After a comment by G0API relating to something he heard from G3WDG after EME testing, I've been pondering...
Take a microwave downconverter with a virtually untuned front end (for example one using a quadrature mixer to give the image rejection and not much else). Use this with a synthesizer having a not-too-brilliant far out phase noise response, like any one with an internal VCO such as the dedicated ones in 'digital' LNBs.and teh LMX2541 etc.
This is exactly what we have in modern low noise LNBs like the Octagon.
Now: all that wide band noise tens and hundreds of MHz away from the carrier is all being converted by the far out phase noise of the synth and some of it is ending up in the wanted passband. In HF - VHF radios with strong signals in contests, this is the classic reciprocal mixing problem. But here we are concerned with much wider bandwidths and accumulated noise.
I've been thinking all day how to calculate it - and haven't come up with an answer. Yet intuitively it seems as if it ought to be straightforward to estimate - after all, classic reciprocal mixing is, when strong off frequency carriers are used.
Take as an example the LMX2541-2690 which has a PN value of around -157dBc/Hz at 20MHz offset. When multiplied *4 to 10GHz this becomes -133dBc/Hz. Assume for the sake of argument and to keep the sums easy, the PN response is flat and stops abruptly at +/- 100MHz. Ie we've integrated it to a brick wall shape . With a flat front end response, that means 200MHz of noise is being mixed down. We'll also assume the close in phase noise is excellent (which is quite reasonable with a decent reference and wide PLL bandwidth) so close in factors don't contribute.
I now want to multiply the 200MHz (effective) PN by the 200MHz wide flat noise input, and work out how much is mixed down and appears in,say, a 2.5kHz reference listening bandwidth. Common sense says bandwidths cancel and the effect appears as a raising of effective noise level (and receiver noise figure) but by how much ?
Answers on a postcard please ...