[Elecraft] Intermodulation in Crystal Filters

[Jack Smith]

I've mentioned before that crystal filters are not perfectly linear 
passive devices and can in fact generate intermodulation products. The 
reason is that a crystal's motional parameters are not constants, but 
rather are a function of the applied drive, something called "drive 
level dependency" in the trade.

I've just completed a quick IMD measurements for a 4-pole 21.4 MHz 
monolithic crystal filter (ECS 21K7.5) and posted data at 
http://www.cliftonlaboratories.com/Updates.htm under the 03 July 
updates. This is a long update, in several stages, so read it all, 
should you find the first part interesting. The ECS 21K7.5 is a $10 part 
in single lots, so it's certainly not an expensive filter.

Monolithic crystal filters are known for IMD, and the test was run with 
a rather high drive level, around +9 dBm. However, the data is quite 
instructive and should convince anyone who doubted that crystal filters 
could themselves be a source of IMD. (These measurements are quick and 
dirty and I will refine them as I move towards a final filter design. )

Bringing the discussion back to the K3, it shows the importance of 
getting the roofing filter correct, as intermodulation generated in it 
will come in "down the pipe" in the DSP stage.

And, it's not just crystals that are non-linear passive devices. 
Capacitors and inductors also have the same problem. In the case of a 
capacitor, some dielectrics have a dielectric constant that is a 
function of applied voltage. Ferrite and powdered iron (but to lesser 
degree than ferrites) are also non-linear. Hence any component in the RF 
path must be suspect and carefully evaluated for IMD performance. (I've 
written about non-linear capacitance. See 
http://www.cliftonlaboratories.com/capacitor_voltage_change.htm, but 
have yet to make serious measurements on non-linear inductors.)



Jack K8ZOA
www.cliftonlaboratories.com
>