I can recommend where to place connections between bottom and top mass planes, as i got the bad PCB working now. Try to keep the hole in each via open for better RF connection.
Make vias :
Close to the border ofthe PCB:
Under : R18, C21, U3, U2, R15, D4,.
To the left of : C3 and D1.
Solder the output connector at the bottom side.
Directly in the cooling fin island of each MMICs.
Directly at the mass connections of C11, C16, C17, C8.
At the left of U1 (mass).
Better solder 100uF/16V capacitors in parallele to : C11, C16, C17. C8.
Solder 100uF / 63V in parallel to C1.
This helps preventing noise below 200kHz from an external SMPS supply and the 38V converter U1.
I also short circuited L6, L7, L8 with a wire. This makes the frequency characteristic less hilly in my experience.
Try changing R11 into 82 kOhms (yes eighty two kilo ohms). It gave me a flatter frequency characteristic.
Check the difference.
My PCB is now still more changed so it is all powered from 12V :
U1 is now disconnected from 12V, and R11 removed.
R15 is short circuited.
R17 is removed.
R19 is now 120 Ohms.
D4 is 4.7V 500mW. D4 gets 37mA via 100 Ohms from a 78L09. Its output is decoupled to maas by 1nF + 10nF + 100nF + 100 uF. The 1nF is important for high frequency characteristic, hold wires short.
Output -50dBm @ 3.7 MHz.
Flat from 25 kHz to 3.2 MHz +/- 1dB.
50 MHz -19dB.
Flat from 60 MHz to 160 MHz within +/- 1dB.
Power : 12V 280 mA. Tested wit 14.6Vdc 320mA, no break down.
NO hand effect anymore when near the border or undeside of the PCB.
NO oscillations anymore.
NO unstable (jumpng) noise output.
When in a box, then metal, and a lot of ventilation holes. One ground connection to the box at the output connector.
But i designed my own new PCB. Must save money for odering it and make a new test version.
I removed this topic from my dite as it is to much work for correcting a junk PCB.
Please report me your findings. Maybe also interesting for others.
73’s, Nico pa0nhc. www.pa0nhc.nl