Topics

QRP-Labs 50 W Amplifier - low power output


Ben Bangerter, K0IKR
 

I constructed my 50 W amplifier For 20 M at the end of June, and the assembly process went well.  I set the bias for 113 mA with A 20.1 V power supply, no RF power input, and both the input and output ports terminated in 50 ohms.  I tested the amplifier briefly driven with my original QCX, which puts out 2.0 W with a 12 V supply.  My Bird power meter with a 50 W slug Indicated ~15 W to a dummy load.  I then set the amplifier aside while I awaited my QCX+.  That was assembled and aligned, and it delivers ~2.5 W (32 V p-p on the scope) to a 50 ohm dummy load.  When the amplifier, supplied with 20.1 V, is driven by the QCX+, it only delivers ~12.5 W (71 V p-p) To a dummy load.  This is not highly dependent on the amplifier bias setting (3.1 V in this case for the standing current threshold), nor on the power from the QCX+, which can be adjusted by adjusting its supply voltage over a 10 - 14 V range.  I looked at rf waveforms at various places in the amplifier circuit.  Here are waveforms at the gates of the IRF510s, Q1 in yellow and Q2 in violet:

 The drive signals differ in phase by ~180 degrees as expected for a push-pull circuit, but the Q1 waveform in particular appears highly distorted.  The waveforms do not change much at higher bias settings.  This is the waveform at the drain of Q1; that for Q2 is nearly identical.

The frequency of this waveform is ~42 MHz, indicating high third harmonic content.  The amplitude between the cursors is 103 V p-p.  The next waveform is at the secondary of the output transformer at the input to the low-pass filter:

The shape of the waveform is the same as that at the Q1and Q2 drains, the p-p voltage here is 246 V.  The output of the amplifier, after the low-pass filter, is a nice sinusoid at 14 MHz, as expected, but just 92 V p-p, and which dropped to 71 V p-p over time as testing continued.  At no point during my testing did the heat sink become the slightest bit warm to the touch.  The final photo shows the voltage at the input to the low-pass filter with markers to indicate the time interval between the indicated peaks, here 71.6 nS, corresponding to 14 MHz.



i checked the amplifier over carefully, and all the components appeared to be the right values in the right places.  I thought the output transformer might have been reversed.  I removed it and unwound it, and found it was installed correctly.  Then I rebuilt it with new wire and reinstalled it.  While it was out, I examined the low-pass filter with my spectrum analyzer/tracking generator, and saw it had a 3 dB cut-off frequency of ~ 16 MHz, and a loss at 14 MHz of 0.25 dB.  I thought the problem might lie in the T/R switching circuitry, as an incorrectly installed or bad diode could clip the rf waveform and create harmonics.  Checking in both T and R modes, the four 1N4007 diodes all had the correct forward bias currents and correct reverse bias voltages, depending on the on/off condition.  These were examined with amp bias set to zero and no rf input or output.  The rf voltage doubler circuit produced 186 volts when the amplifier was delivering its (measley) 14 watts or so to the dummy load.

At this point, I have run out of ideas, other than replacing the two power MOSFETs.  I have been reluctant to do that except in desperation, being concerned about damaging the pcb.  If anyone has any suggestions for resolving this problem, I will be overjoyed to hear them,

73, Ben K0IKR



ajparent1/KB1GMX
 

On Wed, Aug 19, 2020 at 10:46 AM, Ben Bangerter, K0IKR wrote:
I set the bias for 113 mA wit
Why?   The amp is class C so the bias should be what ever the board
draws and no more.

The adjustment procedure has the bias at zero and no signal in (CW practice mode)
and the bias is turn up until the board current inreases then backed off to the point
before that.

the waveforms are seriously distorted.  Maybe input transformer miswired or 
soldered.  you could have an  inter-turn short.

Something is wrong and its likely in the input side.

