I know it's not implemented yet, and it's unknown if it ever will be, and
I have very little interest in SSB anyway, but I can't help to wonder.
What is the original idea?

There's a synthesizer that can produce a single frequency, and a way to
modulate this frequency with amplitude modulation, and that's really all
there is.

Is it even theoretically possible for the MCU to do some magic with this
modulation that will result in voice on a USB or LSB, without any other
hardware tricks?

Pretty sure the idea is to use an EER (envelope elimination and restoration) technique similar to what Guido PE1NNZ implemented on his QCX modification and later (tr)uSDX. There's a big thread here talking about: /g/QRPLabs/topic/29572792
--

73
Andrew KI7FXL

This might give some idea:

It is already done.

73

Barb WB2CBA

Presumably when ssb is implemented other modes like psk31. and the like will become available?? or am I on the wrong track?
. Paul DJ0CU G4ADF?

I asked GPT4 and it gave me a complete code segment written in C using the Hilbert transform, which creates a 90° phase shift and mixes it with original signal. I'm not a C expert so can't say how valid the code is.

Paul DJ0CU G4ADF.

It's pretty intuitive if you think about what an FM signal looks like in IQ representation, and what an AM signal looks like in IQ representation. You can take any signal in IQ space and determine its FM component and its AM component. The QMX can rapidly retune its oscillator thousands of times a second, and it can PWM the transistor feeding the PA, to recreate the FM and AM components respectively. This will be pretty kickass if it can be pulled off with this hardware.

As I see it, the main limitation, once you have the computing power, is the lack of linear amplification. If you had a linear amp, you would simply generate the IQ data and amplify it.?

I may have a simplistic view, and I hope to be corrected, but it seems to me that an EER method can only reproduce a single tone at a time. The code tracks the loudest frequency and duplicates. The wave shaping circuit changes the volume. ?This can not generate a true two tone signal.?
--
Colin - K6JTH?

开云体育

An even more simplistic view is that the audio signal is a single wiggly line, what you would see on an oscilloscope.? The subtle changes in that generate the multiple tones of speech, even those of a full orchestra.

What seems to be planned is to make changes to the frequency and amplitude of the RF signal to replicate that, with the appropriate modulation.? If you were doing AM all that would be needed would be to change the amplitude. For a 3 kHz bandwidth at least 6000 times a second.? For FM changing the frequency, also 6000 times a second.? SSB needs changing both.?? The challenge may be to synchronise the frequency and amplitude changes so they happen at the same time.

Chris, G5CTH

On Sat, Sep 2, 2023 at 12:08 AM, Colin Kaminski wrote:

I may have a simplistic view, and I hope to be corrected, but it seems to me that an EER method can only reproduce a single tone at a time

This is sort of the equivalent of saying "but a DAC can only produce one voltage at a time, how can you reproduce an entire waveform that way?"

The magic happens when you do it many thousands of times a second.

What does the waveform of a two tone signal actually look like? The actual waveform, if you squint at it really hard, just looks like a single tone that's being amplitude modulated at the beat frequency between the two tones, and inverts polarity whenever the amplitude modulation has a zero crossing.

An SSB signal, on the air, if you squint at it real hard, is just a single tone that changes frequency and amplitude very rapidly. You can reproduce this by doing the same thing, retuning an oscillator very rapidly and amplitude modulating it.

On Fri, Sep 1, 2023 at 11:31 PM, Shirley Dulcey KE1L wrote:

The STM32F446 microcontroller has two 12 bit DACs. A DAC output is used to control the amplitude of the transmitter output; it's not done with PWM. The current firmware only uses that to control the rise and fall times of CW transmissions.

If Q507 is being driven linearly then I hope it has good heat sinking. I would have thought that in order to keep the efficiency benefits if class E you'd want to switch the uc pin driving that transistor from an analog DAC mode to a pwm mode.

On Sat, Sep 2, 2023 at 10:18 AM, Stephan Ahonen KE0WVA wrote:

An SSB signal, on the air, if you squint at it real hard, is just a single tone that changes frequency and amplitude very rapidly. You can reproduce this by doing the same thing, retuning an oscillator very rapidly and amplitude modulating it.

Fake news.

On Sat, Sep 2, 2023 at 02:18 AM, Stephan Ahonen KE0WVA wrote:

What does the waveform of a two tone signal actually look like? The actual waveform, if you squint at it really hard, just looks like a single tone that's being amplitude modulated at the beat frequency between the two tones, and inverts polarity whenever the amplitude modulation has a zero crossing.

When I was writing my Teensy version of the uSDX I thought one day I would inject a perfect two tones programmatically and learn some things about the Si5351 frequency generation algorithms but much to my surprise the result was exactly as Stephan says but without any squinting.? Two tone testing the uSDX method of SSB is only a test of the AM modulator.

I found that the SSB method is/was limited by the I2C bus speed and a more powerful processor does not gain you anywhere near as much improvement as one might expect.? It will be interesting to see what Hans can do as he seems like a very smart fellow.

Ron? KE1MU

Surely this just amounts to creating? at any given point in time only one sampled frequency
and then switching to another fast enough in order to give the effect of two or more
simultaneous tones ?

In an analogue system? then ALL the tones of speech are present simultaneously.

--
- 73 de Andy -

On Sat, Sep 2, 2023 at 04:24 PM, Da Amazin' man G0FTD wrote:

In an analogue system? then ALL the tones of speech are present simultaneously.

