[Premium-Rx] Reaction Instruments 685 AM-FM demodulator

[Charles Steinmetz]

Peter wrote:

>The 2N5432 is interesting, while non stocked it is "available" from 
>Digikey for only $46.76 each.  Yikes.  There are some sellers of 
>them on ebay, one guy has them for $4.50 each plus $1 each shipping.

All of the 2N5432s listed on ebay are almost certainly counterfeit -- 
don't bother with them.  Crystalonics sells small quantities 
directly, so it's worth a call to inquire about both 4447s and 5432s.

>Costly, but if that's what it takes to fix this I would do it.  Of 
>course I could re-jigger the circuit to simply use something like 
>2N7000s which are dirt cheap.

Enhancement-mode MOSFETs are unlikely to work without major 
redesign.  Depending on the circuit, they may not work at all because 
of the substrate diode.  2N4447s are high-current JFETs of the type 
often recommended for chopper circuits or other switching duty.  As I 
said, the J106 should work fine but you might need to replace all 8 
(and check the pinout before you solder them in).

>The biggest step will be to design a filter which will give good 
>phase response for the narrow bandwidths I want' for example 10 
>kHz.  That may be tricky.  I suppose maybe using a crystal filter?

Oh, heavens no.  Crystal filters (any filters with a sharp cutoff) 
are the very antithesis of linear phase and will cause greatly 
increased distortion.  As you observed, the filters in the FM 
demodulator have very gentle slopes.  They aren't being used to pass 
some frequencies and reject others -- they set the conversion gain of 
the detector, similarly to how an I-Q detector works.  They could 
well be all-pass filters.  The passband is set elsewhere, probably in 
the IF stages (and that filtering also needs to have a linear phase 
characteristic -- Bessel filters are often used).

You may want to read up on the theory of FM demodulation with a 
balanced mixer, so you have a firm understanding of what to expect 
when you change things.

Best regards,

Charles