Allison

-------------------------------
Please reply on list so we can share.
No private email, it goes to a bit bucket due address harvesting


Ben Bangerter, K0IKR
 

Allison,

You ask why 113 mA.  That is the current drawn by the amplifier at the point where any further clockwise rotation of the bias setting potentiometer begins to increase the current, as the IRF510s begin to conduct.  It is also the current drawn when the potentiometer is fully counterclockwise.  I have read and understand the instructions.  I have set the bias so that the amplifier is operating in class B, that knife-edge between class C and class AB where the output power is maximized with no standing current in the IRF510s in the absence of rf drive.  Thank you for the suggestion that the problem may reside in the trifilar input transformer.  I have ordered some 22 gauge enameled wire, and will rebuild that transformer when it arrives.

73, Ben K0IKR
 


Timothy Fidler
 

1. check the phasing of the phasing choke / transformer.

2.  Get some sort of temperature probe , borrowed or one you have . work out some consistent thermal insul system and see if you can find a difference between the two  power mosfets under bias conditions. If one is notably cooler given the identical bias  then you prolly have your issue pointed at. But in that case check all solder joints in the current path very carefully before you point  the gun at the silicon.

3.  If Power Mosfets need a change out , buy them as  a pair  from Hans or some  known source. There are a lot of white spot and repackaged IRF510s around. You need a pair that come from the same production batch as the bias scheme on this amp is so crude. 

4. Examine the Pot. R5. Is it Exactly the same in the build photos ? I have a theory that this very unusual common bias scheme works because there is some capacitive bypassing from GND to the wiper of the bias potentiometer.  If the potentiometer checks out physically then there is the kludge available of using a 0.1uF multilayer cap (ie RF rated) and a 100R resistor mounted from the wiper to GND to ensure that the drive signal can get from the source of the respective IRF to the cold side of the driving transformer 's Ct.

5. Hans gets his ferrites pounded together by Dwarfs  mit Feuer und Eisen from some Trollish undeground factory.  They have no brand but are very cheap which gets you this kit for USD 25 or so.  Consequently it is possible that the ones in the OPT or even the IP single toroid are not as per the golden test board.  the OPT ones are nominally type 43 but do not have a mafr certification I believe. It is possible that his supplier has mixed cores of different  material types. The colour and dimensions are a good go by.

From a dwarfish troll in the Sth Pacific.


Timothy Fidler
 

Point 5 . of the last  Troll post is from Left Field  (at least in terms of the OPTrafo only ) because all the indications are that the problem may be ahead of the gates of the Mosfets.

regards, TEF


Hans Summers
 

Hi TIm

Your point 5 isn't just from left field it's off the field all together. There is nothing wrong with the supplied ferrites. I spent a vast amount of time on suppliers and material testing in order to choose the best ferrite for the job. I would not be supplying ferrites which are not up to the job. Many constructors have built this amplifier successfully and to spec. As with all QRP Labs kits, if it does not work there is a probability of at least 99% that it is due to an assembly error or fault, not a defective component. 

Regarding transistors, yes, replace them together. I wouldn't agree that the biasing scheme is crude, it is designed for a particular task and it accomplishes this perfectly. Separate bias for the two MOSFETs This is a Class C amp. The answer may be different if it was a Class AB amp where we were aiming for highest possible linearity. Nevertheless I would always recommend replacing both together, and would recommend that for any PA including the Class E finals in the QCX/QCX+. For the pennies these transistors cost it isn't worth taking chances. 

Sourcing the transistors does need care. A lot of cheap IRF510s are just some generic MOSFET onto which is printed various labels purely for selling them - such as IRF510, IRF520, IRF530... they don't necessarily have the correct characteristics of our beloved IRF510. The transistors used in the kit are Vishay brand and are reliable and correct. If you want to replace them then buy IRF510s from a reputable supplier. For example if you buy in US, Digikey IRF510s are also the same Vishay type and are known to be good. It doesn't have to be Vishay, just buy from a reputable place so you know it is is what it says it is. 