Excepting xtal filter group delay params.

That's why SSB always has that SSB sound, unless you're using a DC RX ;-)

--
- 73 de Andy -

开云体育

Yes, but in the RF domain you need frequency and amplitude.?

73, Willie N1JBJ

On Sat, Sep 2, 2023 at 10:24 AM, Da Amazin' man G0FTD wrote:

Surely this just amounts to creating? at any given point in time only one sampled frequency
and then switching to another fast enough in order to give the effect of two or more
simultaneous tones ?

In an analogue system? then ALL the tones of speech are present simultaneously.

--
- 73 de Andy -

In the simple case of two simultaneous tones, you are not modulating frequency at all. You can produce two tones solely through amplitude modulation.

Think about the spectrum of an AM signal - an AM signal produces two full sidebands without changing the frequency of the oscillator at all, simply by modulating amplitude.

To produce an SSB signal, you add some frequency modulation to the above procedure. Not by tuning the oscillator to every frequency in the entire signal, but rather by tuning the oscillator to the "instantaneous frequency" of the signal at any particular moment, the same way you produce a full spectrum of sound by making a DAC produce the instantaneous voltage of the signal from moment to moment.

The real trick to understanding this will be to develop an understanding of the IQ representation of RF signals. Seeing what an FM signal looks like on an IQ diagram, and seeing what an AM signal looks like, and understanding why those signals look the way they do, is really the key to understanding the fact that every arbitrary signal can be broken down into FM and AM components.

On Sat, Sep 2, 2023 at 02:17 PM, Stephan Ahonen KE0WVA wrote:

The real trick to understanding this will be to develop an understanding of the IQ representation of RF signals. Seeing what an FM signal looks like on an IQ diagram, and seeing what an AM signal looks like, and understanding why those signals look the way they do, is really the key to understanding the fact that every arbitrary signal can be broken down into FM and AM components.

Conversely, given sufficient bandwidth and precision of the IQ data stream it is possible to generate any arbitrary signal, or combination of signals, of any modulation types.??
73, Don N2VGU

On Fri, Sep 1, 2023 at 10:08 PM, Colin Kaminski wrote:

I may have a simplistic view, and I hope to be corrected, but it seems to me that an EER method can only reproduce a single tone at a time. The code tracks the loudest frequency and duplicates. The wave shaping circuit changes the volume. ?This can not generate a true two tone signal.?

If you look in the audio domain, you have a two (or multi) tone signal. ?However in the RF domain, you see close to an almost pure single frequency RF waveform with nearly invisible frequency changes (on the 20M band, SSB modulation only changes the RF frequency by 2.5kHz/14MHz or a max of 0.02% change). ?Those changes are very slow (audio rate) compared to the RF frequency, but can include an array of many microscopic sidebands, a lot more than just two. ?So you can change a narrow-band close--to-pure RF waveform in amplitude and phase by tiny amounts at a relatively slow rate, and end up with tons of audio frequencies when those tiny-at-RF sidebands are modulated back down to baseband, where they now become large differences relative to each other in terms of rich audio.
--
73, Ron, n6ywu

One of the earlier instances where the "trick" of getting linear signals through an ostensibly non-linear signal path was brought to the attention of the amateur radio community back in the mid 1970s with the launch of the OSCAR-7 Satellite.

In that case, a method called "HELAPS" was used - and has since been used in one form or another on other amateur radio satellites equipped with linear transponders since then:? A bit of history may be found here:? ? (There's a scan of the HELAPS design document from the '70s out there, but I was unable to find it on a search just now.)

In short, you do this:

- Take the analog signal source - whatever it is, convert it to a lower frequency (I believe that they used 10.7 MHz) and band-pass filter it, then split it into two signal paths.
- On one signal path you do a hard-limit (as is done for analog FM reception) of the signal to eliminate the amplitude component.
- On the other signal path you do an envelope detection to eliminate the frequency component.
- The 10.7 MHz signal from the limiter was frequency converted to the final frequency and applied to the drive of the final amplifier.
- The supply voltage of the Class-C final amplifier was modulated - using PWM (switching supply) techniques with the envelope-detected voltage that we derived earlier.

The result, if done perfectly, will be a faithful recreation of the original signal.? In this case, it was a transponder with many SSB and CW signals - but it could be combination of any type of signal, of any mode.

In the case of the QMX we have the basis of the above:

- A phase/frequency modulatable source of some sort.
- A means of amplitude-modulating the final amplifier.

When I spotted the comment about possible inclusion of non-linear modes (such as SSB) in the QMX manual two questions crossed my mind:

- Is the bandwidth of the drain-voltage modulator of the PA sufficient to permit modulation at audio frequencies?? I thought this noting the presence of C506, the 1uF capacitor.
- Is it possible to phase/frequency modulate the Si5351 at a rate sufficient to permit the production of the spectral components of speech?? I know that there are limitations related to PLL tracking in the '5351 but if one can work around those (e.g. phase adjustment, dividers) this might be doable.

If both of the above are true, it's possible to do this, given sufficient processor power (and it probably doesn't take terribly much:? If Hans has pulled off doing Hilbert transforms of 80 taps or so in the receive path, he can do this!)? If if it turns out that speech isn't practical for whatever reason, other modes involving some sort of envelope modulation (certainly PSK31 - which I have done on a lowly PIC by modulating frequency/phase and amplitude separately) certainly are.

73,
Clint, KA7OEI