The following notes are based on a discussion that was conducted on the UK Five Megs Yahoo group in February 2013.
The exchange was triggered by comments relating to the suitability using NBFM on the band, and if it was possible to operate without unwanted sidebands falling outside permitted frequencies.
I'd considered right from the start that NBFM would be an ideal mode for inter-GQSO's on 5MHz (or 1.9MHz for that matter).
But I didn't want to rise it as possibility in a public forum. As I knew that I'd be burnt at the stake for even suggesting it.
However I've followed previous discussions on this subject. Including the Carsons rule calculations, and some of it just didn't seem correct based on past experience.
About 30+ years ago when I lived in Manchester. I almost exclusively used NBFM on 2m. I was able to regularly work stations at distances of over 100 miles. The trick was to use low deviation +/-2.5KHz (as now required for 12.5KHz channel spacing) and narrow RX IF filters. If the going got tough I'd listen using SSB. This is because low modulation index phase modulation (which is what a lot of rigs actually used at the time, rather than true FM) is practically the same as AM. Just the phase relationship of the first order sidebands is different. See further down the page for more information.
Whilst modifying my FT-857D for operation on 5MHz. I thought I'd take a look at the TX bandwidth in different modes, and capture some plots on the spectrum analyser.
This was quite revealing. So I decided to use the opportunity to take a look at two more radios each of which show slightly different characteristics.
First a plot of an FT-857D.
Peak carrier level is at -10dBm. As you look further away from the carrier center frequency. The skirt of the AM signal is quite considerably wider than that of the FM signal.
I believe this is due to non-linearity of the RF amplifier stages, and the ALC action creating additional sidebands.
In fact at -80 / -90dBC the SSB and NFM bandwidths are practically the same as each other.
Occupied peak bandwidth at -30dBC for each mode is as follows:-
USB = 2.8KHz
AM = 5KHz
NBFM = 5KHz
Occupied peak bandwidth at -60dBC for each mode is as follows:-
USB = 9KHz
AM = 12.5KHz
NBFM = 11KHz
Next a plot of my Flex 3000 - Not quite as good as I'd hoped for.
Hmmm. It looks like the TX image rejection and IQ amplitude phase needs adjusting to achieve a better null.
Note that I used the 2.5KHz deviation setting to obtain this NBFM bandwidth
Last an Icom IC-7000 - Again not as good as I'd hoped for.
One quirk of the Icom is that in order to select narrow deviation on TX. You need to ensure that you have the 7KHz RX filter selected. The deviation is still a bit too wide for my liking. The service handbook doesn't state how to adjust this value. So I had to turn the Mic gain down to 1% (Mic gain adjustment from 1 to 100% doesn't make a lot of difference !) In order to get an acceptable TX bandwidth.
Listening to the IC-7000 transmitting with USB selected on my Flex. I actually couldn't tell much difference between USB, AM or FM transmissions. Note that in common with most radio currently on sale. All three of these radios have a two tone IP3 figure of about -30 to -35 dBC. Which is why the skirts look so bad.
So I'd say that, providing the maximum deviation is set to +/-2.5KHz. It's perfectly possible to fit an NBFM signal into some of the wider 5MHz channels. My experience is that NBFM is comparable to AM in terms of distance per watt, but much easier on the ear. In terms of background QRM /QRN. Just don't run too high a power level if your PA stage isn't up to continuous operation.
I'm just surprised that the is not much use of NBFM on other suitable LF bands such as 160m. Where it could offer quite considerable advantages for local QSO's.
Here are some additional comments from Peter G3PLX:
Martin raises an interesting point. Although the sidebands of an FM signal extend to infinity on each side in theory, it is very easy to design a' perfect' FM modulator that keeps the sidebands to the theoretical level.
Designing a linear amplifier or AM modulator is much more difficult, with the result that the 'splatter' from an AM or SSB signal can be much worse in practice than the higher-order sidebands of an FM transmitter, at least this is true of the 'far out' sidebands, as Martin has verified on the bench. I think this was known in the days of analogue PMR on VHF. It was much more difficult to meet the very severe adjacent-channel emission specifications with AM than with FM, so it was FM that was chosen when congestion forced a move from 25 to 12.5 kHz channels. One might have thought that they would have gone to SSB in due course but they never did, for the same reason the practical non-linearity made SSB splatter far worse than the theoretical high-order FM sidebands. However, this advantage of FM over AM or SSB doesn't apply to 5MHz NVIS, where the propagation is such that we don't need adjacent-channel protection ratios as high as they do on VHF. We can get away with the 20-30dB adjacent channel splatter of a typical SSB transmitter simply because all the signals we work are roughly the same strength with NVIS. As I have said before, if the adjacent-channel splatter from FM is no worse than it is with SSB then it should be acceptable on 5MHz, but that doesn't prove that FM would show any advantage over SSB or AM.
The difference between FM and phase modulation is very subtle. If you took normal audio, gave it a 6dB/octave rising frequency response, and fed it to a true FM transmitter, the result would be indistinguishable from PM (and vice versa). In practice, since comms-quality NBFM audio is given a 6dB/octave boost (pre-emphasis) in the top half of the audio band, the NBFM that we use on VHF/UHF is effectively a mixture of the two - FM for the lower frequencies and PM for the higher frequencies. The two lowest-order sidebands of a true PM transmission are exactly the same as the two sidebands of an AM transmission, and this means that it's fairly readable on an SSB receiver, and indeed 2.5kHz deviation comms-quality NBFM is just about readable on an SSB receiver provided there isn't too much low-frequency content. (i.e. there isn't too much of the FM part which is certainly not readable).
There's an interesting possibility for another kind of modulation which I have never seen used (at least not in the analogue world) but which has some useful features. Imagine an AM signal then remove the carrier to create double-sideband then re-insert it with a 90 degree phase-shift. Draw it on a phasor diagram and you see that 100% modulation only needs twice the peak power of AM rather than 4-times, and there is no downward modulation to zero amplitude so it's far easier to keep the transmitter linear. It's a bit like phase modulation but with ONLY ONE pair of sidebands. It's performance would be identical to that of AM, but with the bonus of improved immunity to impulse noise. Such a signal would splatter far less than AM or SSB because of the relaxed linearity constraint and less than FM because there's only one set of sidebands. This mode would have been a better solution to VHFPMR channel congestion than any of the others, but then everything went digital.