Regarding point 4... somewhat just off the Left Field boundary... do NOT add a capacitor and 100-ohm resistor at the wiper of the pot! You really will reduce the power output of the amp then, I guarantee it! I don't see what is confusing. The pot sets a DC bias that is applied to both gates. It works...

Good luck!

73 Hans G0UPL
http://qrp-labs.com

On Sat, Aug 22, 2020 at 12:29 AM Timothy Fidler <engstr@...> wrote:
Point 5 . of the last  Troll post is from Left Field  (at least in terms of the OPTrafo only ) because all the indications are that the problem may be ahead of the gates of the Mosfets.

regards, TEF


George Korper
 

The toroids are very good.  I use them in making 49:1 antenna transformers for Half Wave End Fed Antenna systems
where they also work well. Never had a bad one in 10 amps. 
That being said I have found that stripping and pre-tinning the enamel wire is essential. 
Checking the continuity of the transformer before soldering on the board, allows for easy soldering with less heat. 

On Sat, Aug 22, 2020 at 12:56 AM Hans Summers <hans.summers@...> wrote:
Hi TIm

Your point 5 isn't just from left field it's off the field all together. There is nothing wrong with the supplied ferrites. I spent a vast amount of time on suppliers and material testing in order to choose the best ferrite for the job. I would not be supplying ferrites which are not up to the job. Many constructors have built this amplifier successfully and to spec. As with all QRP Labs kits, if it does not work there is a probability of at least 99% that it is due to an assembly error or fault, not a defective component. 

Regarding transistors, yes, replace them together. I wouldn't agree that the biasing scheme is crude, it is designed for a particular task and it accomplishes this perfectly. Separate bias for the two MOSFETs This is a Class C amp. The answer may be different if it was a Class AB amp where we were aiming for highest possible linearity. Nevertheless I would always recommend replacing both together, and would recommend that for any PA including the Class E finals in the QCX/QCX+. For the pennies these transistors cost it isn't worth taking chances. 

Sourcing the transistors does need care. A lot of cheap IRF510s are just some generic MOSFET onto which is printed various labels purely for selling them - such as IRF510, IRF520, IRF530... they don't necessarily have the correct characteristics of our beloved IRF510. The transistors used in the kit are Vishay brand and are reliable and correct. If you want to replace them then buy IRF510s from a reputable supplier. For example if you buy in US, Digikey IRF510s are also the same Vishay type and are known to be good. It doesn't have to be Vishay, just buy from a reputable place so you know it is is what it says it is. 

Regarding point 4... somewhat just off the Left Field boundary... do NOT add a capacitor and 100-ohm resistor at the wiper of the pot! You really will reduce the power output of the amp then, I guarantee it! I don't see what is confusing. The pot sets a DC bias that is applied to both gates. It works...

Good luck!

73 Hans G0UPL
http://qrp-labs.com

On Sat, Aug 22, 2020 at 12:29 AM Timothy Fidler <engstr@...> wrote:
Point 5 . of the last  Troll post is from Left Field  (at least in terms of the OPTrafo only ) because all the indications are that the problem may be ahead of the gates of the Mosfets.

regards, TEF


Timothy Fidler
 

Hans you need to be building the QSX not auditing my moronic replies : - ) .  I tell lyou what - I have an HP 48G that has never been abused and now refuses to boot up. Even after the kiss of death the paper clip reset.  So I can't give it to my nephew.  Here's the deal I offer no more free tech advice to the Punters and you repair the HP48 and sent it on to Nelson and I get to be the superhero.  well perhaps not.  Enough chitchat and attempts at subversion - 
========================================================================================
"Your point 5 isn't just from left field it's off the field all together. There is nothing wrong with the supplied ferrites. I spent a vast amount of time on suppliers and material testing in order to choose the best ferrite for the job."

Hans, I am not suggesting QrpLabs was supplying non standard parts but plenty of Chinese suppliers will sell you part X and when they run out of that sack,  will sell you part Y but is a mix of type 61 ferrite and ground up Yalu river mud which looks the same but works like ----well Yalu river mud.

