BATC Forum

Hi,

I'm in discussion with a local radio ham about the modulation characteristics of the mpeg2 RF signal being generated by the Portsdown tx (or any other similar tx).

I'm having trouble explaining that even with QPSK and only 4 points on the constellation there is significant peak to average power variation as the signal phase moves across / round the constellation - which is why we need linear amplifiers and not Class C as this ham wants to use.

He seems hung on there being no amplitude content and no matter how I try to explain it I can't seem to get the message over - any idea as to how I can explain how the modulation works in simple terms and why linear amplification is essential ?

73 Tim

Its difficult when people have fixed ideas and lose the capability to challenge their own assumptions. Especially if they have invested significantly in unsuitable hardware. How about demonstrating the effect on a spectrum analyser - or if that doesn't work, transmit at full power on an adjacent channel while he is trying to receive.

Mike

Hi Tim,

If you look at the I and Q waveforms after the Nyquist filter you will see they resemble sine waves. These signals are applied to balanced modulators. The 2 RF outputs are then combined to become the QPSK signal. When I and Q change at the same time the RF amplitude passes through zero during the transition. The figures at this website should help your friend understand.

John

http://www.movingpixel.com/DemodEye.jpg
http://www.movingpixel.com/main.pl?demodSW.html

Hi John, Mike,

thanks for the 'sanity check' - the basic problem is that he wants to use 1.3GHz YIG oscillator and directly modulate that with an IQ signal.

I'll forward those links on but doubt they'll make much difference,

John,

It's that zero crossing point I've been hammering on about, but I suspect he thinks the signal always goes in a circle and thus constant amplitude.

Clearly there's a number of issues when using a YIG, not least with the instantaneous phase stability and phase pulling of the oscillator itself - despite my best efforts to focus him on the 'correct' path I suspect he's going to be in for a lot of pain, money and wasted time.

Ah well such is life....

73 Tim

Hi Tim,

As long as the YIG osc is followed by an isolation amplifier then something like an AD8346 modulator, there shouldn't any issue with the YIG getting pulled. Frequency stability associated with a YIG is a separate issue, and will also need addressing.

QPSK isn't a constant-power modulation scheme, and yes, a linear amplifier is required. See https://www.allaboutcircuits.com/upload ... QPSK_1.JPG it's a pretty basic graphic; the transition from 45 to 135 degrees hopefully make it quite clear that QPSK has a significant amplitude variation.

If that explanation doesn't ring true, how about a data transition which moved diagonally from top-left to bottom-right of the constellations (through zero), how might that be possible without an amplitude variation ?

Maybe he is getting mixed up with GMSK, which is a constant phase scheme used by GSM, which has little in common with QPSK ?

Why not just send this whole thread to the person concerned ?

Regards, Mark

Hi Mark,

thanks for taking the time respond and explaining the modulation in simple terms - something I'm not too good at.

I agree that a YIG 'might be ok' as a LO source but there was discussion of either directly modulating the yig or using delay lines to create the phase shifts rather than using a real IQ modulator - I think by now there's probably a better understanding of zero crossing and the effect on amplitude etc.

I suspect the person concerned who is a member of this forum may at some point read this thread and as a result understand QPSK a little better and will not have wasted too much time any money trying to re-invent the wheel but end up with a square one !

73 Tim

QPSK has a crest factor of 1 and a PAPR of 0 dB if and only if it has infinite bandwidth.
As soon as you start to filter it, that is no longer the case.

- Charles

Hi Charles,

yes agreed and tried to explain it that way as well which is why base band filtering is essential and not performed at the carrier frequency (as has been proposed) which would lead to all sorts of problems.

I wonder if a section on digital modulation as used by Portsdown and similar tx's would be useful to others trying to grasp the principles of how it works ?

Certainly there's a lot of info on the web but something in simple terms might be more understandable ?

I guess there's parallels here to when SSB first became viable for amateur use and the need to remove the unwanted carrier and sideband before hetrodyning to the final frequency.

Except now were talking about I and Q rather than LSB and USB.

73 Tim

Good luck with that!

Andy G4JNT covers this sort of stuff in his Datacoms column, TAPR published a book on it in 1996
and just about every EE undergraduate communications text book covers it.

Amateur Radio is supposed to be about self learning and part of self learning is going out
and discovering things for yourself.

Here you go this is what we need (if you have £1432.00 to spare)

http://www.open.ac.uk/courses/modules/tm355

- Charles

May I sound a bit of a contrarian? I don't think the fact that QPSK must transit through zero has much to do with the certain need for linear amplification. Morse, for example, switches between RF and zero on every dah and dit but doesn't need linear amplifiers just an LF filter on the key and a LPF after the PA. The real issue is surely intermodulation. The usual QPSK constellation diagram is a blurry time domain view. If we imagine a random digital modulation on I and Q, the power spectral density is the Fourier transform of the autocorrelation function of this random signal. We are familiar with the sin(x)/x form that results from this. The sin(x)/x form tells us we have smeared our modulation right across the section of band we are occupying. Whenever there are intended multiple in-band signals we know that odd-order intermod products will appear on and around the band in use. Those on our band just corrupt our signal but those around our band corrupt adjacent channels. Good practice suggests our signals should be 50dB down in the adjacent channels which implies low intermods, i.e. linear amplification. The 'ears' of sin(x)/x also show that we need some filtering before we plonk the modulation on an RF carrier to avoid building in adjacent channel interference.

That's my take on it, anyway.
73, Ron G7DOE