|
I noticed my high band QDX rev E receive current was double of normal. 320 vs 165 mA normal.
TX power out had also dropped by 1/3.
The case was warmer than normal.
?
The gates of Q8 & Q10 were at .6 volt, should be 0.
?
Removed Q8 & Q10 rx current dropped to slightly below normal (155 mA).
Voltage on the gate connections is back to 0.
I'm assuming the gate puncture issue caught early thanks to my Siglent power supply  This unit has been in operation for a couple of weeks on WSPR 20m thru 10m into a pre tuned inverted "L" antenna and has had no mis matches that I am aware of.
I had ordered extra BS170 from Hans, so I replaced just Q8 and Q10.
The unit is back on 17m putting out full (4.8W) power and drawing the normal 865 mA during tx.
I did note this happened after I was operating WSPR on 12m and 10m, which I hadn't done much of prior, as I was usually on 15m or 20m.
Another question - have most of the PA failures been WSPR ops w/the 2 min key down?
73
Mike N0QBH
|
Mike, That seems to match the theory for L14? inductive voltage spike failure pretty well. I hope you installed a high speed? commutating diode across L14 to prevent recurrence.
JZ KJ4A?
On Wed, Apr 26, 2023, 3:33 PM Mike Berg < mikeberg@...> wrote:
I noticed my high band QDX rev E receive current was double of normal. 320 vs 165 mA normal.
TX power out had also dropped by 1/3.
The case was warmer than normal.
?
The gates of Q8 & Q10 were at .6 volt, should be 0.
?
Removed Q8 & Q10 rx current dropped to slightly below normal (155 mA).
Voltage on the gate connections is back to 0.
I'm assuming the gate puncture issue caught early thanks to my Siglent power supply This unit has been in operation for a couple of weeks on WSPR 20m thru 10m into a pre tuned inverted "L" antenna and has had no mis matches that I am aware of.
I had ordered extra BS170 from Hans, so I replaced just Q8 and Q10.
The unit is back on 17m putting out full (4.8W) power and drawing the normal 865 mA during tx.
I did note this happened after I was operating WSPR on 12m and 10m, which I hadn't done much of prior, as I was usually on 15m or 20m.
Another question - have most of the PA failures been WSPR ops w/the 2 min key down?
73
Mike N0QBH
|
It would seem a possibility according to electronics 101. ;)
I'm planning on running it on 20 - 17m for awhile and observe then test my luck on 12 & 10m to see if it reappears.
As for the mod, I don't have a service monitor w/spectrum analyzer, so I'll wait & see what Hans comes up with.
|
Mike?
Simulation? and theory suggests that lower antenna feed impedance at the TX output port makes things worse. The transistor currents go up? sharply and so does the L14 current and stored energy. At 17.5 ohms antenna load (3:1 SWR) each of the four transistors could be eating over a watt of spike power for about a half microsecond. Worse, the drain-gate voltage goes way above spec.
Put the diode in.?
73 JZ KJ4A.
On Wed, Apr 26, 2023, 4:45 PM Mike Berg < mikeberg@...> wrote: It would seem a possibility according to electronics 101. ;)
I'm planning on running it on 20 - 17m for awhile and observe then test my luck on 12 & 10m to see if it reappears.
As for the mod, I don't have a service monitor w/spectrum analyzer, so I'll wait & see what Hans comes up with.
|
One more thing on this. If you are worried that the commutating diode will introduce bad spectral lines, consider this.
The diode only conducts current for a brief interval after the cessation of transmission. Its current only circulates through L14 and never enters T1 to get to the LPF and antenna. Should any diode current somehow enter T1 it does so through two counter wound paths whose fields cancel so no conveyance to the load. Boogeyman I think, just Iike the idea that the commutating diode will raise the receiver noise floor.?
Just do it.
JZ
The suggestion of?
On Wed, Apr 26, 2023, 5:06 PM John Zbrozek < jdzbrozek@...> wrote: Mike?
Simulation? and theory suggests that lower antenna feed impedance at the TX output port makes things worse. The transistor currents go up? sharply and so does the L14 current and stored energy. At 17.5 ohms antenna load (3:1 SWR) each of the four transistors could be eating over a watt of spike power for about a half microsecond. Worse, the drain-gate voltage goes way above spec.
