Topics

QCX: 30m Rev 4B Output Power Less than 500 mW #qcx

Paul Seger K7AZT
 

Greetings!

I built the 30m QCX Rev 4b transceiver which was shipped post 12 April 2019.  The output power is far below expectation.  I've done troubleshooting recommended in this message group, varied L1 through L4, pulled and measured some of the filter capacitor values, checked for leakage paths, but come up empty.   Does anyone have ideas?  What follows is some of the testing.  Given the measured values, perhaps someone can point me in the right direction.
 
I'm stumped.  Thanks in advance.  
73, Paul
 
 
The kit was built with the reverse protection diode in place.  The QCX operated into a 50 ohm dummy load.  Output power was measured using the internal RF Power function, and confirmed with an MFJ-870 SWR/Watt meter.  The power supply was a 12 AHr 12.6V battery.  Battery voltage droop on transmit was under 25 mV.  Supply current was 250 mA on transmit.  Correct output frequency was verified via Kenwood TS-590 and oscilloscope.
 
1.  This is my starting point:  The power output was 0.45W with coils wound to the specified number of turns in the manual.
 
2.  Removing L1, L2, L3 turns in various combinations got up to 1.30W.  However the resulting filter rolloff at 20 MHz was an unacceptable -11.5 dB down from 10 MHzwhich would enable spurious emissions.
 
3.  Coils L1, L2, L3, L4 were then rewound to the exact inductance values specified in the manual.  The output power went down to 0.18W.
 
4.  I modeled the output filter in LTSpice to get a feel for how the output filter should perform using the inductance values (not turns) specified in the QCX manual.  The filter was flat at -6 dB to about 11 MHz, and was -55 dB down at 20 MHz.  This is an effective filter for the class E amplifier.  My greatly modified filter (see #2 above) that achieved 1.30 W was -8.5 dB at 10 MHz and -20dB at 20 MHz.  The second harmonic is only 11.5 dB down from the fundamental.  This is not an effective filter.  My conclusion is that the QCX filter design is exactly what one would want if I could get the power out of it.  A tested modified filter that got any meaningful output power was hardly a filter at all and would invite an FCC violation.
 
5. I then investigated whether RF choke L4 might be the problem.  With L1, L2, L3 at the specified inductance values, the number of turns on RF choke L4 was varied over a wide range.  The output power increased from 0.18 W at specified 14 turns to 0.28 W when reduced to 4 turns.  The output power was still too low and the L4 inductance value insanely small.  L4 does not appear to be a solution.
 
6.  Component inspection:  The capacitor locations and markings in the power amplifier and low pass filter were were verified to be correct, several times.  Capacitor C28 was pulled, measured, and found to be dead on 270 pF.
 
7.  C33 was pulled to isolate the TR switch from the low pass filter.  This had no effect.  The TR switch is not part of the problem.
 
8.  Waveforms:  At 12.6 VDC and L1, L2, L3, L4 at specified inductance, the drains of Q1, Q2, Q3 for all tests was a clean 32V peak-to-peak signal that look like the positive half of a sine wave clamped near ground for half a cycle.  The waveforms at L1, L2, L3 were clean sine waves viewed with an oscilloscope.  The harmonic content was not measured as I don't have a spectrum analyzer.  When L4 had very few turns the drains were near 50 Vpp and somewhat ragged at the top, but not a square wave.
 
9.  Ground paths:  There were no DC paths to ground in the output stage.  With a 50 ohm dummy load the resistance to ground along L1, L2, L3 was 50 ohms.  With the dummy load removed the resistance was infinite. Resistance to ground on both sides of L4 was infinite regardless of dummy load.
 
