U3S Ultimate3 and 5WPA experience

Sorry enough  after just a  few day's testing the U3S / 5WPA  at 50 / 28 / 21  Mhz  
Output was lost  after 15:42Z  Monday 17-4  
The FET RD15HVF1  died  possible by statics?
1 genuine left but repairing this 5W PA  ?
A lot of effort and trouble  to bring 1W to 4W at 28MHz 
The U3S single BS170  already is 1W at 5V!
I thought that it would be an easy job to have 5W+ with the 5WPA RD15HVF1
But no way! something must be wrong somewhere the power gain is 0 - maximum 6dB only
The 5W-PA is good for 10MHz and lower but very disappointing at higher bands.

The signal from the U3S straight from the Si5351a needs to be filtered before it drives the RD15
and after the 5W-PA again i think 
Far from sinus shaped signal at the output of the PA behind the LPF filter. 

Be aware! the fake RD15HVF1 is still offered at eBay  5 for 5$
It was only my 2nd bad buy at eBay  of many purchases.     
After solder tip eating  SOLDER core wire CF-10 B-2  0.6mm  
that smells like cat pee. 
The good  the  bad and the ugly.. 
Left genuine Mitsubishi,  mid Ebay "good quality RD15HVF1"
what appears to be an   IRFxxx  G-S = 1.29nF    re-stamped type 
pins:  GDS instead of GSD 
seller alice  
right  the IRF510 (G-S = 400pF )  good but low power gain at 28MHz

Frequency drift is near zero using  TAITIEN TS-27.000  VCXO  from DF9NP  http://www.df9np.de/
Bingo! from cold start 50.294500MHz  was decoded with 0 drift. for the first time.

Modified an 2015  version synthesizer Si5351a to the Taitien VCXO as reference.
It fits on a tiny 6x3 pad experimental the 5k trim-pot is for VC  pin 1
This 2015 Si board has no option for TCXO 
Used tiny wire pieces to connect 3V3 gnd  and RF out
SMD TCXO connections seem to be a standard pin 1= VC or nc/gnd  pin 4 =gnd  pin 5=3v3  9=rf out
This Taitien TS type will need an adaption for use on the V4 TXCO prepaired PCB  it needs VC at pin1 around 1.5 V for trimming the 27.000000 MHz
The pads match with  4 pin type SMD TCXO   this Taitien has 10 connections 4 are used 6 nc
It is 7x3 mm  while the pads are ment for 5x3.2 mm TCXO types 

some reception report received this afternoon   PA3BHW 
many decodings by 9H/OZ1LXJ   IZ1ANK IZ2JNN OZ7IT at 28MHz  
and  50.293MHz OZ7IT  DK8NE PE1MBR  NL8992 PA3BHW  mostly 0 1 Hz drift   4Hz maximum
before with the FOX924 a 6m decode was rare . The frequency drift of the FOX a little more mostly 4Hz +
But the real problem with the FOX is the disturbing frequency correcting steps.

Also 144.490500 0 or 1 Hz drift.  That's how a real TCXO should perform!
No missing decodes , the WSPR signal quality is good with Taitien VCXO as reference.
The FOX924-27 as reference resulted in almost 100% fail decoding at 144MHz
The CAL correction in the U3S can be disabled
With the trimmer the exact frequency of 27.000000 can be adjusted.
Next time will rotate trimmer position 90dgs, to be better accessible.  
50.294500 MHz  is  mid WSPR window  
144 is a bit too low in the window, should be in the center too 
checking the IC706  with  144.90000 MHz GPS locked ADF5351a  
showed that the IC706 is still accurate at frequency mid window  at 144.489500 
so that puzzles me now a bit why not exact center at both 

I left the working Synthesizer V4 PCB with FOX924-27 TCXO as it was.
This Synthesizer PCB's have to be treated very careful 
Now 2  defective Synthesizer PCB several repair attempts failed. 

Finaly i can be almost.. satisfied with the results.
It runs 4 to 5W at  6m 10m 12m 15m  .
(using LPF  6m for 10m  and 12m LPF  for 15m)
At the low bands more output:  from 6W to 11W  30m 40m 80m 160m
using the 30m LPF for 40M
at a conservative 12V supply 
The Ultimate stays relative cool at a perfect load  but it will heat up more  if the SWR rises 
 additional cooling is needed when used in warm environment.
                                           RD15HVF1 mounted to bottom with CU plate
CPU cooler and external FAN at TX 12V  out plug for hot evironments

                                             LCD scrambled RFI issue  annoying but does not disturb signal 

13V  gave 5W at 6m   but  4W on 10m selecting its designated 10m LPF   
The 10m LPF must be at the wrong frequency range 
no visible faults   could be  spreading of Torroïds  
6m LPF for 28MHz  gave the expected 5W  at 13.5V
28MHz still marginal interferes the LCD display 
100nF over the LCD 5V - gnd  did not solve it 
Using the 6m LPF improved the output and lowered the display RFI
28 and 21 MHz causes display distortions but the U3S functions normal.
LCD RFI is stronger with case closed.

Remarkable 12W  at 160m  13V   11W at 80m 
low bands  a bit too much for WSPR use
 The 5W PA  is now used without the power modulation option 
An extra  IRF9540 is TX 12V switch with the key signal
The BS170 gate stopper is replaced by a jumper 
It would have been a nice feature to have this BS170 as software adjustable drive level controll
Now the power is descending strong 11W at 160m to 4W at 6m

The PA  is now simplified the whole power modulator DA -converter circuit is
obsolet  for WSPR
 Reduced to 3basic parts  a bias controlled RD15HFV1 
  and the T1 1:9 output transformer

My Ultimate wish list:

1. Higher efficiency  possible class D like at low bands 160 80 40 must be possible
      For  24*7 beacon operation 
2.  Software power regulation low loss per band at the driver side.
     to equalize the output over the used frequency range.
3.  Fix LCD annoying display RFI.
4.  In some boot situation after a short power down the GPS starts it's pps and NMEA 
     is present but the time is not updated and keeps counting.
     Then the TX start time is out of sync and triggered by the random power up of the timer.

Tried with the BS170 as a Source follower to drive the RD15HVF1    
I used IC socket pins for this BS170
Swapping gate and  drain it simple becomes a source follower 
No success at least not at 28MHz and 50MHz
Turning up the bias seems the only way  to get  5W out at 50MHz 12V

T1 modified to 8 turns also did not increase the 10 and 6 output
T1  FT50-43  heaths up     5-10% loss estimate 
Improvement of efficiency is still possible.

                                          12V    about 4W at 50MHz

MW 475 kHz   last night first T.A. decodes since weeks  the last USA i saw at 25 March

 2017-04-15 04:10  WG2XPJ  0.475649  -23  0  FN34lp  1  PA0O  JO33hg  5587  50 
 2017-04-15 03:24  WG2XPJ  0.475649  -26  0  FN34lp  1  PA0O  JO33hg  5587  50 
 2017-04-15 04:02  WG2XPJ  0.475649  -21  0  FN34lp  1  PA0O  JO33hg  5587  50 
 2017-04-15 03:28  WG2XPJ  0.475649  -27  0  FN34lp  1  PA0O  JO33hg  5587  50 
 2017-04-15 04:06  WG2XPJ  0.475649  -21  0  FN34lp  1  PA0O  JO33hg  5587  50 

btw  this new blog layout is missing the login / edit link i can't find it and use a history link to be  able to modify.
Same time the spelling NL DL check disappeared, only Brexit language check is left. 
Always a surprise what you get next day, no future guaranties for a working PC or Wlan system.