AD8191: Crosstalk and isolation

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

I have a customer who is designing the AD8191 into a dual link DVI KVM and need
to know the isolation and crosstalk figures.

They are interested in frequencies up to the maximum single link data rate of
1.65Gbps, but any figures you have especially in the 100MHz to 200MHz range
would be greatly appreciated.


We don't have any explicit data on this. However, I can see several ways to
interpret this.

First, it would be helpful just to get an understanding as to why these
performance parameters are important to the customer. We might be able to
better produce desired data if we understand this.

We have been asked for similar specs in limited situations from people working
on high-security systems.  In such systems it is important that one user cannot
eavesdrop on another by "sniffing" the crosstalk from another signal. If this
is the case, the customer might not want to explicitly say this for security
reasons, and I will understand. But perhaps you know the nature of the business
they are in.

Note that these are digital parts. Therefore, the usual system concern is that
the crosstalk or off-isolation is not so severe that it causes bit errors. We
do not have such problems with these parts. If the customer is asking for more
than this, we would have to know more.

If both inputs are disabled and one is driven by a "normal" HDMI or DVI signal,
then the output will not switch. This is to be expected. However, I believe
that customer wants more than this. Does he want to know the residual
square-wave signal at the output relative to the input? Since he is asking for
frequency content, would it be useful to do this with sine waves and use a
network analyzer, like we do for analog crosspoints?

Further considerations: 

One input is on and driving the output. The feedthrough from the second input
can modify (increase) the jitter of the on channel. Is this what is desired
(i.e change in jitter)?

The isolation can be from one input (say Red) to the Red output. Or it can be
to different outputs (Red to Green, Blue or Clock). Does this matter to the

If just one channel is on, then there can be crosstalk among its channels, e.g.
Red to Green or Blue to Clock, etc. Is this important?

It is also possible for one input (set of 4 diff pairs) to crosstalk to the
other input (set of four diff pairs). Is this what is needed? Does it matter
which of the four signals of one input crosstalk to which of the four of the
other input?

I think you can see that this can get very involved if the situation is not
constrained in some way. As I mentioned, the condition of no bit errors is
sufficient for almost all "normal" users. My suspicion is that this customer's
needs go beyond the normal usage of the part. I want to help out further, but I
need to know better what the customer really needs.