Topics

QCX-SSB: SSB with your QCX transceiver


Jerry Gaffke
 

Steve,

I have no idea what's now broken.
However, I doubt you had proper PWM going on there,
either on a stock rig, or with the integrating cap at C31 removed.

Once things are working again, you might monitor the drain voltage to the BS170's.
See if that voltage follows the modulation waveform you are trying to give it.

Jerry, KE7ER


On Fri, Feb 1, 2019 at 10:32 AM, Steven Weber wrote:

Okay, played around with this some more this morning.

 

The frequency output definitely follows at single tone. And repeats above 3 Khz, then above 6 kHz, etc, so it could use an anti-aliasing filter. Some frequencies sound a little squirrely too.

 

One issue I noticed which needs to be addressed is the PWM stays active after coming out of transmit.  I’m also not getting anywhere near 5W out, it peaks out at about 1.5 watts.

 

So, I put the original QCX chip back in to see what kind of power it should put out and now there is no power out at all. It doesn’t seem to be switching to transmit when I key. I hear side tone and the display shows what I’m keying, but the BS170 aren’t even getting a signal. The receiver is deafer then a paper mill worker  (I live near a former big time paper producing town).

 

I guess I have other issues to address first. Guess it’s time to  “RTFM”.


Steven Weber
 

After I “RTFM” I think the rig is in practice mode. This is a board loaned to me, so didn’t know how it was set up if it was fully calibrated or not.

 

Just now figuring out the menu system and how the buttons work.

 

Steve,

I have no idea what's now broken.
However, I doubt you had proper PWM going on there,
either on a stock rig, or with the integrating cap at C31 removed.

Once things are working again, you might monitor the drain voltage to the BS170's.
See if that voltage follows the modulation waveform you are trying to give it.

Jerry, KE7ER


On Fri, Feb 1, 2019 at 10:32 AM, Steven Weber wrote:

Okay, played around with this some more this morning.

 

The frequency output definitely follows at single tone. And repeats above 3 Khz, then above 6 kHz, etc, so it could use an anti-aliasing filter. Some frequencies sound a little squirrely too.

 

One issue I noticed which needs to be addressed is the PWM stays active after coming out of transmit.  I’m also not getting anywhere near 5W out, it peaks out at about 1.5 watts.

 

So, I put the original QCX chip back in to see what kind of power it should put out and now there is no power out at all. It doesn’t seem to be switching to transmit when I key. I hear side tone and the display shows what I’m keying, but the BS170 aren’t even getting a signal. The receiver is deafer then a paper mill worker  (I live near a former big time paper producing town).

 

I guess I have other issues to address first. Guess it’s time to  “RTFM”.

 


Steven Weber
 

Yep, it was in practice mode, so no output. Power out is still just 2.5 watts in SK mode, so the output filter isn’t quite right, minor issue.

 

Also had the antenna plugged into the power meter port of my test box instead of the signal generator port. No wonder it sounded deaf!

 

Just the usual operator errors 😊

 

KD1JV

 

After I “RTFM” I think the rig is in practice mode. This is a board loaned to me, so didn’t know how it was set up if it was fully calibrated or not.

 

Just now figuring out the menu system and how the buttons work.

 

Steve,

I have no idea what's now broken.
However, I doubt you had proper PWM going on there,
either on a stock rig, or with the integrating cap at C31 removed.

Once things are working again, you might monitor the drain voltage to the BS170's.
See if that voltage follows the modulation waveform you are trying to give it.

Jerry, KE7ER


On Fri, Feb 1, 2019 at 10:32 AM, Steven Weber wrote:

Okay, played around with this some more this morning.

 

The frequency output definitely follows at single tone. And repeats above 3 Khz, then above 6 kHz, etc, so it could use an anti-aliasing filter. Some frequencies sound a little squirrely too.

 

One issue I noticed which needs to be addressed is the PWM stays active after coming out of transmit.  I’m also not getting anywhere near 5W out, it peaks out at about 1.5 watts.

 

So, I put the original QCX chip back in to see what kind of power it should put out and now there is no power out at all. It doesn’t seem to be switching to transmit when I key. I hear side tone and the display shows what I’m keying, but the BS170 aren’t even getting a signal. The receiver is deafer then a paper mill worker  (I live near a former big time paper producing town).