Put the diode in.?
73 JZ KJ4A.
On Wed, Apr 26, 2023, 4:45 PM Mike Berg < mikeberg@...> wrote: It would seem a possibility according to electronics 101. ;)
I'm planning on running it on 20 - 17m for awhile and observe then test my luck on 12 & 10m to see if it reappears.
As for the mod, I don't have a service monitor w/spectrum analyzer, so I'll wait & see what Hans comes up with.
|
Thanks for your insistent suggestion, John.?
I think I'l look over some of my other radio schematics and look for a similar solution tried else where.
This isn't a relay.
73
Mike N0QBH
|
No, it's not a relay but it still makes quite a kick!?
On Wed, Apr 26, 2023, 5:44 PM Mike Berg < mikeberg@...> wrote: Thanks for your insistent suggestion, John.?
I think I'l look over some of my other radio schematics and look for a similar solution tried else where.
This isn't a relay.
73
Mike N0QBH
|
Mike,?? Apologies for my over-eagerness. It's entirely possible that you might do nothing further on this and still enjoy long and trouble free use of your QDX. There are, after all, thousands out there today, running just fine. I am glad your IC5 survived the stress and didn't need replacement.
You are unlikely to find any RF literature specifically addressing the configuration of this power amplifier. The un-bypassed feed choke L14 is somewhat unique.
?The configuration is better known in the realm of Switch Mode Power Conversion. There it is known as a Current Fed Push Pull Converter. You can find articles about it. "Current fed" relates to the fact that the feed inductor current stays constant through the running time of the circuit. An inductive kick at turn off is acknowledged and the designer is advised to either use avalanche-rated transistors or a commutating diode.
Enough on that. May your QDX live long and serve you well.
JZ
On Wed, Apr 26, 2023, 5:45 PM John Zbrozek < jdzbrozek@...> wrote: No, it's not a relay but it still makes quite a kick!?
On Wed, Apr 26, 2023, 5:44 PM Mike Berg < mikeberg@...> wrote: Thanks for your insistent suggestion, John.?
I think I'l look over some of my other radio schematics and look for a similar solution tried else where.
This isn't a relay.
73
Mike N0QBH
|
Wouldn't a Zener from drain to source of each pair of PA transistors, as discussed in this list before, also take care if possible inductive "kicks"? I've mounted?1N4756A ZENER DIODE, 1W, 47V on both mine as shown here. -- Sverre
LA3ZA
http://la3za.blogspot.com
|
Yes, Sverre, that is a very viable solution and furthermore it protects against wild SWR/load related voltage damage. The only downside, if one exists at all, is the small extra capacitance at the transistor drains.?Probably small potatoes.?
Post your before/after output power results by band if you can!
73, JZ KJ4A?
On Thu, Apr 27, 2023, 3:28 PM Sverre Holm < svholm54@...> wrote: Wouldn't a Zener from drain to source of each pair of PA transistors, as discussed in this list before, also take care if possible inductive "kicks"? I've mounted?1N4756A ZENER DIODE, 1W, 47V on both mine as shown here.
--
Sverre
LA3ZA
|
I did a test of power before/after adding Zener diodes and couldn't detect any change neither in my low-band QDX nor in my high-band QDX. -- Sverre
LA3ZA
http://la3za.blogspot.com
|
Excellent, Sverre. Your solution should be considered seriously by all.
On Thu, Apr 27, 2023, 5:33 PM Sverre Holm < svholm54@...> wrote: I did a test of power before/after adding Zener diodes and couldn't detect any change neither in my low-band QDX nor in my high-band QDX.
--
Sverre
LA3ZA
|
On 27/04/2023 22:11, John Z wrote: Your solution should be considered seriously by all. John, Perhaps some careful testing is required before it can be declared a solution? The theory seems good but as in previous discussions no actual research. Can we be sure that there are no unknown pitfalls? have most of the PA failures been WSPR ops w/the 2 min key down? Mike, The impression I have is no. If you look at previous posts it is difficult to establish the cause(s) of BS170 failure. If a spike causes a BS170 to short it will look like an overcurrent failure. It is also difficult to establish a normal operating temperature for the PA, some say 2 minutes WSPR just makes them warm, others say HOT. Personally I throttle back my QDXs used for WSPR . Mainly because I think 5 Watts is too much for that mode on HF. 73 Alan G4ZFQ
|
Alan, Testing is always good. It's helpful though to have some idea as to what you are looking for. What possibilities have you in mind here?
