differential coax driver

I've been driving a 75 ohm coax differentially just fine for years but with my new batch of cable, I'm limited to about 40'. I'm using old
Taxi chip sets from AMD. Signal levels are psudo ECL, ac coupled at the driving end, emitter type drivers. Biasing at the receiving end to about 3.7V, with what was 75 ohms. But now that I'm having to look into the situation more closely, I can see from experimenting with a pulse generator and scope, that the best line match is achieved at 50 ohms. Taxi uses an embedded clock. The recovery circuit extracts the clock and will report link errors with a status line. As the link length increases, so do the link errors. I need to go up to about 80'. Baud rate is 143MHz. 5/4 encoding. 800mV p-p. I can see that as the link gets longer, there is some over and undershoot at the reciever occationally but only on one side of the differential pair. I suspect this messes up the clock recovery and behold the errors. The psudo ECL, emitter driver drives into a .1uf dc blocking cap, and there is is a 440 ohm pull down between the driver and the cap. I tried reducing the pull down to 220 to quicken the fall time, and increased the cap to .2uf to decrease baseline wander. Neither had any effect. I'm driving from 10 to 80 feet of coax depending on the installation. Using a simple resistor biasing network with 62 ohms to +5, and 255 to ground which also serves as the cable termination. This net biases the input a bit high, about 4V but works best of what I've tried so far. Prior to now, I was using a 75 ohm network. The 50 ohm works much better. Also used a 50 ohm biased to 3.7V, but it did not work as well as the 4V bias either. Measured cable capacitance and prop delay at about 24pf/ft and 1.5ns/ft.
Any ideas? since I don't see the over/undershoot at the driving end, and only on one side of the differential pair, I'm looking at the reciever board layout now.
Thanks, Bob
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.