On Thu, 03 Aug 2006 17:23:31 -0500, Ben Jackson wrote:

Most of the time I see RF stages coupled with .1u caps, which are
basically transparent at RF. Sometimes I see very small values
used, like 3-10p.

Depends on the frequency and the impedences.

A couple examples are in:

http://www.amqrp.org/kits/38spcl/schematic.html

C17 (near the middle top, connecting U2 to a tank) and then the output
of that tank to U4A (unmarked, but also 5p).

It's clearly important to the circuit (empirically) but I don't understand
what they're doing.

It's part of the matching network, the upper cap is the 5p and the
lower is 100p. Also by making the value small the amount of RF
coupled is lower ( capacitive attenuator) along with tuning L2.
It is a simple circuit but has a more complex analysis.

For example U2 on recieve is a product detector and the desired
output from pin 4 is audio. On transmit it's used as a mixer to get
the VXO and BFO to produce a ~10mhz rf signal at pin 4. So making
the path for RF use a smaller value to keep any audio out of the path
is part of the circuit "tricks".

Since the design makes many of the parts serve dual duty
the analysis is often less than obvious.

I can sort of see C20 (the unmarked second one) acting
as a high-pass RC filter.

It's doing two jobs, DC blocking and a somewhat high pass coupling.
However the impedence at that point is fairly high so coupling values
can be smaller. The device there is not an opamp but a inverter
(digital logic) from a CMOS IC pressed into service as an amplifier
so there are some things done arount it to make behave are as needed
to make it work.

Allison
KB!GmX