At the outset - just to be clear - you know I love you, and I am not interested in any kind of emotive argument about anything, I am only very interested on the technical aspects, ONLY.
lot of interesting info here
I disagree, as you can clearly sense their agenda. You don?t even need to read between the lines as both make their personal issues with us/me relatively clear.
We decided against "jumping through their hoops", when those hitpieces appeared, as there?s nothing to gain. Best way to deal with trolls: Don?t feed them.
Sorry I didn't know there were personal issues between you and Pascal VK2IHL (who I assume is who you mean by "them"). And I don't know what you mean by hoops. Maybe I misread the VK2IHL page but I only saw someone who is clearly technically competent, and who made a lot of experiments on uSDX, and I thought uSDX and (tr)uSDX are supposed to be intended for experimentation, so it all looked OK for me. Perhaps I missed the things you mention.
?
But briefly: It is easy to provoke this IMD bahavior when you overdrive the input. So my advise would be: Don?t overdrive the input.
I think?IMD3 is a very good measure of a transmitter's SSB capability. Have you got an IMD measurement for (tr)uSDX? It appeared to me that there was very little about this on the internet other than the VK2IHL pages and the other link I included? .?
??
On the other side, I?m still missing ANY proof, that drain modulation is superior to gate modulation, as there was never a practial implementation.
That, I am unsure?of. Maybe I misunderstood you. But the ORIGINAL uSDX implementation was a modification of the original QRP Labs QCX as everyone knows. I believe the original uSDX used drain modulation. PWM was applied via the MPS751 envelope shaping in QCX. Later, it was realized that several more components could be eliminated, if gate bias modulation was used. However the impact on performance is quite significant.?
I am not sure what you would consider as a valid proof but on the?VK2IHL page, if you search for 'FROM THE POLES" it will take you far down the page (which is very long) to a uSDX implementation by some Polish radio amateurs; they did indeed retain a PNP transistor and PWM Drain modulation. VK2IHL had measured with this modification, an IMD3 of -20dBm, and with the gate modulation, -5dBm. It appears to me that there is therefore both a practical implementation, and that the drain modulation example is superior to gate modulation.?
?
I hope, that your SSB Firmware for QMX can finally change that. And I?m really looking forward to that, to be honest !
Until then it?s just a theoretical unproven claim.
See above, there appears to be reliable experimental evidence. It's also very significant, I think, and important, that the measurement was made in the same lab on the same equipment by the same person (VK2IHL) since very often, there can be severe changes in experimental method or equipment from one person to another. From what I can see, VK2IHL appears to be competent and have reliable methods and equipment, and made the measurement on both modulation styles equivalently.?
?
?And we already had personal discussions at length about your initial QMX Claims in your FDIM Paper, if you remember......
Certainly! And I will recap here for the sake of others... in the FDIM paper I said that I was optimistic about the future capability of QMX to implement SSB since uSDX had already demonstrated the concept, and that QMX had a greatly superior hardware. Specifically I mentioned that the amplitude modulation design in QMX uses a very linear control characteristic which has a measured 37dB of amplitude control range. Meaning that the transfer characteristic from the 12-bit STM32 DAC output pin to the RF output Vpp of the amplifier is a very linear relationship with 37dB control range.?
Comparatively I had (since it was my first thought, to try, during the QMX design) measured the uSDX style with gate modulation. I measured 20dB of control range and a very poor linearity.?
So what we discussed, was this "claim", which did not sound too good. But again please understand, I have no personal axe to grind, I have tremendous respect for what Guido PE1NNZ has achieved in the firmware and your own huge contribution to the project. I am only interested in the technical aspects.?
uSDX uses a 10-bit modulation which goes into an RC filter and this then is applied as gate bias to the MOSFETs. There's a rule in DSP that each 1 bit is worth 6dB of dynamic range. Does this mean uSDX has 60dB of control range? NO. It means the maximum theoretical dynamic range would be 60dB. This is a theoretical maximum boundary limit which cannot be exceeded. But the practical realized dynamic range can be less, even much less, depending on the circuit characteristics.?