The Zeners that Sverre has chosen are inactive during all normal transmission conditions. Even an open antenna port does not produce a drain voltage sufficiently high to activate them. They only become active during the high voltage inductive spike at end of transmission.
Simulation has not revealed any negative effects.
Some manufacturers of RF power transistors incorporate Zeners into their die for the very same purpose that Sverre has chosen to use them.
What say, Alan?
JZ
On 27/04/2023 22:11, John Z wrote:
> Your solution should be considered seriously by all.
John,
Perhaps some careful testing is required before it can be declared a
solution?
The theory seems good but as in previous discussions no actual research.
Can we be sure that there are no unknown pitfalls?
?>have most of the PA failures been WSPR ops w/the 2 min key down?
Mike,
The impression I have is no.
If you look at previous posts it is difficult to establish the cause(s)
of BS170 failure. If a spike causes a BS170 to short it will look like
an overcurrent failure.
It is also difficult to establish a normal operating temperature for the
PA, some say 2 minutes WSPR just makes them warm, others say HOT.
Personally I throttle back my QDXs used for WSPR . Mainly because I
think 5 Watts is too much for that mode on HF.
73 Alan G4ZFQ
|
John, do you have a part # for one of those power transistors?
I'd like to look at the data sheet.
I like your basic idea, but I keep getting a nagging thought from my old RF instructor (W0WUG sk) saying something like "you put a non linear device in parallel with the power supply of a transmitter and expected it not to be a mixer?"
Or the "push in poke out theory" from my old manager. No action is without a reaction.
That said, if we always listened to these cautionary types we'd never get anywhere.
73
|
Hi John
I think that you are probably correct that there are no negative effects by adding the 47V zener diodes. Probably meaning, better than 50% probability. I'd say a lot better.
And yet - there is enough to worry me, to not make it an official recommendation. Yet.
Firstly I still don't see any convincing evidence that the spike, which is a fraction of a microsecond in duration, actually causes any failure or increased probability of failure. But fine, it would not be a bad or imprudent thing, to err on the side of caution. So we can roll with that.
But not then to throw all caution to the wind in regard to other potential issues, which aren't things that are or can be revealed by simulation. Specifically I'm referring to noise and harmonics.
Noise means an increase in the noise floor during transmit. I think it's unlikely but I'd not discount the possibility.
Harmonic distortion can arise too. The diode is not a perfect switch. It's a thing with a curve. A very low level of conduction can be enough to create even order harmonics that are measurable and troublesome. Effects due to diode capacitance could also be a factor. The diode sees a varying voltage during the RF cycle. The capacitance is also a function of voltage. That creates the potential for waveform distortion, in other words, it can change the level of harmonic content. In particular, the 2nd harmonic has the potential to cause problems. The low pass filters in QDX are designed in the knowledge that the even order harmonic content is very low. If this assumption becomes invalid or less valid, there's a possibility that the FCC harmonic level requirements may not be met. Again I think this is unlikely. But it isn't a zero possibility and I think it needs to be tested.
If I've learned anything in the last few years of QRP Labbering it's that there are an infinite number of hidden gremlins just waiting in the shadows for the worst moment to jump out of the darkness and bite your posterior. Even low probabilities or probabilities you don't even know exist, are hiding there.
It's all the worse when it's something like in this case, where we are talking about performance characteristics which are important but can't be simulated, and which the vast majority of operators don't have the equipment to measure.
Call me paranoid if you like...
I did order some 47V diodes but they were found to be 9.1V. Supplier error. I complained and eventually got a new set, which I haven't had time to check out yet. I do have the means to measure noise floor and harmonics. But it's probably at least a day's work to do it properly on all bands and the low and high band QDX, and I simply haven't had a chance to make that sized timeslot for it yet.
73 Hans G0UPL
Alan,
Testing is always good. It's helpful though to have some idea as to what you are looking for. What possibilities have you in mind here?