 

I guess I have other issues to address first. Guess it’s time to  “RTFM”.

 

 


Jerry Gaffke
 

I think I'm slowly coming to terms with how EER works.
Some might find the following discussion helpful.

Consider a standard USB transmitter operating at 14.2mhz,
with two constant audio tones of equal amplitude at 2khz and 3khz.
The result can be thought of as two RF carriers, one at 14.202mhz, another at 14.203mhz.
The transmitter output gives the sum of those two carriers, which is also a sine wave
but varies at an audio rate in amplitude.
Since they are very close in frequency, they will sometimes be in phase
(and thus the output signal will double in amplitude), and sometimes be out of phase
(and there will be no transmitter output at all).
Those variations in amplitude will occur at a 3khz-2khz = 1khz rate. 

The trig identity for the sum of two sines (or cosines) lets us know that the sum
of two sine waves of equal amplitude is one sine wave at a frequency halfway between the two,
but modulated by an audio frequency determined by the difference in the original frequencies.
That audio frequency modulation is the variation in amplitude due to phase differences discussed above.
Adding two sine waves of equal amplitude, we get this result:
    sin(a) + sin(b) = 2 * sin((a+b)/2) * cos((a-b)/2)

In the general case, an SSB transmitter will have a continuous spectrum of audio frequencies
between perhaps 300hz and 3khz going into the mike, each component of that spectrum
is of arbitrary phase, and varies in amplitude with time.
But the result out the transmitter can still be thought of as a single sine wave
that varies slightly in frequency, and also varies in amplitude due to both changes in 
the incoming audio levels, and changes in the phase relationships between the
various components.

So I think I've mostly convinced myself that EER techniques can exactly reproduce the output
of a conventional SSB transmitter.  And that the variations in frequency and amplitude
will be at audio rates, and thus could be manageable with a small processor.
I'd guess that the relatively infrequent step-wise jumps in si5351 output frequency
will prove to be the primary compromise in getting SSB out of an si5351.
And I doubt the arrangement of Q6 on the QCX is adequate for the needed
PWM variations in drain voltage.

We can ignore the required changes in frequency, and just modulate the voltage
into the final.  That gives us an AM phone signal, which can be clearly received
by an SSB receiver.   We have some extra artifacts though, namely the carrier
and the opposite sideband.

We can apparently choose to ignore changes in amplitude and concentrate
instead on steering the frequency about.  Guido, Allison, and Steve all report that this 
is sufficient to produce an intelligible signal for an SSB receiver.
Exactly what the extra artifacts will be, I'm not quite sure.

Very curious stuff!

Jerry, KE7ER

 


On Wed, Jan 30, 2019 at 01:39 PM, ajparent1/KB1GMX wrote:
Q6 depending on the mod it may be an issue.  Doing EER means getting the
amplitude phase correct with frequency change or you get all manner of artifacts.


Alvey Street
 

I remember an article written in 'The Bulletin' many years ago, when transistors were 'poor' devices.  The article showed that an SSB signal can be split into an amplitude component, and a frequency component.  The author limited the SSB signal producing a complex square wave.  This signal was amplified by a class C stage at high efficiency, and restored to a 'normal' SSB signal by applying amplitude modulation to the class C stage.  This modulation was obtained by diode-detecting the original SSB signal.  The author reported good results and very high efficiency.  I came accross a commercial use of this principle some years later.

Regards Vin g4ksy


Jerry Gaffke
 

Excellent!
This would give a highly efficient SSB transmitter, curious the technique has not been used more.

On the positive side:  No need to steer the si5351 frequency around at audio rates if that proves difficult to do accurately enough.
Also, no computations required.

On the negative side:  Still need to build the SSB exciter, the EER method gets around that.

Jerry


On Fri, Feb 1, 2019 at 01:04 PM, Alvey Street wrote:
I remember an article written in 'The Bulletin' many years ago, when transistors were 'poor' devices.  The article showed that an SSB signal can be split into an amplitude component, and a frequency component.  The author limited the SSB signal producing a complex square wave.  This signal was amplified by a class C stage at high efficiency, and restored to a 'normal' SSB signal by applying amplitude modulation to the class C stage.  This modulation was obtained by diode-detecting the original SSB signal.  The author reported good results and very high efficiency.  I came accross a commercial use of this principle some years later.