 

YOu end up with mud on yer face but whoflungdung walks off with your greenbacks in his back pocket.  IN general if you have a 2 year or more relationship with your suppliers and pay on the nail this won't happen.  But beware... The CHinese remember the Opium wars and who started that .. and where were you born ?  Of course the cores might come out of Germany or France for all I know.
The story of the vulcan makers of the ferrites is sort of a joke with me because all the printed material from Amidon and others talks of proprietary milling and sintering operations and it is clear there is a lot of IP in the manufacturing  ferrite business so it is just as easy to say Dwarfs  mit Feuer und Eisen run the process. It is also a hint at the poor bastards who died under Von Braun et al at Dora Mittelwerk (underground) building V2 rockets when WWII was clearly already lost......

My suggestion of  the 0.1uF MLM capacitor and 100R  to the wiper and the other end  of the string to ground . RE ; "do NOT add a capacitor and 100-ohm resistor at the wiper of the pot! "  Really ?  remember these additional Weakware components (untested in the real world )  are in series. That is obvious to blind freddy  (in my 'umble opinion)  but  you might have thought or Assumed I meant in parallel.

 

  Wass fur -this circuit ? well  the CT (gnd) RF signal induced over the Trafo has to go through the active MOS gate and back to GND and then get back to the CTap.  Considering  the R5 preset  in the expected Bias posn I'd expect the wiper to be sitting at 2260 R.  That means that the RF drive signal has to go through that to get to the CT of the IPT. A 0.1uF cap will have a Z of -j 0.159 ohms at 10 Mhz . I added the  100R to make sure that the RF current did not see too much of a good thing  (the rest of the drive ac voltage is lost across the IRF's gate and in the body resistance of the IRF and the inductive losses in the legs and PCB traces, not to mention in the Transformer coil and Rs generated by the flux in the ferrite core).

That core is likely type 43 material but could be closer to type -61 given you need the amp to work well on 20 m too. ) .  I suspect you thought the intent was to bring the DC bias down to 0.204 V or so,  which of course is effectively no bias at all and the Mosfets  won't dance for much of the drive cycle.  Definitely not. 

Hans - It might be worth testing out this capacitor and resistor addn idea in Turkey on the Golden Build reference amp and seeing what the power out consequences were.  I'd be very interested in  a PMessage advising the results.




Timothy Fidler
 

I almost lost my complete post on the issues  hence I start a new one to finish 

"labels purely for selling them - such as IRF510, IRF520, IRF530... "

At 80 and 40 m the IRF530 is almost as good as the IRF 510 and will produce 5-6 W out Linear from a CL A amp.
above 8 Mhz the 510 starts to shine.  BUT 
what would be really interesting is for an existing builder or Hans with the Golden Build to take a IRFZ24N which is a current production switching MOS approx equiv to the IRF 510 and see what the performance is of the amp with the substit. at 20 V. All these later production upgrade Industrial N Mos seem to have VDS max =60V versus 100V for the older parts. Hence my suggestion of a comparison at 20V supply. Why ? because IRFZ24N is in current production and builders in the UK US and Australia are able to buy them readily knowing that they are unlikely to be fakes.

Eg in Australia you'd get them From **Rockby Electronics  VIC for about AUD 2.2 or 2.5 Ea. This might seem steep but it is better to be dealing with a known part.  People like the CarandJay  (Australia ) will by the cheapest dreck they can find on the grey market for their semiconductors they can get away with.  IRF 510s included.   ** Rockby will  buy in oddball parts as part of their bulk orders on Soanar and the like as long as you pre pay or are known to them. And no I am not Eugene's brother in law :-_)


Hans Summers
 

Hi Tim

I'm not sure what the problem is that we're trying to fix here. My goal was to design a 50W Class-C amp kit for the QCX, that produces 50W without stress and at low cost; furthermore it should be reproducible easily by the kit builders. Common components so spare parts are easy. The IRF510 is about as ubiquitous as it gets. So I'm not sure what problem is being aimed at here... and it seems to me best to prioritize other projects! 