10.  The gates of Q1, Q2, Q3 have approximately 5V pp square waves on transmit.
 
11.  Oscilloscope measurements were again taken @ 12.8 VDC 12 AHr battery supply, 50 ohm dummy load, power measured with the QCX, all coils at specified inductance:
 
DC input to L4:  12.4 VDC when keyed (scope measured)
Input to L3:    4.80W, 12.8 V zero-to-peak (near ground to 12.8V), half quasi sine wave
Input to L2:    2.15W, 28 Vp-p sine wave  (1.96W from RMS voltage)
Input to L1:    0.84W, 18 Vp-p sine wave  (0.81W from RMS voltage)
Output:         0.18W,  8 Vp-p sine wave  (0.16W from RMS voltage)
 
12.  Losses: Using values from set #11 above, power lost per stage is L3 to L2 -3.5 dB, L2 to L1 -4.1 dB, L1 to output: -6.7 dB.   The losses are high.  Loss through the last stage (L1/C28) appears particularly excessive.  The value of C28 had been confirmed to be 270pF as specified, and L1 was wound to exact inductance.
 

Tim Raven
 

Hi Paul, 
I built the QCX for 60m and also had a lot of problems getting the expected power out. I tried quite a few of the things you have tried, without success. Then I came across a posting that mentioned the output matching of a class E amplifier could be unpredictable. So I put all the toroids back to there exact value, ie the number of turns in the assembly manual, apart from L3, which I removed 2 turns from, which resulted in the output power doubling from 2 to 4 watts with 13.8 volts through the reverse bias diode. 

A couple of other people have tried this, and it's resolved their problems also. 

It's a brilliant CW rig, so I do wish you every success

73 de Tim G4ARI

Alan G4ZFQ
 

Paul

Supply current was 250 mA on transmit.
I think this is half of that normally reported? 3W total.
Presuming they have 12V on their drains are all the BS170s conducting?

My conclusion is that the QCX filter design is exactly what one would want
Quite. It is an established design used in many projects.
The only thing is, as you say, the cutoff is close so the inductors often need a turn or two removed if AL value is at the higher end of tolerance.
It seems that in a few cases a LPF capacitor has been at fault, there are many posts on this subject, difficult to draw a conclusion.

73 Alan G4ZFQ

Richard G4TGJ
 

I think the capacitors supplied are the problem. I'd replace the caps in the LPF and class E amp with NP0.
--
73
Richard
G4TGJ

Alan G4ZFQ
 

I think the capacitors supplied are the problem. I'd replace the caps in the LPF and class E amp with NP0.
Richard,

Possibly but do the symptoms reported match others who have said the capacitors were faulty?
No harm in trying but as Hans has said only a few have had that problem.

73 Alan G4ZFQ

Kelly Jack
 

Do the caps get hot? Seems like they might if you're losing that much power between the inductors. Also, it seems like the only thing you haven't tested yet.

73

Simon
VK3ELH

Richard G4TGJ
 

My 30m QCX had low output power. I changed the LPF to the improved GQRP design including using NP0 caps. Unfortunately I don't think Hans supplies suitable components as they are not NP0 and too many people have reported a problem.
--
73
Richard
G4TGJ

Hans Summers
 

Hi Richard

The supplied capacitors in all QRP Labs kits are NP0/C0G type. The number of people that have reported any problem is small proportionately to the number of kits produced. I can replace any capacitors which are defective. 

73 Hans G0UPL

On Wed, May 29, 2019 at 6:17 PM Richard G4TGJ <rpt@...> wrote:
My 30m QCX had low output power. I changed the LPF to the improved GQRP design including using NP0 caps. Unfortunately I don't think Hans supplies suitable components as they are not NP0 and too many people have reported a problem.
--
73
Richard
G4TGJ

Arv Evans
 

Several potential problems with filters that do not seem to be working correctly.

  1)  Inductance is wrong

  2)  Capacitance is wrong

  3)  Soldering problems

  4)  Impedance mismatch on input or output, or both

Many seem to have overlooked the possibility of output filters seeing an impedance
mismatch.  This can be compensated for by changing filter components so that the
filter also transforms impedance for a better match, but then you need to re-scan the
filter to verify it still has the proper response characteristic.

Before blaming the capacitors we probably need to take a step back and look at the
larger picture to see what the real problem might be.  To-date there apparently has
not been a definitive test of those supposedly bad capacitors to determine if they
really are the problem. 

However, capacitors are easy to change so that can still be the first step if one lacks
test equipment to prove them bad or good. 

