Thank you all for this great conversation. I want to remind everyone that we need long term coherence in the receiver, which can only be accomplished by low phase noise, not for the purity of the tone needed to be heard by a human ear, but by the need to correlate for long known symbols to recover frequency and timing in the SMS text messaging system and to do COHERENT detection of the symbols transmitted. We want the analog transponder to support this. This is also contributes to the requirement for a much larger dynamic range in the system than we have had heretofore on our spacecraft.
This system needed was added to the requirements for the receiver after the initial requirements but not before the design was done and certainly not before it was built.
73's Bob N4HY
Grant Hodgson wrote:
Jim Sanford wrote:
Do you have a different answer to my original question, "What is a good phase noise number to shoot for in a microwave narrowband (ssb/cw) system?" If so, I'd like to engage in that conversation, both for Eagle and for my terrestrial microwave station.
Needless to say, there is no right and wrong, it's very subjective. However, there are a couple of factors that can be used in calculating phase noise requirements.
For analogue systems, i.e. good old CW and SSB, there are two basic phase noise requirements for the local oscillators.
The first is that the LO needs to produce a tone that sounds 'clean' when listened to on an analogue receiver. The term T9 is often used, but is rarely defined. The parameter of Residual FM can be used to specify the cleanliness of a tone - it's subjective, but a value of 20Hz RMS IMHO is more than good enough for narrow-band voice communications. As it's a noise measurement, the bandwidth needs to be specified, and so 300Hz - 3kHz is appropriate for SSB. (Note that FM is different as much wider bandwidths are needed).
So, I would suggest a residual FM spec. of 20Hz from 300Hz-3kHz. This can be achieved with a 'flat' phase noise performance of -76dBc/Hz at offsets from 300Hz to 3000Hz, for example.
Residual FM can be measured directly, but not easily, or it can be determined from an integration of phase noise over the required bandwidth. A Google search will give the maths of converting phase noise to FM, I'll post a link if required. Software such as that from KE5FX has the conversion built in. Note that for the purposes of determining residual FM, it doesn't really matter what the shape of the phase noise plot looks like, it is the total area under the curve that counts. Also, the value of 40Hz is based on the response of the human ear - it is not based on the actual frequency of the local oscillator. Therefore, a 10GHz local oscillator with 40Hz FM will sound just as good as one with 40Hz FM at 1.8MHz. However, maintaining the required level of FM at microwave frequencies obviously becomes more challenging as the frequency is increased.
The second requirement is that for the phase noise spec. applies to reciprocal mixing; the process whereby a weak signal is buried in the noise of a strong signal which is at a different frequency. For the purposes of the Eagle analogue receivers, consideration needs to be given to the relative strength of any strong signals. The SAW filters will do a good job of reducing out of band signals, so we are left with the consideration of strong in-band signals.
For satellite work, the levels of the incoming signals should be much closer together than would be the case for terrestrial work. If we assume that the strongest signals will be 30dB higher than the weakest, at an offset of 5kHz, and that we don't want the strong signal to degrade the S/N of the weak signal by more than 3dB, then we can say that the phase noise requirement at 5kHz offset would be approximately -30 - log(2.7kHz) = -64dBc/Hz.
That might seem surprising, but it is based on a relatively narrow range of input signals. It assumes that the uplink signals are perfectly clean with no phase noise, and ignores radar interference. If the range of input signals is larger, then the phase noise needs to be better than this. Terrestrial stations need much better performance than this as they have to deal with a much larger range of input levels. I've got a feeling that this subject has been anlaysed before, but I couldn't find a mention of it whilst browsing EagleP.