Grant :
Thank you very much. I will attempt to put together all your remarks, answer the questions, and then embody the result in a spreadsheet and upload it to EaglePedia some time next week. I am off today to Seattle for a few days. That will also give Jan a couple more days to comment if he wants to.
I agree with your statement that my argument neglects phase noise from the "interferer". I am making the assumption that most at > 20 kHz away. This comes about because most users seem to have their training from HF ham bands where the USB users are above the middle of the band. In our case, with the SDX our desire to put the beacon in the middle since that will otherwise be degraded performance territory anyway ( the 1/f - DC centered noise of the A/D's).
We can then use the analysis tool and expand it to encompass all of our LO/Mixer combinations.
Bob
Grant Hodgson wrote:
Bob
I think your statement
and we worry that
RN (reciprocally mixed noise) = S * L * B will appear and hurt us to the point of signal being covered and for example, SMS not working.
is not quite right, because it doesn't take into account the gain of the mixer or the level of the LO signal.
A much easier way to analyse the effect of reciprocal mixing is to just consider the differences, rather than absolute levels. That way, constants such as mixer noise, gain and LO power get cancelled out.
Considering the phase noise of the satellite's receiver, and making the following assumptions :-
the phase noise of the LO is flat across the receiver bandwidth at the offset of the interferer there is only one interferer the mixer is not being driven into compression the mixer has a flat frequency response
then, for a given on-channel input level which you called Text, and working in dB, the noise power in the IF channel due to reciprocal mixing is :-
X = (I - Text) + [10log(B) + L] {eqn 1)
where I is the interferer level in dBm, Text is in dBm, B is the Rx bandwidth in Hz and L is the Rx phase noise in dBc/Hz at the appropriate offset.
For example if the interferer is 45dB higher than the wanted signal, the LO's phase noise is -82dBc/Hz at the interferer's offset frequency and the bandwidth is 500Hz, then X = 45 + (27-82) = -10dB. I.e. in this case the noise contribution to reciprocal mixing would be an extra 10%, degrading the signal/noise ratio by 0.4dB.
However, the above model is slightly too simplistic, as one has to also consider the phase noise of the interferer. This will fall right into the Rx passband, and no amount of filtering will remove it. We don't know what the phase noise of the interferer will be, but we can guess. If the interferer is another ham transmitter, then a reasonable first assumption is to assume that it's phase noise will be similar to the receiver's. Therefore the equation above needs modifying :-
X = (I - Text) + [10log(B) + L + 3dB] {eqn 2)
which in the above example would give 20% extra noise contribution, reducing the S/N ratio by 0.08dB
Equation 2 can be modified to accommodate more than one interfering signal.
What are the offset frequencies of the interferers likely to be?
regards
Grant
Robert McGwier wrote:
We have done all of these calculations for the links and published these spread sheets. The spread sheets show the typical bias towards taking thermal noise and some "guesstimate" as to interference levels but not really taking into account LO noise.
"Everybody" can understand that noise performance in a mixer that has a 50 ohm input port (which is hooked to a 50 ohm load) with the pretend noise power density of -174 dBm/Hz (k T0). The noise factor of the mixer is taken into account in the usual way
Noise Factor = kT0 (F-1) where k is Boltzman, T0 is room temperature and F is the noise factor in the mixer.
But, we have these oscillators that we say we want tunable. I doubt we actually want them to be tunable but let's start to go through this so we can calculate if I am right.
Reciprocal mixing is the crap that gets thrown into your receive pass band by a strong signal (strong compared to our weak SMS text signal for example) and raises the receiver noise floor. We have these computed signal strengths at the spacecraft and we do not want to raise the noise in our signal bandwidth by more than a few dB (none?) ;-)
Let's make some simplifying assumptions that the bottom end of the passband will be as empty as it has always been and the upper half band will have the usual ten suspects in them and not much more. The usual 20 people using the linear transponder on the satellite.
So we will assume that we are out in the "flat part" of the LO noise and that a signal of level S is out there.
Our bandwidth of our desired signal (the undesigned SMS text message signa for example) is B and the LO noise floor is L.
and we worry that
RN (reciprocally mixed noise) = S * L * B will appear and hurt us to the point of signal being covered and for example, SMS not working.
SNR(in)/SNR(out) = F + RN/(kT0 B)
If we wish to see signal level Text at our receiver then we can estimate the noise floor required of the LO (L) to be (using the previous formulae)
L = Text - S - C/I - 10log10(B)
where C/I is the carrier to interference ratio required at the output of the mixer. Notice that the reciprocal noise specification depends directly on the input blocking signal strength S. So this is a bit more complex than one might think at first so some more thought needs to be put into this before we leap to say the 70 cm design is good, bad, or indifferent.
We need to resolve this in the next few days so John, Juan, and the rest of our receiver friends can have a design to build to. Please everyone, check over my formulae before I go and do a ton of work that is all wasted. If you don't like the form of them, suggest something different. I suggest we build a spread sheet that does these calculations in a "mutually coupled" way. I am off to Lyle's on Thursday and will not be able to do much more on this before I return. Somebody jump all over it please if you want to.
Bob
Via the Eagle mailing list courtesy of AMSAT-NA Eagle@amsat.org http://amsat.org/mailman/listinfo/eagle