Arv
_._


On Wed, May 29, 2019 at 9:17 AM Richard G4TGJ <rpt@...> wrote:
My 30m QCX had low output power. I changed the LPF to the improved GQRP design including using NP0 caps. Unfortunately I don't think Hans supplies suitable components as they are not NP0 and too many people have reported a problem.
--
73
Richard
G4TGJ

Dennis AG7IN
 

Hi Paul,
I built a 30m rev 4 kit and have T1 and L1 through L4 with windings per the manual, no changes there, and used silver mica caps for C25-28, C30 and C8.
Have 3.3 watts out, internal RF measure, with 13.8 volts, and drawing 0.46 amps.
Check that C5 and C8 are not both in place for 30m.
Posted some pics of both board sides a month ago if interested, search 30m mods.

73,
Dennis
AG7IN

Richard G4TGJ
 

Sorry Hans, I take it back. With my experience and those of others it seemed likely the capacitors were to blame but I accept that the incidence of problems is very low. Unfortunately, people are more likely to complain than praise so these problems have been unfairly multiplied.


On 29 May 2019 4:26 pm, Hans Summers <hans.summers@...> wrote:
Hi Richard

The supplied capacitors in all QRP Labs kits are NP0/C0G type. The number of people that have reported any problem is small proportionately to the number of kits produced. I can replace any capacitors which are defective. 

73 Hans G0UPL

On Wed, May 29, 2019 at 6:17 PM Richard G4TGJ <rpt@...> wrote:
My 30m QCX had low output power. I changed the LPF to the improved GQRP design including using NP0 caps. Unfortunately I don't think Hans supplies suitable components as they are not NP0 and too many people have reported a problem.
--
73
Richard
G4TGJ



--
73
Richard
G4TGJ

Adam Goler
 


8.  Waveforms:  At 12.6 VDC and L1, L2, L3, L4 at specified inductance, the drains of Q1, Q2, Q3 for all tests was a clean 32V peak-to-peak signal that look like the positive half of a sine wave clamped near ground for half a cycle.  The waveforms at L1, L2, L3 were clean sine waves viewed with an oscilloscope.  The harmonic content was not measured as I don't have a spectrum analyzer.  When L4 had very few turns the drains were near 50 Vpp and somewhat ragged at the top, but not a square wave.
 
Out of curiosity, when you made these scope measurements, what specific location on the PCB did you use for a "ground" reference?

When you follow the calibration procedure in the manual, does everything go as expected?

73 and GL!

--
73, Adam KM6PHD NAQCC#9838

Paul Seger K7AZT
 

Hello All,
 
Thanks to all who have offered suggestions.  They have been very helpful in eliminating possible issues and in pointing direction for further investigation.  With your help I'm confident this rig will see air time.
 
Below is current status and my reply to suggestions.  The brief summary is that the input impedance of the low pass filter is suspect.  Thanks for putting me on that track.  I'll look into it and report.  
 
73, Paul
K7AZT
 
----------------------------------
 
Since my first note I did additional testing with an Astron power supply set to 13.95 VDC replacing the battery.  I remeasured voltages, power, and waveforms throughout the power amp / output filter.  Scope confirmed Q6 collector is rock solid power when keyed.  There is no significant difference from my first note with respect to voltages, power, or waveforms.
 
To specific suggestions:

Capacitors C5 & C8: C5 is in place, marked 220, 22pF as specified.  C8 not in place, as specified.
 
Capacitor Temperature: There was no discernible heating of capacitors due to transmitting.
 
Solder joints: all inspected.  None cold. No haze. Leads fully enclosed by solder.
 
Scope probe ground point: used BNC connector ground and the paddle jack shield.  No difference between the points.  Scope leads are short. Waveforms are cleaner than in assembly manual.
 
Calibration: Calibration went smoothly.  Receiver sections is really good.  I didn't expect this level receiver in such an inexpensive kit.  There were no start up problems of any kind.
 