Jerry Gaffke
 

If updating the si5351 through the i2c interface is not fast enough to create smooth transitions,
we could instead jerk around the 27mhz reference oscillator:
    https://groups.io/g/BITX20/message/34020
Using the varactor diode to pull around the crystal could possibly be made linear enough
through a lookup table inside the uC.

Or, if we're willing to spend big bucks, we could choose the si5351B
in the QFN20 at $2, Mouser pnum 634-SI5351B-B-GM.
The si5351B provides a "VCXO" input, we feed it a voltage between 
0 and 1.65v, it can pull the internal VCO around by up to 240ppm.
That 240ppm is not quite enough on 80m and 40m, but should be
sufficient on 20m:    14.0mhz * 240/1000000 = 3360hz.

Either way, we tune the rig as usual with writes to the i2c port of the si5351
and then leave the i2c bus parked while transmiiting SSB audio.
The si5351 is fine tuned through a control voltage from a PWM or DAC channel from the uC
when transmitting.

That, plus a buck mode switcher that can be programmed for an output voltage from zero to Vcc
using a second PWM or DAC channel from the uC, and we have a complete EER SSB transmitter.

Datasheet for the si5351B says that the "VCXO Modulation Bandwidth" is 10khz, so should work fine.
The si5351B has the VCXO reference oscillator, plus the usual crystal oscillator controlled VCO,
so could use that same part to simultaneously create other clocks that are not modulated.

The QFN20  package has multiple Vdd pads, a large ground pad, and can drive outputs differentially,
all of which may help reduce crosstalk between channels.
So might be a better choice than the MSOP10, when using multiple clocks
even if we don't need the VCXO.

Jerry, KE7ER



On Fri, Feb 1, 2019 at 12:25 PM, Jerry Gaffke wrote:
I'd guess that the relatively infrequent step-wise jumps in si5351 output frequency
will prove to be the primary compromise in getting SSB out of an si5351.
And I doubt the arrangement of Q6 on the QCX is adequate for the needed
PWM variations in drain voltage.


Jerry Gaffke
 

Guido's method has been labeled as "EER" in this forum, an acronym 
for "Envelope Elimination and Restoration".

Seems the EER acronym applies best to the scheme that Alvey just described,
where we first create an SSB signal using a conventional SSB exciter, then send
just the frequency information into a class C, D, or E amp, then restore the
envelope by modulating that amp to follow what we see coming out of the exciter.

We can avoid the need for the SSB exciter by instead creating the frequency and
envelope information through calculations based on values from an ADC that
monitors the incoming audio. 

Jerry, KE7ER




On Fri, Feb 1, 2019 at 01:04 PM, Alvey Street wrote:
I remember an article written in 'The Bulletin' many years ago, when transistors were 'poor' devices.  The article showed that an SSB signal can be split into an amplitude component, and a frequency component.  The author limited the SSB signal producing a complex square wave.  This signal was amplified by a class C stage at high efficiency, and restored to a 'normal' SSB signal by applying amplitude modulation to the class C stage.  This modulation was obtained by diode-detecting the original SSB signal.  The author reported good results and very high efficiency.  I came accross a commercial use of this principle some years later.

Regards Vin g4ksy


Jerry Gaffke
 

One could add the complication of a mixer such as an SA612 to down convert
that 20m square wave coming out of the VCXO Si5351B, bring it down to 40m or 80m. 
Then amplify class E, modulating that amp as before. 
LO for the SA612 could come from one of the other Si5351B clocks.
Total additional cost is the $1 for an SA612.

Jerry


On Fri, Feb 1, 2019 at 04:03 PM, Jerry Gaffke wrote:
That 240ppm is not quite enough on 80m and 40m, but should be
sufficient on 20m:    14.0mhz * 240/1000000 = 3360hz.


Steven Weber
 

Jerry,

 

I think you hit on a good solution. It would be an interesting project to pursue.

Wish I was better at coding in C. The hardware is straight forward. Might want to upgrade the processor too.

