Is there anything wrong with driving a comparator differentially

Is the input signal big enough to overcome any input offset voltage and input overdrive requirements?

The input signal is about 200 mVpp in each leg, i.e. 400 mVpp differentially, but it is a little noisy...

/Mikhail

The LT1719 isn't quite fast enough to handle 50MHz reliably - page 15 of the data sheet sets an upper limit of 62,5MHz on the output toggling frequency. and recommends the L1394 for application requiring fast toggling - but that part typically falls over at 100MHz, which doesn't leaving much margin to accomodate worst case parts.

The original Am685 comparator fell over at about 100MHz in the same sort of way, but the various 9685 follow-on parts got to 200MHz and above.

I'd be inclined to use a PECL line receiver or two folowed by a fast PECL-to-TTL converter - the Synergy SY10100ELT23 part is guaranteed to make 160MHz, and I believe that it was designed by our own Jim Thompson, whose circuits seem to be okay, even if his politics suggest incipient Alzheimer's.

Use an LVDS-to-TTL line receiver. Dirt cheap and blindingly fast.

John

Guys,

These are all great advices, but it's a production board I am talking about and I need to fix it without respinning. In all of my new designs I actually do use PECL and/or LVDS receivers...

Thanks, /Mikhail

A few things- the LT1719 does not like high impedance drive on the inputs, which go negative resistance at switch over, such as RF transformer leakage inductance. You would be better off opening the CT and placing 2x resistors each of value Rterm/2 in series directly across the full secondary and running the 2.5V bias to the resistor center tap.

That type of output usually indicates a transient common disruption on the comparator inputs. The LT1719 likes this on the low end- so you might consider lowering the 2.5V to more like 1V. There is also a partial latchup issue at turn on where the LT1719 does not like coming up with active signals in excess of 200mV being applied to the inputs. Since the comparator is fully characterized at 20mV overdrive, you might consider an anti-parallel Schottky clamp fed by nominally small damping resistors like 10 ohms or so from the input, and bypass the common mode bias point with a 0.001uF. The LVDS receiver might be fast but it stinks compared to the LT1719 insofar as sensitivity.

If it is a production board, you really shouldn't be using the LT1719 - to repeat myself

"The LT1719 isn't quite fast enough to handle 50MHz reliably - page 15 of the data sheet sets an upper limit of 62,5MHz on the output toggling

frequency. and recommends the L1394 for application requiring fast toggling - but that part typically falls over at 100MHz, which doesn't leave much margin to accomodate worst case parts. "

The 62.5MHz figure in the data sheet is clearly a typical limit, rather than a worst case guarantee.

Fred Bloggs has come up with an independet problem with the input stage, which strikes me as an equally valid reason for replacing the LT1719 with something that can be guaranteed to work.

Production hate it when you change a printed circuit board, but they hate it even worse when they have to junk a bunch of fully loaded and tested printed circuit board because one batch of one of the components came in at the low end of the manufacturer's guaranteed performance range. When they have to junk boards, they come rampaging through engineering, looking for scapegoats.

Thanks a lot Fred. That's the kind of input I was looking for!

/Mikhail

an RF

circuit

especially

distorted,