Transmit Current: My transmit current is low by about 50% compared with others.  Drains Q1, Q2, Q3 have zero to 33V peak-to-peak waveforms, pulling down near to ground one half cycle, and near half-sine on other half of cycle. This is in line with others.  I believe the drains are correct, but suspect that the low pass filter is not accepting power.  See next.
 
Impedance mismatch between power amp and filter:  This is the likely culprit which I'll follow up.  Here is why.  I used LTSpice to calculate the impedance looking into C29 with all low pass component values at specified inductance (not turns) and capacitance.  I was surprised to see 48.0 ohms!  When I increased L3 from 1.1 uH to 1.3 uH the input impedance went up to 75 ohms.  When I decreased L3 to 0.8 uH the input impedance when down to 27 ohms. A couple hundred nanohenry delta made a big difference in impedance. A few hundred nanohenries can happen by squeezing windings.  Conclusion: this is quite sensitive to getting the inductance correct.
 
The next step is to remeasure the as-wound L3 inductance, measure it again while squeezing in and spacing out the turns to get a feel for sensitivity.  Then squeeze and stretch while in-circuit, and if needed modify the number of turns.  Repeat as needed.

Alan de G1FXB
 

Re:- PA / LPF modelling

computer based simulations are only so good....
Real life unfortunately is not always represented and a fudge factor may / may not need to be applied.
In this case, see comments from page 117 of the latest assembly manual. PA device capacitance & Variation according to drive waveform, device conduction Vs. supply voltage etc......


The inductor build manual instructions err's on the side of having too many turns to account individuals own construction variations. it's easier to remove wire than add.....?
If squeezing the turns is not enough then remove a turn and start again from a point of stretched turns.
Add to that, the LPF element characteristics will interact
Throw in the topic of capacitors
Real world is fun.....

It's not an exact science.
However with 7600+ kits supplied, the odds point to tolerance variations either through component or construction variables rather than a fundamental design error?
How did we ever get by, before everyone had desktop access to a sim.?? :-[


Alan


On 29/05/2019 23:05, Paul Seger K7AZT wrote:
Hello All,
??
Thanks to all who have offered suggestions.?? They have been very helpful in eliminating possible issues and in pointing direction for further investigation.?? With your help I'm confident this rig will see air time.
??
Below is current status and my reply to suggestions.?? The brief summary is that the input impedance of the low pass filter is suspect.?? Thanks for putting me on that track.?? I'll look into it and report.????
??
73, Paul
K7AZT
??
----------------------------------
??
Since my first note I did additional testing with an Astron power supply set to 13.95 VDC replacing the battery.?? I remeasured voltages, power, and waveforms throughout the power amp / output filter.?? Scope confirmed Q6 collector is rock solid power when keyed.?? There is no significant difference from my first note with respect to voltages, power, or waveforms.
??
To specific suggestions:

Capacitors C5 & C8: C5 is in place, marked 220, 22pF as specified.?? C8 not in place, as specified.
??
Capacitor Temperature: There was no discernible heating of capacitors due to transmitting.
??
Solder joints: all inspected.?? None cold. No haze. Leads fully enclosed by solder.
??
Scope probe ground point: used BNC connector ground and the paddle jack shield.?? No difference between the points.?? Scope leads are short. Waveforms are cleaner than in assembly manual.
??
Calibration: Calibration went smoothly.?? Receiver sections is really good.?? I didn't expect this level receiver in such an inexpensive kit.?? There were no start up problems of any kind.
??
Transmit Current: My transmit current is low by about 50% compared with others.?? Drains Q1, Q2, Q3 have zero to 33V peak-to-peak waveforms, pulling down near to ground one half cycle, and near half-sine on other half of cycle. This is in line with others.?? I believe the drains are correct, but suspect that the low pass filter is not accepting power.?? See next.
??
Impedance mismatch between power amp and filter:?? This is the likely culprit which I'll follow up.?? Here is why.?? I used LTSpice to calculate the impedance looking into C29 with all low pass component values at specified inductance (not turns) and capacitance.?? I was surprised to see 48.0 ohms!?? When I increased L3 from 1.1 uH to 1.3 uH the input impedance went up to 75 ohms.?? When I decreased L3 to 0.8 uH the input impedance when down to 27 ohms. A couple hundred nanohenry delta made a big difference in impedance. A few hundred nanohenries can happen by squeezing windings.?? Conclusion: this is quite sensitive to getting the inductance correct.
??
The next step is to remeasure the as-wound L3 inductance, measure it again while squeezing in and spacing out the turns to get a feel for sensitivity.?? Then squeeze and stretch while in-circuit, and if needed modify the number of turns.?? Repeat as needed.