 

73, Steve KD1JV

 

If updating the si5351 through the i2c interface is not fast enough to create smooth transitions,
we could instead jerk around the 27mhz reference oscillator:
    https://groups.io/g/BITX20/message/34020
Using the varactor diode to pull around the crystal could possibly be made linear enough
through a lookup table inside the uC.

Or, if we're willing to spend big bucks, we could choose the si5351B
in the QFN20 at $2, Mouser pnum 634-SI5351B-B-GM.
The si5351B provides a "VCXO" input, we feed it a voltage between 
0 and 1.65v, it can pull the internal VCO around by up to 240ppm.
That 240ppm is not quite enough on 80m and 40m, but should be
sufficient on 20m:    14.0mhz * 240/1000000 = 3360hz.

Either way, we tune the rig as usual with writes to the i2c port of the si5351
and then leave the i2c bus parked while transmiiting SSB audio.
The si5351 is fine tuned through a control voltage from a PWM or DAC channel from the uC
when transmitting.

That, plus a buck mode switcher that can be programmed for an output voltage from zero to Vcc
using a second PWM or DAC channel from the uC, and we have a complete EER SSB transmitter.

Datasheet for the si5351B says that the "VCXO Modulation Bandwidth" is 10khz, so should work fine.
The si5351B has the VCXO reference oscillator, plus the usual crystal oscillator controlled VCO,
so could use that same part to simultaneously create other clocks that are not modulated.

The QFN20  package has multiple Vdd pads, a large ground pad, and can drive outputs differentially,
all of which may help reduce crosstalk between channels.
So might be a better choice than the MSOP10, when using multiple clocks
even if we don't need the VCXO.

Jerry, KE7ER



On Fri, Feb 1, 2019 at 12:25 PM, Jerry Gaffke wrote:

I'd guess that the relatively infrequent step-wise jumps in si5351 output frequency
will prove to be the primary compromise in getting SSB out of an si5351.
And I doubt the arrangement of Q6 on the QCX is adequate for the needed
PWM variations in drain voltage.

 


Steven Weber
 

 

 

Alright, my final conclusion on this is that it almost works.

 

The issue is the PWM envelope reconstruction. For one, getting the integration time just right is tricky. Second, there isn’t enough dynamic range.

 

The PWM has a rep rate of 25 uS (about 40 kHz) but has an ON pulse 3 uS long with no audio output. Assuming 256 step, the on time is 30 out 256 steps. That results in a strong carrier and really cuts into the dynamic range. Power output flat tops really quickly.

 

A possible solution is to run the PWM into a low pass filter, then amplify it to the level needed to modulate the PA. The offset could be nulled out. Or that long on time could be reduced to something more reasonable, like 0.1 uS

 

Time to study the program…

 

73, Steve KD1JV


GM4CID
 

Some may remember the G2DAF linear which successfully used this technique. And in point to point HF circuits Lincompex (linked compressor and expander) was used starting in the mid 1960's where the voice signal was quite heavily compressed and the amplitude sent as a frequency modulated carrier which was demodulated at the receiving end and used to to reintroduce the amplitude variations. This increased average TX power and gave a much improved receive s/n .

73 Bob GM4CID


Steven Weber
 

 

More progress.

 

I found where the PWM is set up in the Sketch and changed the settings to give me the full dynamic range. I also found out I was biasing the A/D port wrong. I didn’t realize he only switches to the 1.1V reference on transmit, so I thought it was always at 5V for some reason. So I had the bias resistors powered by 5V instead of the Vref pin. One should really follow the directions 😊

 

With R41 = 10K, R42 = 3.3K and C41(?) 0.01 ufd (10 nF) I now have no residual carrier and decent looking modulation which sounds ok on the monitor receiver. I can still easily overmodulate and flat top, have to hold the mic like 8” away and talk softly. I think he has a way to adjust that in firmware.

 

I’m still going to try putting the PWM signal through a low pass filter, scale it with an amplifier to drive an emitter follower to modulate the carrier. I think that will make a cleaner modulation envelope and maybe make this sound really decent.

 

So, am I the only one to have tried this mod?

 

All in all, I think this is pretty amazing.

