开云体育

Hi John

In the attached "before" and "after" scope screenshots the top trace (blue) is RF output across a dummy load, the bottom trace (yellow) is a BS170 Drain voltage. "After" has the 1N4148 commutating diode across the FT37-43 10-turn choke. NOTHING else changed.?

Attached are the before and after "spike" screenshots. Measurements of power, current, efficiency, harmonics:

? ? ? ? ? ? ? ? ? ? No diode? ? ? ? ? 1N4148?
? ? ? ? ? ? ? ? ? ? ("before")? ? ? ?("after")

? ? ? ? ? ? ? ? ? ? ----------? ? ? ?---------

Supply voltage:? ? ?11.94 V? ? ? ? ?11.94 V

PA current:? ? ? ? ? ?596 mA? ? ? ? ? 618 mA

RF output:? ? ? ? ? ?4.80 W? ? ? ? ? 4.49 W

PA Efficiency:? ? ? ? ?67 %? ? ? ? ? ? 61 %

2nd harmonic:? ? ? ?-59.0 dBc? ? ? ?-57.0 dBc

3rd harmonic:? ? ? ?-70.0 dBc? ? ? ?-70.0 dBc

4th harmonic:? ? ? ?below measurement noise floor (-90dBc)

5th harmonic:? ? ? ?-81.6 dBc? ? ? ?-80.8 dBc

The addition of the 1N4148 diode had the following relative effects:

• Substantially attenuated the spike, though not completely, it still rises about 10V above the drain peaks but stays below the 60V Vds rating
• Reduced RF output power from 4.80 W to 4.49 W
• Increase PA current from 596 mA to 618 mA
• Reduced PA efficiency from 67% to 61%
• Increased the level of the 2nd harmonic by 2 dB

I did go back and forth several times to make sure the differences I saw were real, not just whether I was breathing or not; the measurements were the same each time.

So - the diode significantly improves the spike but at the expense of a little power, 2nd harmonic and PA efficiency. I'm still not putting a 1N4148 on mine!

73 Hans G0UPL

On Fri, Sep 15, 2023 at 12:56?PM John Z <jdzbrozek@...> wrote:

Hans,

That is a very good experiment!

I can see though that it will have the QRPp crowd rushing to make the power supply separation mod! :-)

It is clear that avalanche is occurring as you describe it, and the transistors appear to survive. Good! It would be interesting to see the gate drive voltage in that hot interval.

I would argue with you though about any significant power being delivered to the load. The level of the ringing voltage is quite low and the rate of damping is very slow. More than 7 complete cycles are seen in your pics.

In principle the output transformer should prevent energy from being coupled to the load. If the avalanche current is the same on both banks of transistors, the fields produced at the transformer will cancel (common mode operation) and nothing will couple to the secondary.

Imbalances will create a bit of coupled output of course, and some of that is probably happening here.

I would like to ask you to take this a step further!

Repeat the experiment with a commutating diode across L14.?

Then, if you are willing, fire up your spectrum analyzer and review performance with and without the diode. A 12 V only run should be adequate I would think.

If the stress of that spike can be removed from the BS170 array without ill effect, then I think everyone wins.

JZ

On Fri, Sep 15, 2023, 5:21 AM Hans Summers <hans.summers@...> wrote:

Hi JZ, all

I had an idea that it would be interesting to snapshot the drain voltage spike at various different supply voltages. Normally the QDX won't operate properly down below somewhere around 6-7V. But I separated the supply choke and fed it separately with its own power supply, and 100nF capacitor (intentionally small). Then I could take it down to 2V supply for the BS170 PA, and keep the rest of QDX running at 12V.

QDX is a 12V low-bands QDX with RWTST transformer, experiments done at 7MHz using the transmitter test feature in the terminal. PA supply voltage and current are read from the display on the variable bench PSU.?

Supply? ? Output? ?PA? ? ? ?Spike
voltage? ?Power? ? Current? Voltage

-------? ?------? ?-------? -------

1.95V? ? ?170mW? ? 100mA? ? ? 44V

3.46V? ? ?0.49W? ? 175mA? ? ? 71V

4.51V? ? ?0.76W? ? 226mA? ? ? 83V

6.40V? ? ?1.54W? ? 324mA? ? ? 85V

9.05V? ? ?3.03W? ? 455mA? ? ? 84V

9.84V? ? ?3.38W? ? 494mA? ? ? 86V

11.9V? ? ?4.75W? ? 592mA? ? ? 95V

In the attached 'scope traces, the supply voltage is the file name.?

I'm not sure what this tells us. At the very least it tells us that the spike is dissipated in the load, not only in the BS170s... this is shown by the 1.95V supply case where the spike is 44V, not enough to breach the 60V BS170 Vds. Here the spike is dissipated into the load, which then rings a bit at some frequency a little higher than the 7MHz operating frequency - which is also quite believable in a 40m LPF.?

The 3.46V supply case shows 71V spike and doesn't look a different shape to the 1.95V case.?

At 4.51V the spike is 83V and it looks like a sudden drop from there down to maybe 80V or so and thereafter it looks like the lower voltage cases with the dissipation into the ringing?of the load.?

At 6.4V and above, we see a trend - the peak spike voltage appears to go up very slowly from here, 85V at 6.4V supply to 95V at 11.9V supply. It appears to get quickly limited by one or more spike events, down to around 80-85V, and then it falls down to zero dissipating energy into the load ringing.?

I think one plausible set of interpretations of these observations could be:

• Avalanche breakdown is occuring at somewhere around 80-85V.?
• The transistor doesn't remain in the avalanche condition while the spike dissipates but there appears to be a hysteresis?
• The avalanche condition tends to act to limit the spike to approximately 80V.?
• The majority of the energy (from 80V down to zero) gets dumped into the load which rings at a frequency of something a bit higher than the operating frequency

If this is the case then it appears that the avalanche breakdown is a normal occurrence at every key-up for any normal operating conditions, since we start to see evidence of it at 6.4V supply. Given the large number of QDX users operating normally for months or years including me, I would suggest that this avalanche breakdown evidently does not, by itself, cause any failure.?

One might suggest that the avalanche breakdown is undesirable generally, I would not argue with that. One might say that it is a contributing cause perhaps making the QDX more delicate to other forms of abuse like high SWR; still there is not any evidence of this so I'd be hesitant to go that far. And I'm hesitant now to use the term "speculation" but...?

There are numerous other dimensions of investigation that could be pursued but all this takes a lot of time, which is in scarce supply and high demand already. For example, using BN61-202 appears qualitatively to reduce the spike; one could also experiment with smaller choke inductances, 10 turns on an FT37-43 is probably a bigger inductance than is necessary even down to 80m band.?

Anyway I hope the observations and measurements provide some further advance in this area.?

73 Hans G0UPL