The Zeners that Sverre has chosen are inactive during all normal transmission conditions. Even an open antenna port does not produce a drain voltage sufficiently high to activate them.
They only become active during the high voltage inductive spike at end of transmission.
Simulation has not revealed any negative effects.
Some manufacturers of RF power transistors incorporate Zeners into their die for the very same purpose that Sverre has chosen to use them.
What say, Alan?
JZ
On 27/04/2023 22:11, John Z wrote:
> Your solution should be considered seriously by all.
John,
Perhaps some careful testing is required before it can be declared a
solution?
The theory seems good but as in previous discussions no actual research.
Can we be sure that there are no unknown pitfalls?
?>have most of the PA failures been WSPR ops w/the 2 min key down?
Mike,
The impression I have is no.
If you look at previous posts it is difficult to establish the cause(s)
of BS170 failure. If a spike causes a BS170 to short it will look like
an overcurrent failure.
It is also difficult to establish a normal operating temperature for the
PA, some say 2 minutes WSPR just makes them warm, others say HOT.
Personally I throttle back my QDXs used for WSPR . Mainly because I
think 5 Watts is too much for that mode on HF.
73 Alan G4ZFQ
|
Mike,?
Here is a reference design for an amplifier that uses such transistors:
Reference Design for 13.56 MHz Push-Pull 600 W RF Amplifier ...
??
The QDX transmitter already comes into contact with many non-linear devices. Consider all the PIN diodes in band switching, or the MOSFET t/r switch, or the two diodes in the voltage doubler reverse bias generator connected to the output of T1. Oh my! Those diodes are definitely active during transmit as the reverse bias generator thieves power from the transmitter.
Sverre's two Zeners that are nearly always nonconducting seem pretty benign in comparison
JZ
On Fri, Apr 28, 2023, 4:37 AM Mike Berg < mikeberg@...> wrote: John, do you have a part # for one of those power transistors?
I'd like to look at the data sheet.
I like your basic idea, but I keep getting a nagging thought from my old RF instructor (W0WUG sk) saying something like "you put a non linear device in parallel with the power supply of a transmitter and expected it not to be a mixer?"
Or the "push in poke out theory" from my old manager. No action is without a reaction.
That said, if we always listened to these cautionary types we'd never get anywhere.
73
|
Here is a reference design for an amplifier that uses such transistors:
Reference Design for 13.56 MHz Push-Pull 600 W RF Amplifier ...
??
Nice amp :-D? ? ? So far the largest power I've ever transmitted in my ham life has been about 60W, from the QRP Labs PA kit ;-)? ?? ? The QDX transmitter already comes into contact with many non-linear devices. Consider all the PIN diodes in band switching, or the MOSFET t/r switch, or the two diodes in the voltage doubler reverse bias generator connected to the output of T1. Oh my! Those diodes are definitely active during transmit as the reverse bias generator thieves power from the transmitter.
Sverre's two Zeners that are nearly always nonconducting seem pretty benign in comparison
Yes and as I said, I think you are most probably correct and that?there will be no ill effects. HOWEVER - the important difference is that the QDX in contact with all those other non-linear devices has all been carefully designed and real-world tested with those devices. That's all I'm saying - I think it would be prudent to TEST this in the real world, rather than just ASSUME no ill-effects based on a limited simulation prediction. Testing results would make me comfortable.?
73 Hans G0UPL
|
Hello Hans,
I'm very glad to see that you are gearing up to test a Zener solution. Excellent!
Your point about the non-linear capacitance of the off state diodes is interesting and potentially germaine. I point out though that is no different than the voltage dependant drain capacitance of the BS170's, and quite a bit smaller than the transistor spec maximum capacitance X2. Did you select worst case BS170's for your testing?
Regarding the spike, avalanche related damage is indeed time and energy dependent. Gate oxide puncture is virtually instantaneous given sufficient voltage. Your own testing showed a spike magnitude of about 100 volts. I think that will do it.
The BS170 is spec'd for 60 volts Vdg. The Vdg spec is not seen much on datasheets these days. It has been subsumed by the Vds avalanche spec on the assumption that if you are violating that you are already in trouble.
