Single ended 50MHz sine to square wave

[walker1]

We need to convert a 50MHz sinewave (maybe 1Vpp) from a built in synthesizer to a 3.3 V clock signal to be used in our logic on another board in the other end of the box.
The 50MHz sine will most likely come in a shielded coax cable with ground.

We 1st thought about using a comparator, but we havn't found any yet, that is fast enough and available. Besides it will require some external components for biasing etc.

Line receivers, on the other hand, do most of this, are available, and plenty fast enough.
I have found DS90LV018A (National) from one of our suppliers, but the data sheet says, I should not connect any of the two inputs to a fixed voltage, as the device then will go into some fail-safe mode.
Or have I misunderstood something here.

Can I connect one input to, say, 1.5V and the other to my 50MHz source ?
I will still have to provide some external bias and termination, but we will be up and running.

Are there other solutions?

 

[Skogsgurra]

I have used small broad-band transformers for the same purpose. They give you complete isolation and a differential signal - if you need that.

I used a 30 MHz component, but there are transformers available for more than 200 MHz. Look at Fair-Rite, minicircuits or do a generic search for "broadband transformer" on google.

 

[VEBill]

In general, all you need to convert a sin wave into a square wave is a lot of gain and clipping.

Also, look at 'zero crossing detectors'.

With respect to moving the reference around, it is quite easy when you're dealing with 50 MHz because you can use resistors to set the DC level and then couple in the 50 MHz through a small value capacitor.

 

[nbucska]

You can use MC1651 or MC1652 comparator.

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[walker1]

I havn't found much about MC165?, except they seem to be for ECL.

Do they work for 3.3V logic as well ?

I know practically nothing about ECL, besides from its excistence, and that it involves -2V

 

[nbucska]

It is ECL, made by MOTOROLA but they don't support it
Most likely it has 3.3 V version, too.


<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[Ronimax]

Have a look at National's LMV7219. You will need some R's and C's for biasing. I don't know how you can avoid this.

 

[IRstuff]

a number of Analog Devices comparators should work:

http://www.analog.com/en/subCat/0,2879,759%5F776%5F0%5F%5F0%5F,00.html

TTFN

 

[nbucska]

ONSEMI makes MC100ELT21 (or MC10...): It needs discrete
parts (bias resistors, C cupling and back-to back shottkies
to limit the input ) And it is not cheap...



<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[sreid]

You might check out some of these.

http://www.linear-tech.com/pc/categoryProducts.do

 

[Carb0n14]

Shmitt Trigger maybe ?

 

[nbucska]

yes,sreid, the LT1715 would do it. It has hysteresis, but it needs passive input network.

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[Skogsgurra]

Having seen all variations on the 50 MHz sine to square theme, I still say that the HF transformer gives you most flexibility and the choice to use whatever logic family you fancy.

If you use a Pulse Engineering 55 MHz transformer and bias the low side of the secondary about 0.5 V below your logic threshold then you have galvanic isolation and you need no other circuit than a spare logic gate to convert the sine to a square. The isolation reduces interference so that a schmitt trigger isn't needed.

Costly? Depends on volume. Five dollars for one off. But no other problems. And, as I said, the galvanic isolation could be gold if you want to go between boards.

 

[Carb0n14]

Why not just use an op amp to get the level then use a schmitt trigger to convert it to a square wave, why are all the other componants needed ?

 

[jon1]

I would use a cheap 3X step-up RF transformer into a generic 3.3V schmidt trigger. It is simple, low power, low impedance, and very robust.

 

[walker1]

I think we still have some TT4-1a on stock somewhere.
We used it for 10MHz clock years back.

They are good for at least 180 MHz (-1dB) and can deliver 1:2 ratios, so ....