 

73, Steve KD1JV

 

 


ajparent1/KB1GMX
 

Hi Steve,

I've done some playing with the technique and it has potential and issues.
One being if the Envelope restoration is not quite in phase with the frequency 
it creates sidebands so filtering is a "done very carefully" thing for the envelope.
IF the filtering is not adequate it will show as sidebands (at the PWM rate).
Also the envelope in the software is only 64 discrete steps so the normal audio
should sound somewhat compressed.  A better PWM or D/A that updates
faster with more bits allowing for finer filtering with smoother amplitude steps
should work better.

The next area a is the frequency increment and the dynamics of that.  Again using
an 8 bitter introduces dynamic range limitations and math limitations for any filters
for frequency information extraction.  There are a host of possible issues with the
updates to the 5351 and what it may or might do.

I want to see what it looks like on a SA for both close in (IMD) and wide for
artifacts and spurs.

The class E is not a requirement and at most a feature for efficiency.  It is
the relationship of DC in to Power out that is more important.  At least for
the scheme to be clean.  The other is the phase relationship of the frequency
component to the power component.

All the research (white papers) and all make it out  to be the next sliced bread.
It has been that way for at least 3 decades with "more cpu" should help.  
To date any device using this has had a very short market life when it
made it that far.  The one I'm most familiar was the 500W mobile amp (EER)
that SGC put out then withdrew not long after.  In the mean time other mainline
SDR-DSP techniques have advanced and are now in mainstream products
some for decades.    So that makes me ask what was wrong?

Allison


ajparent1/KB1GMX
 

On Sat, Feb 2, 2019 at 09:42 AM, Steven Weber wrote:
So, am I the only one to have tried this mod?

No, I'm waiting for a QCX.    I've taken a raw 328P/5351/CODE and a voltage controlled
attenuator to bench board it.  The VCA allows me to drive it with the PWM and is fairly good
as the VCA has a very fast response, meaning low phase error but then it requires a
decent PWM filter.  List of things to try is a fast 8 bit D/A as that should help with
reducing compression.

The result is 10mW of rf peak. and it makes it across the room on a clip lead.
Looks very compressed on the SA and scope.  That in itself is an audio quality issue.
The other is if the filtering of the PWM is not right you see that as a sideband and
a form of intermod (IMD products).  Same for the update rate and increment
rates for the 5351.  Still early in the measurement realm.

Allison


Skip Davis
 

You might be Steve or you are the only one writing about it. I’m waiting for my QCX kit to arrive so that I can also try it out. In fact I order one just to try this.

Skip Davis, NC9O


Steven Weber
 

 

I made a quick dead bug LPF and scaling amp and it has promise.  The audio is starting to sound acceptable. If someone does decide to talk to you and you ask how’s it sound? I’m sure the comment won’t be all that complementary, it sure isn’t HI-FI. I wouldn’t want to put a linear on the PA and run it at 100 watts, but at 5 watts or so, I would think it’s of acceptable quality. For something of this level of simplicity, you can’t expect a whole lot. The fact it works at all is remarkable.

 

73, Steve KD1JV

 

 

I’m still going to try putting the PWM signal through a low pass filter, scale it with an amplifier to drive an emitter follower to modulate the carrier. I think that will make a cleaner modulation envelope and maybe make this sound really decent.


Jerry Gaffke
 

I'm very curious how well the i2c writes into the si5351 do.
Looking forward to your notes on how clean you can get the signal.

The fact that there has been very little commercial success suggests this may
require something better.

The si5351b with VCXO pin is quite cheap at $2, and there are cheap breakout boards:
    https://www.banggood.com/CJMCU-5351B-Si5351B-Clock-Signal-Generator-Module-I2C-Programmable-27MHz-VCXO-p-1296422.html
Not too bad at $17.56, especially considering that you get ten end-launch SMA's in the deal.

PJRC has a 16 bit 44.1khz stereo-in/stereo-out board at $14.25:
    https://www.pjrc.com/store/teensy3_audio.html
for use with their Teensy ARM processor boards.
The core part, an SGTL5000, is $3 for one up on Mouser, $2 if you buy a bunch.
Some of the STM32F parts have dual 12 bit DAC's.

Would be nice to find a good programmable buck mode switcher breakout board,
output voltage under the control of a DAC from the PJRC  audio board.
Failing that, perhaps hack in an op-amp to one of the fixed buck mode switcher breakout boards.
But for initial experiments at 0dBm or so, most any voltage follower after the DAC filter should be fine. 