73 Hans G0UPL



On Sat, Aug 22, 2020, 14:34 Timothy Fidler <engstr@...> wrote:
I almost lost my complete post on the issues  hence I start a new one to finish 

"labels purely for selling them - such as IRF510, IRF520, IRF530... "

At 80 and 40 m the IRF530 is almost as good as the IRF 510 and will produce 5-6 W out Linear from a CL A amp.
above 8 Mhz the 510 starts to shine.  BUT 
what would be really interesting is for an existing builder or Hans with the Golden Build to take a IRFZ24N which is a current production switching MOS approx equiv to the IRF 510 and see what the performance is of the amp with the substit. at 20 V. All these later production upgrade Industrial N Mos seem to have VDS max =60V versus 100V for the older parts. Hence my suggestion of a comparison at 20V supply. Why ? because IRFZ24N is in current production and builders in the UK US and Australia are able to buy them readily knowing that they are unlikely to be fakes.

Eg in Australia you'd get them From **Rockby Electronics  VIC for about AUD 2.2 or 2.5 Ea. This might seem steep but it is better to be dealing with a known part.  People like the CarandJay  (Australia ) will by the cheapest dreck they can find on the grey market for their semiconductors they can get away with.  IRF 510s included.   ** Rockby will  buy in oddball parts as part of their bulk orders on Soanar and the like as long as you pre pay or are known to them. And no I am not Eugene's brother in law :-_)


Ben Bangerter, K0IKR
 

Tim and Hans,

Thank you for your suggestions.  I did replace the IRF510s, not because I seriously thought they were the source of my problem but because I had some on hand and it was easy to do.  These were obtained from Arrow in the US, a major component supplier.  They are from Vishay, and all identically marked so presumably from the same manufacturing lot.  The performance of the amplifier was identical with that before replacing them, except that the bias voltage for ~class B operation was a bit lower, 2.95 V compared with 3.1 V for the originals.

Hans,  I agree that a bad component is the least likely problem.  I have gone over my build carefully with a loupe, and have found no opens, shorts or suspect solder joints.  The magnet wire I ordered (#22 AWG, 0.644mm) should arrive today, and I will rebuild both the trifilar input transformer and the bifilar drain feed transformer, and closely inspect the ferrites for cracks.  I will see where that leads.  I am rather surprised that bypassing the wiper of the bias pot to ground leads to a reduction in power output.  I imagine you were surprised by this finding as well.  Without a bypass capacitor, the pot and associated board traces are now part of the rf path I should think, up to the point where C4 (2.2 nF) goes to ground.  Puzzling.

73, Ben K0IKR


Hans Summers
 

Hi Ben

Yes, a reduction in power when the capacitor was connected was not what I was expecting. But it was what I saw on my measurement equipment. RF is black magic. So I chose to believe my eyes not my theories :-)

73 Hans G0UPL

On Sat, Aug 22, 2020, 19:21 Ben Bangerter, K0IKR via groups.io <bwbangerter=yahoo.com@groups.io> wrote:

Tim and Hans,

Thank you for your suggestions.  I did replace the IRF510s, not because I seriously thought they were the source of my problem but because I had some on hand and it was easy to do.  These were obtained from Arrow in the US, a major component supplier.  They are from Vishay, and all identically marked so presumably from the same manufacturing lot.  The performance of the amplifier was identical with that before replacing them, except that the bias voltage for ~class B operation was a bit lower, 2.95 V compared with 3.1 V for the originals.

Hans,  I agree that a bad component is the least likely problem.  I have gone over my build carefully with a loupe, and have found no opens, shorts or suspect solder joints.  The magnet wire I ordered (#22 AWG, 0.644mm) should arrive today, and I will rebuild both the trifilar input transformer and the bifilar drain feed transformer, and closely inspect the ferrites for cracks.  I will see where that leads.  I am rather surprised that bypassing the wiper of the bias pot to ground leads to a reduction in power output.  I imagine you were surprised by this finding as well.  Without a bypass capacitor, the pot and associated board traces are now part of the rf path I should think, up to the point where C4 (2.2 nF) goes to ground.  Puzzling.