Alan G4ZFQ
 

The brief summary is that the input impedance of the low pass filter is suspect.  Thanks for putting me on that track.  I'll look into it and report.
Paul,

As another Alan has said, why should your QCX require a re-design when all the reported cases here seem to show a reasonably large tolerance to LPF variations?

Transmit Current: My transmit current is low by about 50% compared with others.
The RX current is ~120mA so that leaves 130mA for TX ~1.5W at 100% efficiency. Do you ever see more during your varied tests? Unless it increases you are not going to get more power out.

Drains Q1, Q2, Q3 have zero to 33V peak-to-peak waveforms,
Which is somewhat lower than Hans' result. Don't take offence but you are sure of the scope readings?

73 Alan G4ZFQ

Alan de G1FXB
 

Loads of pre and postamble,
nothing of real interest, bar one suggestion to the query "is it the LPF" just jump to the blue text what do you think of?
some perhaps of minor interest to a few........


In addition to the general everyone has access to a desktop simulation package comment & not aimed at any singular post,
the availability of "affordable" spectrum analysers provide the distractions of " it cost $300 " therefore must be correct.
And then being led down the garden path, the latest / greatest just provides more opportunity for error if misinterpreted not less.....

(As the QCX is not ham band only TX limited, as part of it's suite of built in test equipment it's effectively a high level, (output) manually tuned tracking generator in it's own right)
Therefore one is able to easily characterise their own particular PA / LPF build combination, importantly in-circuit with nothing more than :-

1/.?????? A dummy load (*)

2/.?????? A repeatable way of measuring power (*) it doesn't need to be accurate to ANY standard, just indicative & repeatable in your tests

3/.?????? A piece of graph paper (*1)

4/.?????? A little time.


Plot the indicated RFout Vs. Frequency at perhaps 50KHz(*2) increments through 14Mhz and past the band edge, it costs nothing but time and?? is 100% repeatable & dependable,

and valid for however long it remains attached to your termination load.....


(*) the QCX has an inbuilt RF probe, personally to minimise shuffling to and throw menu's, and for the costs involved in obtaining a junk box diode and capacitor I would build an outboard one used with a DVM.
(*) Actually one could use the QRP-Labs own dummy load kit it has a RF measurement option. It's almost if it was designed for the purpose.
(*1) feel free to substitute the old fashioned graph paper for a graph plotting / powerpoint application, it adds gravitas when presenting the findings to the R&D team in a commercial environment should you need to prove them wrong.
(*2) Actually to get good resolution and a smooth curve at the roll off point, one may need to reduce the increments..... just takes more time. Do it once & do it right, the more measurement points the better......

(All unlike circuit simulation or hanging the LPF between a separate none representative Tracking Generator and Spectrum Analyser or employing input / output pads to simulate a fixed impedance to mimic your calculated figures
each provides the potential for different characteristics than that found in YOUR real application..... )
From this you should be able to identify if the LPF shape is a problem, the probability is that it is.
In the PA stage, IC3 is there to guarantee the PA devices are driven into saturation, as long as IC3 has it's 5v and the PA's have their supply through L4. there is little else.
Any discrepancies in gate drive waveforms are most likely as a result of scope artefacts / differences in grounding
Attempting to capture the Drain waveforms is again heavily dependant on scope & probe specification / adjustment, moon phase, etc, etc.....
You are probably unlikely to be able to overlay your trace on Hans example regardless of you perhaps both having the same model scope, just a big ugly signal is probably enough to confirm.


