Recreating a SSB carrier for RZSSB phase noise cancellation

Is there a way to re-create the carrier when a SSB signal has the carrier completely suppressed?

I'm interested in implementing a RZSSB (Real Zero Single Sideband) reception circuit using the zero crossings and taking advantage of the possibility of using a phase noise cancellation circuit.

References:

http://en.wikipedia.org/wiki/Reconst...zero_crossings

http://www.itu.int/dms_pub/itu-r/opb...2001-PDF-E.pdf

If I understand correctly, using RZSSB reception can eliminate atmospheric generated amplitude noise and then if a carrier can be recreated from a received SSB signal then the phase noise in the carrier can be subtracted from the phase noise in the signal and the resultant signal's noise is limited primarily by the noise in the circuitry only.

The RZSSB schemes I've read about look like they need a small amount of carrier signal to be transmitted but if there is a way to recreate the carrier when none at all is transmitted then it ought to be possible to take advantage of the RZSSB reception noise improvements with any SSB signal.

I read that a Costas loop cannot be used with SSB to recreate the carrier but I didn't understand why not. I also read that a squaring loop, which I assume means using a squaring detector, can be used to recreate the carrier. Does this mean sending the squaring detector output through a LPF and then using the output to control a VCO LO?
Does anyone know how this works?


Are there other ways to recreate the carrier so that it contains the same phase noise as the signal?

Else, is there a way to extract just the phase noise signal without recreating the carrier?

Thanks,

73
AF5IE

thinking aloud (40 years since I have been to College) :- squaring detector, carrier = SinωcT, modulation = CosωmT . SSB = sinωcT +SinωmT. This is like A+B, squaring gives A^2 +2 A B + B^2. i.e. a term contains carrier only components, albeit squared. Since sin^2 (x) = ?[1 ? cos(2x)], Cos 2 X ωcT can be filtered out. How does this rate?
Frank

Frank,

You remember your college math a lot better than me.

Are you saying:

Mix the received signal with itself to square it and this will output a result that includes the carrier component at twice its original frequency (A^2)?
Then low pass filter out everything at the carrier frequency and below (2 A B + B^2)?
Then do I divide the filtered doubled carrier to get just the carrier?
If that is correct, what circuit does a divide function, a mixer?
Then do I use the phase difference between the recovered carrier versus the phase of the L.O. to get a phase error signal?
Do I take that error signal and use everything below about 200Hz to lock the VCO L.O. and everything above 200Hz becomes my phase error that I subtract from the phase changes of the real zero crossings of the received modulating signal before I detect it with the RZSSB method?

You could just use a digital divider, or I would think just double the PLL locked oscillator, or if you are REALLY ambitious a regenerative divider (thats a blast from the past ).
I was thinking about if it could be possible, by using two reference oscillators, say nom. IF +- 100KHZ, to get the modulation side bands to cancel but not the frequency/phase shift of the missing carrier, not immediately obvious. I'll have to do some serious thinking)

Frank

I am pretty sure that with a perfectly supressed carrier and fully supressed unwanted sideband this cannot be done.

Consider that with a single tone modulation of frequency m, and a carrier of frequency f, an ideal SSB transmitter will generate a single sinusidal output at frequency f+m (upper sideband) or f-m (lower sideband), thus there is no way to identify the carrier frequency much less its phase.

Now if you are pepared to deliberately leak a little carrier power at the transmitter then locking a very narrowband PLL onto the carrier becomes a simple matter of filtering and required SNR, and this can in fact work very well (About the only radios that do not have some carrier leakage are the DUC SDR transmitters).

Also, atmospheric noise within the information bandwidth is indistinguishable from the signal with ssb, so I dont see how it can be cancelled in a scenario which inherently lacks redundancy.

73, Dan.

Yes , I am not convinced either, its just the words "squaring circuit" caused me to dredge up that thing about trig identities,I can't visualize how with an upper side band of 1.001 MHZ, you can regenerate 1 or 2 MHZ, thats why My thoughts wandered off to the two oscillator solution. Intuitively, I can visualize, the difference between two side bands one with a + frequency error, the other with a - frequency error. Its just the practicality of figuring out how to unscramble the frequency error among all the modulation. My thoughts are leaning towards, limiting the two demodulated signals.and putting them into a double balanced mixer with a low pass filter at its output. The only time you will get zero output is when the +- frequency errors are zero. I can see that it could work, but whether it's practical is another thing!
Frank

Yeah, I realized that about the digital divider after my post. I also started thinking that it would be hard to use I Q outputs of a low IF receiver like I was hoping because the carrier would be at or near zero Hz so its square could still be less than the signal information's frequency range.

I think I understand what you all are saying in your subsequent posts about the difficultly in extracting the carrier.

If I use the Weaver method at the transmitter end and at the receive end and I use 1.4KHz as quadrature input into the 2nd set of mixers then I'll have a hole in the audio at 1.4KHz.

