Hello,

I am currently looking for a way to make a zero crossing detector (i.e. a detector that generates a TTL signal each time an input sine or square signal crosses the zero axis). The signals that I want to process have a very large bandwidth (from 0 to 100 MHz), so I need the generated TTL signals to correspond with a good precision to the real zero crossing points. I found a website showing ZC detectors based on XOR gates or comparators, but I was told that for high frequencies (above several MHz) some of the components are too slow and do not give reliable results.
Does someone have an idea on the way to make it?

Another challenge is that I would like also to be able to count (or to make a signal that is proportional to) the elapsed time between two successive zero crosses. This gives the time duration (i.e. a frequency measurement)for each frequency cycle. I do not know if there exists a circuit that is able to do that for such high frequencies.



Thanks a lot for your help.

 

Get the data sheet for this comparator - http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1154,C1004,C1012,P1225. It should be fast enough.

Rather than try to get the interval between crossings, why not count pulses over time and get the frequency that way?

 

The only practical use for period style measurements I know of is slow waveforms, something in the realm of 1Hz or slower. That way you can get a frequency measurement in one pulse. Fast frequencies, where you can get 6 digits with 1 second samples, don't really need period measurements.

Given you can calculate accurately period or frequency if you have the other I guess I don't see the point. Most high end counters will measure a pulse width, which is pretty much the same thing.

Most folks already know what I've said, just asking the quesion, why this way?

 

If you are hoping to reach 100Mhz in TTL you need to use the 74XXS series, none of the others are fast enough including 74XXLS.

Also be prepared to supply plenty of juice, speeders are thirsty.

As an alternative you could bone up on your FM radio theory. There were a few pulse counting discriminator designs in the past.

 

I suggest you use ECL or PECL logic. With 100MHz signals, your pulse width even with 50% duty cycle is 5ns. A meaningful zero crossing pulse will be around 2ns wide. I believe this is beyond the capability of most (all?) TTL logic-level families, as well as TTL-output comparators.
Here is a way to do it with ECL.

MAX9691 datasheet.
MC100EL07 datasheet.

 

Get the data sheet for this comparator - http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1154,C1004,C1012,P1225. It should be fast enough.

Rather than try to get the interval between crossings, why not count pulses over time and get the frequency that way?

Hey, how would can you design a circuit for the count pulses over time scenario ?

 

RonH beat me to it. In the 70's I made an FM limiter out of ECL running on +5 V rather than -5.2 V (PECL hadn't been invented yet). The input was color video modulated on a 7 MHz to 11 MHz FM carrier, recovered from videotape and DC biased to the middle of the ECL transition region. Worked like a charm. The same thing should work today with newer parts.

ak

 

RonH beat me to it.

Yup. By nine years!

 

Yup. By nine years!

Haven't done that for a while.
RTPD - Read The Posting Date!

ak

 

A long time ago,needing a very fast zero crossing detector I used a 74XXS inverter (similar to sstudiot's mention in post #4) biased to slice a high speed video signal.

Later I found that ASIC designs can do 1 ns analog comparators easily. Along the line of analog comparators consider giving the LM10 a look.