Re: LO Phase Noise and wideband front ends
On 13 June 2016 at 20:44, Michel Bourdon michel.m.bourdon@... [ukmicrowaves] <ukmicrowaves@...> wrote:
Yes indeed, I think my thoughts about wide input bandwidth and wideband Phase Noise were one of those cases of intuition and instinct going wrong.
After posting in parallel to the RSGBTech Yahoo Group, the two Peters, 'RZP and 'PLX, responded with sensible suggestions. G3PLX confirmed the thoughts of Michel, above, which I was I was coming round to anyway, after getting nowhere with a calculation method; that mixing noise alone has no contribution. Bandwidths cancel and there is no net increase in noise in a given bandwidth.
G3RZP may then have hit the nail on the head when he mentioned phase noise falling into the Rx passband. And more than likely explains the real observed effect of carriers being less "readable" with a poor wideband PN synth as the LO, over a crystal / multiplier source, in spite of the fact the carrier itself is 'clean'.
For microwave systems with a low IF, like 144MHz, a something-like -140dBc/Hz phase noise at 144MHz will leak across the mixer rejection. A simple two diode balanced mixer may only show 20dB rejection, so for a synth delivering +10dBm into the mixer, there is -116dBm of noise in a 2.5kHz bandwidth. [*] Which, unless there is plenty of front end gain is very definitely going to be noticed.
[*] -140dBc/Hz + 10dBm - 20dB mixer rejection + 10.LOG(2500Hz) = -116dBm
The comments from the two Peters in their two RSGBTech posts are copied below
From the RSGBTech Group :
If you start by supposing that your mixer is a straight multiplication (in
DSP terms) of the signal and the local oscillator, and further suppose that
your signal is a carrier poking up above flat wideband noise, and your local
oscillator is ALSO a carrier with flat wideband noise behind it, then I
think you should be able to calculate the contribution to the intermediate
frequency output which arises from the signal-path noise multiplied by the
local-oscillator noise. This will be an infinite sum if your two wideband
noises are infinitely wide, so you probably do have to guess at a realistic
upper limit to it. This will be a 2nd-order term, in that it will increase
by 2dB for every 1 dB that you increase both the signal-path and local
oscillator noise levels.
Then you subtract 3dB, on the basis that your mixer is being driven hard by
the local oscillator so that the amplitude of the local oscillator doesn't
come into the equation, only the phase.
I suspect that you will find that if your local oscillator phase noise,
measured in the conventional way is -P dBc/Hz then the magnitude of the
noise that is bothering you will be of the order of -2*P. So small that it
shouldn't keep you awake at night.
The situation is also complicated by the fact that the phase noise of the LO
may not be negligible on the receiver frequency, and is thus also down
converted to the IF: it can be the dominant factor and was one reason for
later WW2 microwave radars going to a balanced mixer using a hybrid T.
Similarly, phase noise at the image frequency will be rejected in the
same way only by the amount of image rejection and balance. The problem
is very similar to that of phase noise in Direct Conversion receivers
being mixed down to base band and limiting achievable sensitivity.
Somewhere, I have reference for the effect of phase noise at signal
frequency in early microwave radars, but it isn't immediately to hand
and I'm not going through 20 feet of shelved books to try and fi