In this case, I found - by MEASUREMENT - that once the DC gate bias voltage goes beyond 1.25V, there is no further increase in RF output of the PA. Therefore the 2 most significant bits of the PWM are already rendered obsolete, since whether they are 0 or 1, makes no difference to the output power. 8 bits remain. Which still doesn't mean we have 48dB of dynamic range. Because now we have to examine the opposite boundary, the zero end of the scale.?
So at zero DC bias what happens? There's still about 4 to 5V Vpp output of the PA. Why? This is also possible to understand by considering what happens with the gate voltage. The BS170 datasheet states the typical gate threshold voltage is 2.1V. However, you are driving it via a coupling capacitor with a 5V squarewave. When you apply 0V DC bias at the gate, the squarewave hitting the gate goes from -2.5V to?+2.5V. The?+2.5V side is partially switching on the BS170s, it is higher than the threshold voltage. This is why there is still 5Vpp RF output of the PA even when the DC bias is zero. So this determines the lower boundary.?
This is why the control range is 20dB. Because at maximum 1.25V DC bias, the RF output is say 5W (45Vpp); and at minimum 0V DC bias of the control range, RF output is 4-5Vpp; and this is a range of 20dB. 20dB comes from measurement of the practical implementation of the circuit. It cannot be assumed to be 60dB based on a theoretical 10-bit PWM and 6dB per bit.?
In between the 0V and 1.2V ends of the DC bias control range, the transfer characteristic is of course heavily non-linear, the classic fallen-over S-shape we expect from a MOSFET. Which, I do understand, you can somewhat correct for in software with a pre-distortion lookup table, as you do, I believe, in uSDX.?
So this is the basis of my "QMX Claims" in my FDIM paper... it was all based on hours of lab time, blood sweat and tears. And burns, too, of course.?
Similarly I could not claim, on QMX, that I had 72dB of amplitude control because I use a 12-bit DAC peripheral of the STM32F446. I do not use the entire upper limit of the DAC's range because I have to consider that a user may be using different supply voltages, and cater for that variation. And at the bottom end of the control range, I also do not get zero RF output. So again, measurement of the practically realized range is required an in this case, I measured 37dB control range.?
The preference for drain modulation also makes sense from a thought-experiment or you might say, theoretical perspective. When using drain modulation, to a large extent the Class-E nature (or switching Class-D in QMX' case) of the amplification is mostly independent of the supply voltage. The transistor continues to act as a saturated switch, regardless. Therefore you have a lot of control over the operating state of the transistor. This is not the case when you use gate modulation, because now you put the device into the non-saturated region of its characteristics. Now you have a Class-C amplifier and a lot more dependence on device characteristics and their variation, as well as the non-linear behaviour. It all got a lot more complicated and that means I think, that it also got a lot more difficult to control the behaviour to obtain good performance.
When designing QMX my first thought, for the amplitude modulation aspect of the PA, was to try to apply a similar scheme of gate bias modulation as used in uSDX. I have huge admiration for the uSDX achievements in the firmware and the concepts which it has popularized among radio amateurs. It certainly opened my eyes. This is WHY I measured the gate modulation performance.?
But, the results aren't satisfactory. I could not accept the performance. This is why I searched for a better way and found it in my own repertoire, in the form of the drain modulation circuit used in the Ultimate3S 5W PA kit - that PA modulation circuit is interesting to read about?(starting at page 9); and it was the work of Alan?G8LCO. It has a good amplitude control linearity and wider range.?
So this is the background to the FDIM QMX comment. And as a result of my?experiments, I concluded that
1) The uSDX could have been somewhat improved merely by a different choice of resistors in the PWM output filter, that would have meant 10-bits full scale covered only the range 0 to 1.25V instead of 0 to 5V; though most probably this increase in usable resolution from 8 to 10 bits makes little practical impact on performance.?
2) The uSDX could have been significantly improved by the use of drain modulation rather than gate modulation, as it provides the opportunity (even without more complex modulation circuits such as QMX has) to provide a wider control range and with better linearity.?
3) For QMX to?obtain a good performance, I should use a CPU with a good DAC (as I did not want to have PWM sidebands), so I chose STM32F446 (based on that and of course a whole long list of other requirements) and I should use drain modulation based on my 5W PA kit circuit.?
73 Hans G0UPL
?
Steve G4EDG
Manuel; DL2MAN
va3rr
Bruce Akhurst