[time-nuts] Fine delay generator

pablo alvarez pabloalvarezsanchez at gmail.com
Wed Nov 12 16:25:30 UTC 2008


Dear nuts,

I am designing a card that should be able to delay a trigger from 25ns
up to several seconds in 10ps steps. The card will use an external
10MHz as frequency reference.

I have thought of two architectures. One is a counter clocked by a
keyed oscillator followed by a fine delay and the other is based on a
Time interval meter (TIM), a  counter clocked by a TCXO followed by a
fine delay. Let me just explain a bit more about both options.

 1. Keyed oscillator Counter + fine delay. The keyed oscillator starts
oscillating phase allinged to the input trigger and frequency locked
to the external 10MHz.   This is a beautiful scheme, but sofar I have
not found any comercial keyed oscillator or startable oscillator. Do
you know of a design that could do the job? Keeping an stable
oscillator phase and the frequency locked to an external reference at
the same time is not an easy job.  An example of module that does it
is the V850 by highland:
http://www.highlandtechnology.com/DSS/V850DS.html

The fine delay can be implemented with a digital delay line such as
ON's MC100EP196B or Micrel's SY89297U.
http://www.onsemi.com/pub_link/Collateral/MC100EP196B-D.PDF
http://www.micrel.com/page.do?page=/product-info/products/sy89297u.jsp

A bad point is that the time interval between trigger's cannot be
smaller than the generated delay.
A very nice feature is that its monotonicity is garanteed by design.

 2. Time interval meter + TCXO Counter + fine delay
In this scheme I measure the trigger phase respect to the internal
TCXO and calculate the corresponding fine delay that I have to add to
the counter ouput. I expect the time interval measurement latency to
be of the order of 200ns-300ns, so for small delays it would be
necessary to use a separeted fine delay and multiplex the outputs.
Locking the TCXO to the external 10MHz should not be a problem.

For the TIM I could use a ACAM's TDC-GPX which offers 10ps resolution.
http://www.acam-usa.com/Content/English/gpx/gpx_1.html

As an analogue option I was thinking of latching the input trigger
with a flip-flop, low pass filter it and sample it with a high speed
ADC such as the  AD9626. The TIM could be callibrated at the startup,
but I do not have a feeling of how stable it can be.

http://www.analog.com/en/analog-to-digital-converters/ad-converters/AD9626/products/product.html

Generating the small fine delays from 25ns up to 300ns is perhaps the
most diffucult one IMHO. I am not very sure if the classical scheme of
integrator followed by a comparator can generate delays of up to 300ns
with low jitter.  On the other hand chaining 24 MC100EP196B to
generate a 300ns delay seems a bit scary too...

Thanks in advance for your comments

Cheers

Pablo



More information about the time-nuts mailing list