(There is no indication you are using anything other than what was supplied in your early April supplied kit, and using the passage of time as a positive indicator.
There have being no increase in "low power" complaints in the QCX20 or perhaps most tellingly in other previously UN-affected bands to indicate any Hans has changed supplier / bad batch of none bans specific components?
(The gate drive IC needs to be performing as specified, Likewise the BS170's as they are not normally used at RF,?? it would be best to ensure if they should ever require replacement to obtain full specification same branded devices so the device characteristics match Hans calculations and replace in complete sets
ideality off the same tape to have the best chance of obtaining ones from the same batch.))
In the real world, it's reported the tuning of L4?? is many times LESS critical than the LPF tuning, Worst case, is there is no class E "gain" but it will behave as a straight class D stage?
(Checkout the threads where users simply operate their QCX on bands other than that specified using outboard LPF's but yield results.)
again this can be characterised if you feel your actual peak is far removed from the target frequency?
Remember the peaked response at your build frequency is of interest. Actual level measurements are not relevant, they will comprise of the wanted freq and much harmonic energy


That said your 50% low TX current draw & low O/P RF measurements suggests the PA efficiency is far from what it should.
Be it the cause, or a symptom of something elsewhere?.....
(ie. for easy numbers. If you were making the hypothetical full fat 5W(*3) but only a tenth of that was making it's way through a mis-tuned LPF, you are still making the initial 5W
4.5W is lost in the LPF and that would be in the form of heat, most likely noticeable in one or more of the capacitors......
and you could expect the current draw to be in the correct ballpark associated with the production of 5W??
+ You report no heating of capacitors in your tests?



Many builders achieve acceptable results. Everyone wants the book 5W but perhaps neglect to acknowledge that figure is achieved with something more than the " shack standard " 12 / 13.8v supply Vdrop through the series protection diode & Q6 junction.
Perhaps most builders goal should be to achieve a true 3W(*4) @ 20Mtrs on a std shack supply?
Hans is best placed to comment on all this? QRP-Labs techs have supplied pre-assembled kits together with confidence reports of each build.
(Builders of multiple bands is probably common, multiple builds of the same band by the same person and almost guaranteed to be measured on the same setup is probably less so? )
(*4) Remember that RF power calculation requires the knowledge of AC Voltage it can be differently measured as RMS, Pk or PK-Pk, it's why the different, erroneous?? / mis-interpreted equations get flagged up
choose the correct one, as per another recent "how to measure power" thread.

E&OE / YMMV etc.....


Alan

On 30/05/2019 00:23, Alan de G1FXB wrote:
Re:- PA / LPF modelling

computer based simulations are only so good....
Real life unfortunately is not always represented and a fudge factor may / may not need to be applied.
In this case, see comments from page 117 of the latest assembly manual. PA device capacitance & Variation according to drive waveform, device conduction Vs. supply voltage etc......


The inductor build manual instructions err's on the side of having too many turns to account individuals own construction variations. it's easier to remove wire than add.....?
If squeezing the turns is not enough then remove a turn and start again from a point of stretched turns.
Add to that, the LPF element characteristics will interact
Throw in the topic of capacitors
Real world is fun.....

It's not an exact science.
However with 7600+ kits supplied, the odds point to tolerance variations either through component or construction variables rather than a fundamental design error?
How did we ever get by, before everyone had desktop access to a sim.?? :-[


Alan


On 29/05/2019 23:05, Paul Seger K7AZT wrote:
Hello All,
??
Thanks to all who have offered suggestions.?? They have been very helpful in eliminating possible issues and in pointing direction for further investigation.?? With your help I'm confident this rig will see air time.
??
Below is current status and my reply to suggestions.?? The brief summary is that the input impedance of the low pass filter is suspect.?? Thanks for putting me on that track.?? I'll look into it and report.????
??
73, Paul
K7AZT
??
----------------------------------
??
Since my first note I did additional testing with an Astron power supply set to 13.95 VDC replacing the battery.?? I remeasured voltages, power, and waveforms throughout the power amp / output filter.?? Scope confirmed Q6 collector is rock solid power when keyed.?? There is no significant difference from my first note with respect to voltages, power, or waveforms.
??
To specific suggestions:

