High-speed signal isolation

[Time]

I am designing something where the application requires isolation of a some what high speed signal (1 MHz).  It is essential to isolate the source signal from the rest of the device and preserve the waveform characteristics at the same time.  The catch is that even though the frequency is only 1 MHz, the risetimes are on the sub-nanosecond (2ns to pico-seconds to be exact) order and the rise and fall times of the signal are the most crucial aspect of the signal to this application.

I have spent some time online sorting through various optocouplers and have found the fastest Tphl and Tplh (time to propagate high to low and vice versa) is in the 10s of ns realm.  I gave up on the notion of an optocoupler for this reason and began to search for another isolation method but have not really had much luck.

I was hoping someone here had any good ideas for isolating a 5V MHz signal with sub-nanosecond to pico-second rise/fall times without impacting these times and providing atleast a few kV of isolation.  The isolated signal can be TTL or CMOS, either would work just fine.

[Zero999]

You could try a pulse transformer but finding one with a wide enough bandwidth still might be a problem.

[jahonen]

Might be easier to decode the risetime information before isolation and then transfer it across the isolation barrier (whatever it is).

Regards,
Janne

[Zad]

How about a wide band RF transformer?

http://www.minicircuits.com/cgi-bin/modelsearch?model=ADT1-1WT&search_type=model
http://www.minicircuits.com/pdfs/ADT1-1WT.pdf

As an aside, notice the pricing breaks.

Buy 1 to 9 and pay $14:95 each
Buy 10 or more and it is $2:95 each

[Time]

I've been looking at those.  I have previously used some minicircuits x-frmrs for matching networks.  I am just not quite sure they have the few kV of isolation I need, though.

[alm]

You might be able to find suitable parts in the AD icoupler series or similar parts from Silabs (there's a third source I can't recall). Not sure about rise times, they support signals up to 100Mhz, but I'm not sure about rise time fidelity. Isolation up to 6kV or so. If rise time is critical in an analog way, modulating it might be the best solutiom.

[Time]

Those analog devices have the best specified performance I've seen so far at 3 ns rise/fall times.  I will probably give those a whirl.