It's illuminating though to look for a moment at the Vgs max voltage spec for BS170. It is +/- 20 volts as it attempts to protect the gate oxide from damage. It is probably guard-banded by a factor of 3. The Vdg-max spec of 60 volts probably comes closer to the true oxide failure voltage. Note that neither spec says " ...for N or more nanoseconds".
I am very much looking forward to your findings in your testing!
Kind Regards, JZ
Hi John
I think that you are probably correct that there are no negative effects by adding the 47V zener diodes. Probably meaning, better than 50% probability. I'd say a lot better.
And yet - there is enough to worry me, to not make it an official recommendation. Yet.
Firstly I still don't see any convincing evidence that the spike, which is a fraction of a microsecond in duration, actually causes any failure or increased probability of failure. But fine, it would not be a bad or imprudent thing, to err on the side of caution. So we can roll with that.
But not then to throw all caution to the wind in regard to other potential issues, which aren't things that are or can be revealed by simulation. Specifically I'm referring to noise and harmonics.
Noise means an increase in the noise floor during transmit. I think it's unlikely but I'd not discount the possibility.
Harmonic distortion can arise too. The diode is not a perfect switch. It's a thing with a curve. A very low level of conduction can be enough to create even order harmonics that are measurable and troublesome. Effects due to diode capacitance could also be a factor. The diode sees a varying voltage during the RF cycle. The capacitance is also a function of voltage. That creates the potential for waveform distortion, in other words, it can change the level of harmonic content. In particular, the 2nd harmonic has the potential to cause problems. The low pass filters in QDX are designed in the knowledge that the even order harmonic content is very low. If this assumption becomes invalid or less valid, there's a possibility that the FCC harmonic level requirements may not be met. Again I think this is unlikely. But it isn't a zero possibility and I think it needs to be tested.
If I've learned anything in the last few years of QRP Labbering it's that there are an infinite number of hidden gremlins just waiting in the shadows for the worst moment to jump out of the darkness and bite your posterior. Even low probabilities or probabilities you don't even know exist, are hiding there.
It's all the worse when it's something like in this case, where we are talking about performance characteristics which are important but can't be simulated, and which the vast majority of operators don't have the equipment to measure.
Call me paranoid if you like...
I did order some 47V diodes but they were found to be 9.1V. Supplier error. I complained and eventually got a new set, which I haven't had time to check out yet. I do have the means to measure noise floor and harmonics. But it's probably at least a day's work to do it properly on all bands and the low and high band QDX, and I simply haven't had a chance to make that sized timeslot for it yet.
73 Hans G0UPL
On Apr 28, 2023, at 10:44 AM, John Z < jdzbrozek@...> wrote:
Alan,
Testing is always good. It's helpful though to have some idea as to what you are looking for. What possibilities have you in mind here?
The Zeners that Sverre has chosen are inactive during all normal transmission conditions. Even an open antenna port does not produce a drain voltage sufficiently high to activate them.
They only become active during the high voltage inductive spike at end of transmission.
Simulation has not revealed any negative effects.
Some manufacturers of RF power transistors incorporate Zeners into their die for the very same purpose that Sverre has chosen to use them.
What say, Alan?
JZ
On 27/04/2023 22:11, John Z wrote:
> Your solution should be considered seriously by all.
John,
Perhaps some careful testing is required before it can be declared a
solution?
The theory seems good but as in previous discussions no actual research.
Can we be sure that there are no unknown pitfalls?
?>have most of the PA failures been WSPR ops w/the 2 min key down?
Mike,
The impression I have is no.
If you look at previous posts it is difficult to establish the cause(s)
of BS170 failure. If a spike causes a BS170 to short it will look like
an overcurrent failure.
It is also difficult to establish a normal operating temperature for the
PA, some say 2 minutes WSPR just makes them warm, others say HOT.
Personally I throttle back my QDXs used for WSPR . Mainly because I
think 5 Watts is too much for that mode on HF.
73 Alan G4ZFQ
|
On 28/04/2023 07:43, John Z wrote: Testing is always good. It's helpful though to have some idea as to what you are looking for. John, I had something typed but not sent. Hans has said what I was going to and a lot more. It will be good to see a thorough check, the downside is another day to wait for each of the many other things on the list:-) 73 Alan G4ZFQ
|
Mike Berg