73, Ben K0IKR


George Korper
 

RF may travel in paths you have not forseen. That does not make it black magic, 


On Sun, Aug 23, 2020, 8:34 PM Hans Summers <hans.summers@...> wrote:
Hi Ben

Yes, a reduction in power when the capacitor was connected was not what I was expecting. But it was what I saw on my measurement equipment. RF is black magic. So I chose to believe my eyes not my theories :-)

73 Hans G0UPL

On Sat, Aug 22, 2020, 19:21 Ben Bangerter, K0IKR via groups.io <bwbangerter=yahoo.com@groups.io> wrote:

Tim and Hans,

Thank you for your suggestions.  I did replace the IRF510s, not because I seriously thought they were the source of my problem but because I had some on hand and it was easy to do.  These were obtained from Arrow in the US, a major component supplier.  They are from Vishay, and all identically marked so presumably from the same manufacturing lot.  The performance of the amplifier was identical with that before replacing them, except that the bias voltage for ~class B operation was a bit lower, 2.95 V compared with 3.1 V for the originals.

Hans,  I agree that a bad component is the least likely problem.  I have gone over my build carefully with a loupe, and have found no opens, shorts or suspect solder joints.  The magnet wire I ordered (#22 AWG, 0.644mm) should arrive today, and I will rebuild both the trifilar input transformer and the bifilar drain feed transformer, and closely inspect the ferrites for cracks.  I will see where that leads.  I am rather surprised that bypassing the wiper of the bias pot to ground leads to a reduction in power output.  I imagine you were surprised by this finding as well.  Without a bypass capacitor, the pot and associated board traces are now part of the rf path I should think, up to the point where C4 (2.2 nF) goes to ground.  Puzzling.

73, Ben K0IKR


Ben Bangerter, K0IKR
 

For those who have been following this thread, the saga continues.  Over the weekend I removed the input transformer T1 and the drain feed choke T2.  Both had been constructed correctly, and there was no indication of a short or other anomaly.  I rewound both transformers with new wire and reinstalled them in the amplifier.  After putting things back together, my results were the same:  ~10-12 W output from the amplifier (with 20.1 VDC supplied) when driven by ~2.5 W from the QCX+.

While making subsequent voltage measurements, I found that bias was applied to the IRF510s when the PTT signal was NOT asserted (+5V) as well as when it was.  Not good.  I then found that the drain of Q4 was at 12.48 V in receive mode, when it should have been zero volts.  The gate of Q4 was at 18.3 V rather than 20.1 V as it should have been.  The 1.8 V drop across R10 (and also R11) indicated that something  was drawing ~180 microAmps through R10 and R11, partially turning Q4 on.  It turns out that D7, with 12.57 V at the anode and 16.54 V at the cathode, and thus reverse-biased by ~4.0 V, was nevertheless passing a current of 177 microAmps and causing the unexpected behavior.  Diode D1, also in the transmit path and also with a reverse bias of ~4.0 V, showed no detectable reverse current flow.

Now this may have little or nothing to do with my low rf power output problem, but the issue needs to be resolved before I can proceed.  Though I have several 1A rectifier diodes on hand, none are 1N4007s.  I will have some in a couple of days, and will see where things stand after D7 is replaced.

73, Ben K0IKR


Timothy Fidler
 

Ben this occurred to me days ago and I was remiss not to put it on the wire.

IN order to eliminate as much as possible in the input cct what 
about pulling the IRFs and replacing the G to S with one off 180 pf Capacitor (SM pref for low loss as per a Mosfet gate ) , in parallel with a half watt 10 K resistor and the sum node take to the Source pad with say a 4 ohm R Now if you stick the juice into it you have two guaranteed linear junctions that look to the world  as an approx of an IRF 510 input node.  If you still get that scope trace set then you know that the IRFs are not implicated an that the input circuitry is probably not implicated. It seems you are way off chasing the issue down and it looks to be in other semi conductors but that's still a guess just now. The idea above might be useful to AN Other. Might have been useful to you too if I pulled my finger out , eh ?


