Hi - I am still ruminating over my rx design (look here -
http://g4oep.atspace.com/qrss/qrss.htm ) and am beginning to realise
that the inductors in the front end are a weak point re IP3. In the
absence of equipment to measure this, I have been thinking about the
1dB compression point, which is supposed to be 10 to 12dB below the
ip3. Compression is readily measurable, so I will eventually get
round to measuring it.

Meanwhile, can anyone point me in the direction of establishing a
theoretical derivation of the relationship between these quantities ?

The thing about the inductors is this - If we estimate 44dBm for the
IP3 of this type of mixer in its unmodified form, then take off 12dB
you are left with 1W for the 1dB compression point. This is 10Vpk
into 50 ohms, and I just don't think the cores I have used are up to
this with just 1 turn on the primary - they are almost certain to
saturate, or at least be close enough to it to produce some
compression. There is also the question of how the crystals would
perform at 3 times this level (connected to the secondary of the input
transformer).

Using tuned matching transformers in the filter is attractive because
they enable the filter to be tuned accurately on to the required
passband frequency, and supress spurious responses to an acceptable
level. But I feel that this part of the design needs a re-think if
IP3 is to be optimised. This raises the general question of front-end
filter design - even if a conventional LC filter is used, the cores
are definitely in the firing line and need care.

Andy G4OEP