Capacitors C5 & C8: C5 is in place, marked 220, 22pF as specified.?? C8 not in place, as specified.
??
Capacitor Temperature: There was no discernible heating of capacitors due to transmitting.
??
Solder joints: all inspected.?? None cold. No haze. Leads fully enclosed by solder.
??
Scope probe ground point: used BNC connector ground and the paddle jack shield.?? No difference between the points.?? Scope leads are short. Waveforms are cleaner than in assembly manual.
??
Calibration: Calibration went smoothly.?? Receiver sections is really good.?? I didn't expect this level receiver in such an inexpensive kit.?? There were no start up problems of any kind.
??
Transmit Current: My transmit current is low by about 50% compared with others.?? Drains Q1, Q2, Q3 have zero to 33V peak-to-peak waveforms, pulling down near to ground one half cycle, and near half-sine on other half of cycle. This is in line with others.?? I believe the drains are correct, but suspect that the low pass filter is not accepting power.?? See next.
??
Impedance mismatch between power amp and filter:?? This is the likely culprit which I'll follow up.?? Here is why.?? I used LTSpice to calculate the impedance looking into C29 with all low pass component values at specified inductance (not turns) and capacitance.?? I was surprised to see 48.0 ohms!?? When I increased L3 from 1.1 uH to 1.3 uH the input impedance went up to 75 ohms.?? When I decreased L3 to 0.8 uH the input impedance when down to 27 ohms. A couple hundred nanohenry delta made a big difference in impedance. A few hundred nanohenries can happen by squeezing windings.?? Conclusion: this is quite sensitive to getting the inductance correct.
??
The next step is to remeasure the as-wound L3 inductance, measure it again while squeezing in and spacing out the turns to get a feel for sensitivity.?? Then squeeze and stretch while in-circuit, and if needed modify the number of turns.?? Repeat as needed.

Paul Seger K7AZT
 

Greetings again,

I'm sorry if anyone felt I had suggested a design defect.  I did not mean to imply that.  My concern was that the input impedance of my realized filter was at fault, not the design.  The focus has been on getting the low pass filter correct because other notes suggested this the place to start.
 
Since my last note I was able to get output power up to about 3W (13.95V supply, diode in line) by rewinding L3, L2, L1, pruning, and stretching/squeezing turns starting at L3 proceeding to L1.  Unfortunately the resulting low inductance values produced rolloff between 10 MHz and 20 MHz of only -14 dB (calculated from Vpp across 50ohm dummy load) and calculated as -15 dB in LTspice using the measured values of the final inductors.  It should be closer to -48dB.
 
   Note: thanks for the suggestion to use QCX to sweep its own filter.
 
Today I rewound L3, L2, L1 to the specified number of turns (again) spreading/squeezing. This produces 1.6W.  The measured 10MHz to 20MHz rolloff is -20dB.  Far too small.  The measured inductance at specified turns is about 20% low, as it was in my first build.  Adding turns reduces power greatly.  These numbers don't fit well.
 
I suspect that filter coil turns is not the problem, at least not the entire problem.
 
I'd be thankful if anyone could suggest other areas to explore, such as possible component damage during installation, cooked capacitors in the filter, etc.  Here are some high-level values which might help:
 
The Q6 base and collector voltages are what I would expect.  I had previously varied L4 over a great +/- range, all the way down to 4 turns with little effect.  The drains of Q1, Q2, Q3 clamp within 1 V of ground when driven.  The peak drain voltages are 35-37 V. Current during receive is 0.11A.  Current during transmit is about 0.37A.  Q1, Q2, Q3 get too hot to touch if key down exceeds 8 - 13 seconds.
 
Thanks, Paul
 

Kelly Jack
 

Its been said already but capacitors have been an an issue in a small number of cases.

Have you tried replacing the filter caps?

73 
Simon 
VK3ELH

Joe Street
 

Just a thought but since the reactance inside the transistors is an important part of the overall circuit tuning....what if something is wonky with one of your BS170's?  They are cheap.  I would try replacing them at this point if it was me.  It may not help, but there is a chance.