Timothy Fidler
 

Ben , have you made any more progress  or are you still waiting for parts ?  

In terms of my cryptic note to GK I Now have more time. 


If the driven Mosfet at one point of time is A and non driven  is B then any current coming from A to GND has to ret to the CT of the transformer via either the  Source of B or via the high Z path of the 5K trimmer. (the fraction that the wiper connects to the CT when bias is set.) I suspect this bias voltage even for class C opn of a 510 is as high as 3.5 V.   so the main path is reverse flow via the  B IRF .  Is this possible ? 

The  Awful GroupsIO software LOST the rest of my tech workings. Only the above notes were done off line.
stream of consciousness dump follow - 
 
Input power requirement to drive the IRFs about 700 mW . Corresp.  V peak about 5.5V
 
Hints  180 pf Of Gate capacitance.  30 to 90 degrees conduction angle has to be accounted for.
Other IRF based non CT type circuits work and so the reverse flow path must exist even  if possibly  at somewhat
higher Z as bias goes negative.
 
50 : 200 ohms versus the -j176 ohms of the two IRF gates in series at 10 Mhz - not a bad match but of course the use of 180 pf IN the gate calculations under temporary  reverse bias  is problematical .  I f you have any EE training
you can work it out ;  if not sorry bad luck . I am fed up with all the workings being lost by being dumped out of GIO when  I hit the Send button.
 
TEF 


Ben Bangerter, K0IKR
 

This is a followup to my post of Aug 24 on this thread.  I replaced the faulty diode D7 with a new 1N4007, which I determined showed a negligibly small reverse current flow when having a reverse bias of 20VDC applied.  The drain of Q4 is now at 0 V in receive mode, as it should be.  I next characterized the T/R switching by taking a set of voltage measurements under the following conditions:  20.1 V amplifier power supply, Q1/Q2 gate bias voltage 0.0 V (bias pot fully ccw), 50 ohm terminations on both input and output of amplifier.  The findings were as follows:
In receive mode, D1 and D7 reverse biased at 20.1 V, unmeasurably small reverse current flow (<1 uA); D3 and D5 forward biased with 10 mA forward current.
In transmit mode, D1 forward biased with 19.3 mA forward current, D7 forward biased with 40.6 mA forward current; D3 and D5 reverse biased at 18.4 V with unmeasurably small reverse current flow.
These findings are suitable, and as predicted from the observed power supply voltage and component values.

I then disconnected the hot end of the secondary of T3 and connected the output of the QCX+ to the input of the low-pass filter, and with the amplifier in transmit mode and a dummy load on the output of the amplifier, observed a loss of ~1 dB through the filter and the T/R switch associated with D7.  About what might be expected.

The amplifier was reassembled and showed the same output level as previously observed, ~10-12 W with a ~2.7 W input from the QCX+, with a 20.1 V power supply.

At this point I have rebuilt the three transformers, with new wire, and replaced the two IRF510 transistors.  The original transformers had been constructed and installed properly.  I have carefully examined both sides of the pcb and have found no shorts, opens, bad solder joints or any other anomalies.  I have verified the locations and marked values of resistors, inductors and some capacitors.  The markings on most of the ceramic capacitors can not be read with those components mounted on the pcb.

I am essentially out of ideas, apart from experimenting with modifying the basic design of the amplifier circuit.  I am a bit surprised than no one with a QRP-Labs 50W amplifier, other than Hans, has contributed to this discussion.  Anyone?  I do not disagree with Tim's analysis of the input circuit, but that is not unique to my copy of the amplifier.  It is clearly highly reactive (capacitive) and nowhere near 50 ohms resistive.

73, Ben K0IKR


George Korper
 

This is a difficult problem and pretty depressing. I have had the same situation myself on one amp build. 
My conclusion was there was a fault in the board. A trace even a trace that looks good with a DVM is not
good with RF. I discarded the amp and went on with other builds, changing my method.

