[time-nuts] Reverse isolation

John Day johnday at wordsnimages.com
Sun Mar 8 22:47:07 UTC 2009

At 06:14 PM 3/8/2009, Didier wrote:
>This question is directed at Bruce, but if anyone else has a contribution,
>feel free to speak.
>What is the best way to measure the reverse isolation of an amplifier
>(particularly a buffer amplifier for a 10 MHz reference), when it is
>expected to be in the order of 100 dB or more?
>Feeding the output with a known signal and measuring at the input with a
>spectrum analyzer comes to mind, but I am sure there must be something wrong
>with that technique, it sounds too simple.

No, not really. How else would you measure it? Reverse isolation is 
basically the reverse gain - S12 - of the amplifier. How does a VNA 
measure S12? Essentially inject a  signal at port 2 and see how much 
comes out of port 1.

If you want to get so picky as to determine S12 with a signal in the 
forward direction then you have a problem. Because the forward gain - 
S21 - is going to effectively swamp the signal going the other way. 
So this is almost impossible to measure if the signals are at the 
same frequency. In this case measure the S-parameters of the 
amplifier in its 'normal ' configuration, then de-embed the 
S-parameters of the device. Assuming the device is the only 
non-linear element you are dealing with then from the S12 & S22 
values you can also figure out how the device reacts in the reverse direction.

Then if you don't mind solving a large matrix you can figure out how 
the device might react to passing a signal in both directions. The 
reality is however that if the device is within its linear range, 
which it is likely to be if you want to distribute a reference or 
some such, the reverse behaviour of the well terminated amplifier 
will approach the nominal S12 value. The difficult part is to 
determine what happens when the amplifier is not nicely terminated!


>The presence of a signal at the input (or not) may affect the operating
>point of the amplifier, so measuring from output to input without such
>signal may not give a true result.
>time-nuts mailing list -- time-nuts at febo.com
>To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>and follow the instructions there.
>No virus found in this incoming message.
>Checked by AVG.
>Version: 7.5.557 / Virus Database: 270.11.9/1989 - Release Date: 
>3/7/2009 6:43 PM

More information about the time-nuts mailing list