Could I insert a small amount of 1.4KHz pilot tone into that hole and then lock into that at the receive end?

It seems like I don't even need the tone synchronized with the RF carrier. I just need the receive tone locked with the transmit pilot tone. Then once I have lock I can reduce the bandwidth of the receiver tone's VCO control signal to 10Hz or less and that ought to make its noise floor fall away. All phase noise above that frequency would then become the phase noise that I subtract from the phase noise in the signal.

But maybe the pilot tone's close proximity to the desired audio would make it hard to extract its phase noise independent of the desired audio. I'm still trying to figure it all out.

I want to come up with a way to do it in software using the audio range I and Q inputs to a PC.

73
George

Note that RZ SSB requires that the carrier not be fully supressed, as it needs to be recovered in the recever.

It may be possible to make this work without a carrier if both transmitter and recever are locked to the same master oscillator (Say a GPS derived reference), at least over paths not suffering from atmospheric fading (where the phase shifts due to multipath can only be compensated of the carrier reference is also transmitted over the same path).

I guess the 1.4KHz hole at the receive end would mean the pilot tone would be missing from my I and Q outputs? Maybe I'd need separate hardware to detect the 1.4KHz pilot tone.

If you control the transmitter as well then all this becomes fairly straightforward.

Lets say the LO is set to say 14.2MHz, and your I/Q modulator has a response down to DC, then by simply injecting a small amount of DC into the I channel you will get some carrier present.

If your I/Q source does not go down to DC, then simply modulate your SSB onto a 'local oscilator' in the I/Q domain at maybe a 10 KHz, with the baseband again carrying a small DC offset, if your resulting digital 1st IF is an I/Q Pair at say 10KHz, then retuning the rig to 14.190 USB will result in your signal appearing at 14.2 Mhz with a little carrier present (DC in the basemabd turns into 10KHz in the digital IF with USB audio above it, 14.190 + 10KHz = 14.200).

Regards, Dan.

If you insert some 1.4 KHz into the final balanced modulators, you will be engineering in carrier leakage, unless it is fed in to both and has the 90° phase difference between the modulators. It also would work (I think) if you just added the tone at USB carrier +1.4KHZ. One other technique could be to insert a LF tone (22HZ?) which would be inaudible with ordinary Comms equipment.
Frank

I think I see what you and Dan are saying. I was hesitant to consider adding some carrier on 1 side or the other of the SSB band signal because it would increase the total transmitted bandwidth. That is why I was thinking that if I insert it in the 1.4KHz Weaver hole it would be better. I was wrong about needing separate hardware. I forgot that I would not have the hole coming out of the 1st set of I and Q mixers. The tone would still be part of those I and Q outputs. It is only after the 2nd set of I and Q mixers that the hole would stop the tone from coming through. But if that 2nd set of mixers are in software then I can also use other software circuits to access the pilot tone.

If I have the pilot tone at the carrier (zero Hz) or at 22Hz or at 1.4Khz, I can't visualize how much BW I can have to the phase noise of that pilot tone. I mean, can I detect 3KHz BW of phase jitter when the pilot tone is less that 3KHz from the signal I want to subtract the phase jitter from?

It seems like the jitter would be including some of the voice signal and then when I subtract it from the voice signal then I'd be taking away the voice signal?

Or maybe only a small amount of the voice signal would show up in the phase noise signal so I wouldn't lose the voice signal when I subtract the phase noise signal from it.

Can this scheme work with the voice signal and the pilot tone in close frequency proximity?

Somehow I want to do it without using more transmission BW that the typical SSB would use, if I can help it anyway.

73
George

Anticipated that a single tone would be annoying to SSB listeners I was thinking, if it is possible to have a pilot tone in close proximity or even intermingled with the voice signal then not all the energy need be at 1.4KHz. More energy can be at 700Hz, 350Hz and 2.8KHkz and even more combinations in a kind of spread spectrum manner over the BW of a single SSB signal. As long as there is a known set of pilot tones whose energy can be recombined then it could be used to extract phase noise to be subtracted from the voice signal.

But when I try to visualize time domain and frequency domain aspects of the phase noise cancellation I can't figure out if the pilot tone(s) can be in close proximity and share the BW of the voice signal. It kind of seems like they can but I'm not sure.

This is just a thought - ignoring the technical issues.
Listening to 'raw' SSB through an AM demodulator it is clear that there is a correlation between the original voice and the resulting RF output, even though it isn't intelligible. Tuning through an SSB signal changes the recovered amplitude as the IF filter bandwidth is imposed on the baseband signal but it doesn't change the pitch of the audio. Would it be possible, maybe using DSP techniques, to 'translate' the sound from transmitted envelope back to the original modulating audio without re-inserting a carrier at all?

Deep maths is required... and it's late at night here.... and I've had a glass of wine.... maybe I'm just rambling...

Brian.