Already replaced the U310 once, just in case I cooked or zapped it in my experimentation. I have a J310 that I could also try.
Bias current is set for ~15mA, closer to 12mA once the part heats up since my “current source” is a resistor.
I’d gladly test another unit if someone cared to build one and show me what I did wrong. 
I’m going to set this one aside and build a different design using a common-drain design. This has the disadvantage of being slightly more noisy (higher NF).
A great paper on noise, and why an LNA is important when listening to signals close to the noise floor:
http://www.w1ghz.org/noise/noise99.pdf
He clearly explains the concept behind the Friis Noise Formula.
Could the resistor be the problem?
What does spice predict as the power disapation?
I added a couple of parts to my latest Mouser order to let me build one. We will see how it does.
BTW, solved the problem with the audio amp for my SSB transceiver build. Basically I was an idiot.
I even changed that out… (I had to, since the bias was different between the two)
All three times, they were your standard carbon composite.
U310 is dropping ~10 volts at ~13mA on the current iteration, giving 130mW. The first one was getting a bit hot for my liking, so I put this one upside-down on the board, in contact with the copper ground plane. My PIR meter says the board isn’t getting much above 40C.
Based on your recommendation, I bought one of those fancy DE-5000 LCR meters. I’m impressed so far! The wound coils measure 103 nH, which is what I’d expect. Q factors anywhere from 20-34, depending on how close the windings are compressed together (probably due to varying capacitive coupling).
I would never use a carbon resistor in anything intended to be low-noise. Metal film might run you another ten cents, but worth it.
Yeh, its a great meter. Since it uses Kelvin measurement the probes can be upgraded…
That is likely way too much for a J310…
I think your problem is that you have it mounted upside down. The gravitational effect on a Chinese transisters beta is known to be asymetrical. 
Depends on the application; in this case, the variations in the DC bias current of the transistor are not multiplied by the gain of the device, so they are below the “magic” 10dB line where that current noise would contribute. It is also bypassed by the 220pF cap at HF and VHF frequencies.
I agree that metal film would be much better for precision DC or near-DC circuits. Analog Devices AN-940 has some useful insight on this, too.
ON Semi claims 350 mW at ambient 25 degC.
LOL
Ah, but was that claim written by their engineers or marketing folks?
If the U310 metal can was hot to the touch, the to92 plastic of the J310 would be a bad radiator of that much heat, IMO.
Don’t forget that even though the package might burn you, the silicon is happily working. It may not be the ideal situation for you, but often it isn’t a problem.
The rated junction temperature goes up to 125 C which will register as a lot worse than hot!
Bingo! It was the trimmer cap (the white one) … I swapped it out for a Murata red-colored one (4.2-20pF) and was able to find the +12dB peak gain point.
If I were to re-build this, I’d invest in a piston trimmer to make it a bit easier to align.
What was the cap value when you achieved the 12dB gain?
Was it the value of the cap or something to do with the construction of it that caused the issue?
Two problems conspired to make it hard for me to find the gain peak: the trimmer cap I was using had a range of 4.9pF to 75pF in one turn, and the Q of the output match circuit is fairly high (narrow b/w). Using a non-metal tuning stick, of course, I wasn’t able to get the peak.
I swapped out the cap with a 4.2pF - 20pF one turn model and, lo and behold, I could get the peak!
I’m planning to swing by Tanner Electronics today to get a 5-20pF piston cap, which should it far easier to align it.
The input tuning appears to be much broader and easy to align.
Here’s a good reference on Resonance, Q, and matching for those who need a refresher (or an intro).
Dunno, as it is currently soldered in, but simulation predicts 9pF. That’s in the range of the 1-10pF trimmer specified in the schematic.
Ah, I thought you were using the 3-15 pF cap in the schematic and needed more range.
Nope, too much cowbell!