Coil losses

I think Tektronix uses that method in their diff input scopes. If you

Silicon delay lines are horribly noisy. I don't know this particular chip and it is very expensive but everywhere I found silicon delays I ripped them out and replaced them with analog "home-brew" versions.

Tempco does not matter if you complete the servo loop. Or in modern speak feedback via a 2nd path across the barrier. It's needed anyhow because analog methods are not very linear by themselves.

It was originally advertised as a pure ECL propagation delay device. The data sheet lists the propagation delay jitter as around 1psec - typical - which is regular ECL jitter, coming from Johnson noise and the edge speeds involved.

I worked out a calibration scheme which involved generating pulse-width modulated waveforms and digitising the DC levels with a decent A/D converter. I figure that one could calibrate 128 delays in about 1msec.

This was back around 1996 for the MC100E195, and we never actually built it, but it looked entirely practical.

The problem was likely cooling. There's no good way to get the heat out of a thin piece of FR4, with the heat flow in-plane.

With modern SMD packages such as DPAK I have that challenge often. 3oz or 4oz copper out to an area where you can sink into a metal mount with a large enough surface usually solves it.

I asked him this afternoon. The problem was the permativity of the FR4 and capacitance.