On Fri, May 31, 2019 at 7:05 PM Paul Seger K7AZT <k7azt@...> wrote:
Greetings again,

I'm sorry if anyone felt I had suggested a design defect.  I did not mean to imply that.  My concern was that the input impedance of my realized filter was at fault, not the design.  The focus has been on getting the low pass filter correct because other notes suggested this the place to start.
 
Since my last note I was able to get output power up to about 3W (13.95V supply, diode in line) by rewinding L3, L2, L1, pruning, and stretching/squeezing turns starting at L3 proceeding to L1.  Unfortunately the resulting low inductance values produced rolloff between 10 MHz and 20 MHz of only -14 dB (calculated from Vpp across 50ohm dummy load) and calculated as -15 dB in LTspice using the measured values of the final inductors.  It should be closer to -48dB.
 
   Note: thanks for the suggestion to use QCX to sweep its own filter.
 
Today I rewound L3, L2, L1 to the specified number of turns (again) spreading/squeezing. This produces 1.6W.  The measured 10MHz to 20MHz rolloff is -20dB.  Far too small.  The measured inductance at specified turns is about 20% low, as it was in my first build.  Adding turns reduces power greatly.  These numbers don't fit well.
 
I suspect that filter coil turns is not the problem, at least not the entire problem.
 
I'd be thankful if anyone could suggest other areas to explore, such as possible component damage during installation, cooked capacitors in the filter, etc.  Here are some high-level values which might help:
 
The Q6 base and collector voltages are what I would expect.  I had previously varied L4 over a great +/- range, all the way down to 4 turns with little effect.  The drains of Q1, Q2, Q3 clamp within 1 V of ground when driven.  The peak drain voltages are 35-37 V. Current during receive is 0.11A.  Current during transmit is about 0.37A.  Q1, Q2, Q3 get too hot to touch if key down exceeds 8 - 13 seconds.
 
Thanks, Paul
 

geoff M0ORE
 

You seem to be very concerned about the 20MHz attenuation which is the second harmonic. You haven't mentioned the attenuation at 30MHz which is the third harmonic.

You also quote attenuation figures in the negative i.e. -14dB, but a negative attenuation is a gain, not a loss.

On 01/06/2019 00:05, Paul Seger K7AZT wrote:
Greetings again,

I'm sorry if anyone felt I had suggested a design defect.  I did not mean to imply that.  My concern was that the input impedance of my realized filter was at fault, not the design.  The focus has been on getting the low pass filter correct because other notes suggested this the place to start.
 
Since my last note I was able to get output power up to about 3W (13.95V supply, diode in line) by rewinding L3, L2, L1, pruning, and stretching/squeezing turns starting at L3 proceeding to L1.  Unfortunately the resulting low inductance values produced rolloff between 10 MHz and 20 MHz of only -14 dB (calculated from Vpp across 50ohm dummy load) and calculated as -15 dB in LTspice using the measured values of the final inductors.  It should be closer to -48dB.
 
   Note: thanks for the suggestion to use QCX to sweep its own filter.
 
Today I rewound L3, L2, L1 to the specified number of turns (again) spreading/squeezing. This produces 1.6W.  The measured 10MHz to 20MHz rolloff is -20dB.  Far too small.  The measured inductance at specified turns is about 20% low, as it was in my first build.  Adding turns reduces power greatly.  These numbers don't fit well.
 
I suspect that filter coil turns is not the problem, at least not the entire problem.
 
I'd be thankful if anyone could suggest other areas to explore, such as possible component damage during installation, cooked capacitors in the filter, etc.  Here are some high-level values which might help:
 
The Q6 base and collector voltages are what I would expect.  I had previously varied L4 over a great +/- range, all the way down to 4 turns with little effect.  The drains of Q1, Q2, Q3 clamp within 1 V of ground when driven.  The peak drain voltages are 35-37 V. Current during receive is 0.11A.  Current during transmit is about 0.37A.  Q1, Q2, Q3 get too hot to touch if key down exceeds 8 - 13 seconds.
 
Thanks, Paul