My suggesting a $1 mixer to get around the 240ppm limitation of the Si5351B VCXO input
was being a bit glib.  There's also the small matter of adding $20 worth of band specific filters
after the mixer and a way to select the appropriate filter. 
 
However, the Si5351B datasheet states it's +/- 240ppm, so we should also be able to cover 40m.
    7mhz * 2*240/1e6 = 3360 Hz.
 
Moving the 27mhz reference oscillator around instead will likely allow coverage on 80m and 160m as well.
Perhaps supply 27mhz from a cheap DDS, the Si5351's PLL loop filter should take care of the spurs?
Is there a cheaper way that is accurate enough?


 
I doubt a varactor diode in with the 27mhz crystal will give as much range as the Si5351B VCXO pin.
 
Fig 3 in the Si5351 datasheet shows the VCXO to be independent of PLLA and PLLB,
but I believe that is misleading.  Elsewhere, it's clear that PLLB is what gets controlled
by the VCXO pin.  AN619 on page 10 states "PLLB must be used as the source for any
VCXO output clock. Feedback B Multisynth divider ratio must be set such that the denominator,
c, in the fractional divider a + b/c is fixed to 106." 
From that I gather that the VCXO pin is somehow digitally adjusting
the feedback multisynth divider for PLLB.

The Si5351 specs an "Output Frequency Transition Time" of 10us worst case.
I assume that is for updates to the PLL Multisynth (and thus changes to the VCXO pin?).
Worst case would be skewing from a 600mhz VCO to 900mhz?
Since this is a measure of how fast their low pass filter in the VCO's PLL can react,
I suspect the response time for changes to the 27mhz reference should also be plenty fast.

I think the Si5351B VCXO pin shows the most promise for accurately scooting the frequency about.
Having both the Si5351B and the modulator controlled through twin DAC's and LPF"s
should help to keep them in sync.
 
I'm about to order the PJRC audio board and Si5351B breakout board.
Already have a Teensy.
All of those strike me as generally useful, even if this EER stuff does not pan out.

But don't hold your breath.
I have a long history of working out how to do something interesting,
then getting distracted by something else once it comes down to the grunt work.
Occasionally I do follow through.

This afternoon I have to finish up an attic floor and access hatch,
get hay out to the creatures, move more firewood close to the door,
clear snow from the driveway, get Liz set up to where she can start potting
garden plants for spring, recover from that beer last night,  ....

Jerry, KE7ER


Steven Weber
 

Life’s lttile tasks do have a way of getting in the way of the fun stuff…

Steve KD1JV

 


This afternoon I have to finish up an attic floor and access hatch,
get hay out to the creatures, move more firewood close to the door,
clear snow from the driveway, get Liz set up to where she can start potting
garden plants for spring, recover from that beer last night,  ....

Jerry, KE7ER

_._,_._,_

 


ajparent1/KB1GMX
 

I'm very curious how well the i2c writes into the si5351 do.
Looking forward to your notes on how clean you can get the signal.<<

Jerry,

All you need to test the signal generation is a raduino.  There are a few things to be aaware of as its clocked at 25mhz
and the there is no amp after it but the code after matching the LCD lines runs.  Of  Encoder, An Uno or nano and a 5351 board.

Some of you speculation is about how the si5351 works and much is not given away.  The PLL VCO is controlled
by the PLL divider or multisynth and the VCO tracks that.  The VCO based on my PLL work should settle fast but
you have some long counter chains and that is a factor.  Then you have the overriding thing that moving 8bits via SPI
even at 800khz takes about 11us.  That last item is the fastest you can change any parameter of the 5351, 1 byte 11us.

Putting a varactor on the crystal would create FM, you don't want that.  You would have to process the mic input to
a D/A output and then your guessing on frequency as its in the analog realm.  then the output of the 5351 will multiply
that (crystal to VCO then divide to output frequency).  On a good day that would be band limited.

The use of a mixer, really?  Mix the 5351 and I presume audio, then a crystal filter as you have both sidebands.
Again that is a ground up bog standard filter SSB.  The EER part could be added but if you get the phasing
wrong the sideband spurs will give a IMD of terrible.

Allison