ECL together with discrete CML 
switches.

A good PCB layout preferably using a 4 layer board or equivalent will 
probably be required.

A ringing LC style TDC should also easily achieve the desired noise and 
resolution if the results (by an italian group) using a crude switch 
(its relatively easy to devise a switch with much higher performance) to 
excite an LC circuit are anything to go by.
However a pipeline or similar ADC sampling the resultant decaying 
sinusoid at 80MSPS or more are required together with an FPGA (or DSP) 
to process the ADC samples.
The LC circuit should resonate at around 5-10MHz or so. The exact 
frequency isnt critical as long as its relatively stable in the short 
term as the ringing frequency can be one of the parameters fitted to 
sample burst of the exponentially decaying sinewave.

Simulations indicate that repetition rates of 10KHz or more should be 
readily achievable.

Bruce