 1) I tinned all the transformer leads, so that I would not need much heat on the PCB itself to get good flow

2) I cleaned all of the PCB and some leads with alcohol before soldering.

3) I didn't heat the board too much.


You could try hard wiring some traces. It doesn't take that long or just start over. 


Others may give you more helpful suggestions. Mine to start over using great care
worked for me. 


On Sun, Aug 30, 2020, 2:54 AM Ben Bangerter, K0IKR via groups.io <bwbangerter=yahoo.com@groups.io> wrote:

This is a followup to my post of Aug 24 on this thread.  I replaced the faulty diode D7 with a new 1N4007, which I determined showed a negligibly small reverse current flow when having a reverse bias of 20VDC applied.  The drain of Q4 is now at 0 V in receive mode, as it should be.  I next characterized the T/R switching by taking a set of voltage measurements under the following conditions:  20.1 V amplifier power supply, Q1/Q2 gate bias voltage 0.0 V (bias pot fully ccw), 50 ohm terminations on both input and output of amplifier.  The findings were as follows:
In receive mode, D1 and D7 reverse biased at 20.1 V, unmeasurably small reverse current flow (<1 uA); D3 and D5 forward biased with 10 mA forward current.
In transmit mode, D1 forward biased with 19.3 mA forward current, D7 forward biased with 40.6 mA forward current; D3 and D5 reverse biased at 18.4 V with unmeasurably small reverse current flow.
These findings are suitable, and as predicted from the observed power supply voltage and component values.

I then disconnected the hot end of the secondary of T3 and connected the output of the QCX+ to the input of the low-pass filter, and with the amplifier in transmit mode and a dummy load on the output of the amplifier, observed a loss of ~1 dB through the filter and the T/R switch associated with D7.  About what might be expected.

The amplifier was reassembled and showed the same output level as previously observed, ~10-12 W with a ~2.7 W input from the QCX+, with a 20.1 V power supply.

At this point I have rebuilt the three transformers, with new wire, and replaced the two IRF510 transistors.  The original transformers had been constructed and installed properly.  I have carefully examined both sides of the pcb and have found no shorts, opens, bad solder joints or any other anomalies.  I have verified the locations and marked values of resistors, inductors and some capacitors.  The markings on most of the ceramic capacitors can not be read with those components mounted on the pcb.

I am essentially out of ideas, apart from experimenting with modifying the basic design of the amplifier circuit.  I am a bit surprised than no one with a QRP-Labs 50W amplifier, other than Hans, has contributed to this discussion.  Anyone?  I do not disagree with Tim's analysis of the input circuit, but that is not unique to my copy of the amplifier.  It is clearly highly reactive (capacitive) and nowhere near 50 ohms resistive.

73, Ben K0IKR


Ben Bangerter, K0IKR
 

Well George, that's discouraging.  Did you eventually succeed in getting a working amplifier?  Was it with a new kit?  I hate to give up on a project, but I am out of ideas.  Perhaps it is time to admit defeat and move on with my life.

73, Ben K0IKR


George Korper
 

Yes, I have built 10 or 11. I was working on improving my build quality and used them for the education. I wanted reliable methods so future projects would work first time. I made myself simple rules to follow.  Now all of them work well or it's a bad PCB. The AMP themselves all come out identical like a production line 

I think if you purchase another it would be well worth it. The amps are difficult to repair because one slip and your liable to start a cascade of difficulties and smoke.

I can send you my rules if you like. 

On Sun, Aug 30, 2020, 3:29 PM Ben Bangerter, K0IKR via groups.io <bwbangerter=yahoo.com@groups.io> wrote:
Well George, that's discouraging.  Did you eventually succeed in getting a working amplifier?  Was it with a new kit?  I hate to give up on a project, but I am out of ideas.  Perhaps it is time to admit defeat and move on with my life.

73, Ben K0IKR