Square wave from sine -- high stability

[mekpeacenfood]

lpro-101

Hi, we have a very accurate sine wave source that we'd like to use to sync our whole apparatus with. However, not all components take a sine input--some use a trigger port and want a square wave.

What is the best way to convert a very clean 10 MHz sine wave to a square wave? As a complication, it will need to drive 50 ohms ~1V.

It seems like such a common task--surely everything from oscilloscopes to timing standards themselves need this. What is the accepted strategy? Is a schmitt trigger alone enough, or will phase jitter still be more than a couple ns? If so, what can I add to "take it to the next level"?

 

[FvM]

square wave to sine

A fast comparator would be the usual solution.

 

[biff44]

square wave comparator

I wouldn't use a comparator unless timing is not critical! The slightest voltage noise at the comparator threshold will turn into clock phase noise.

I would make sure the sine wave is big (linearly amplify), and then either clip it at ground and +Vcc with very fast schottky diodes, and then buffer with a very fast (1 ghz) CMOS gate.

You might get away with just driving the sine wave, AC coupled, into the appropriate gate--without the diode clipping. sn74lvc1g17dbvr looks like a good one to use.

 

[FvM]

clipped sine to square wave

The suggestion should be verified by comparing real device data. I assume, that the said very accurate sine wave has a sufficient amplitude anyway. In this case I insist on my opinion, that a fast comparator (as ADCMP551 or similar) promises the best performance in most cases. Fast CMOS gates or LVDS receiver can be also used, but a less exactly defined threshold and higher temperature dependency of treshold voltage and delay may be a disadvantage.

 

[sergio mariotti]

lpro 101 manual

In my opinion the lowest jitter circuits are proposed by atomic clock manufactures.

Take a look pag 17 of the following doc
**broken link removed** ( or google "LPRO User Manual")

You have to pay attention not only to phase noise and jiiter; drift is also important.
In order to minimize drift, the circuit board should be electrically short as possible and thermostatic.