Is it possible to measure some LNA's NF manually using poor man slide screw tuner?

I'm trying to measure optimum NF of MP4T6365 manually using Maury 1643D screw tuner. Manufacturing Data sheet doesn't show Gamma opt for opt NF. My set up is as followed

Noise Source -> Bias Tee -> Screw Tuner -> Dut -> Bias Tee -> HP-8970A
f=1GHz

I can calibrate with everything in with the tuner at default 50Ohm but when I tried to tune the tuner, my reading is always as high as 2.8dB (supposed to be around 1.3dB). Use 3dB attenuator padding at input/output won't help. I think the tuner introduces different loss at different setting so it throws the measurement off. Can you guys show me the way to do this thing correctly (besides using an automated tuner system)?

Thanks
Lanny

Stub tuner loss depends on impedance transformation settings but unless it is a really bad tuner should not be 1.5 db.

If tuner is old and dirty it might have excessive loss.

Accuracy of NF measurements are very dependent on noise takeover to NF meter or spectrum analyser (for Y method).

Noise figure meter is about 1-3 MHz bandwidth and 4-6 db NF. The narrower the bandwidth of the DUT the more noise takeover is required to prevent the NF meter's wide noise bandwidth from degrading NF measurement.

Look at noise takeover when noise source is turned on and off into a spectrum analyser (Y factor method). With a spectrum analyser you can take reading at narrow bandwidth. Your LNA needs to create at least 10 db of narrow band takeover on spectrum analyer (without noise source on) to get accurate Y factor reading. Since a typical spectrum analyser input is >20 db NF you need significant gain in front on spectrum analyser. This can be accomplished with low noise amps between your LNA and spectrum analyser.

Another common issue is RF pickup from environment causing spurious signals in passband of NF measurement. Using a NF meter should be done in a shielded box or room.

Using a stub tuner and a HP8970A to find the gamma optimum (this is actually what you are looking for) should be fine.
A good quality stub tuner has very low insertion loss at 1.6GHz, which is the maximum frequency of 8970A. No more than 0.3 dB.
Use a calibrated network analyzer to measure the real insertion loss which you have to take into account during noise figure measurement.
Also the network analyzer will tell you what is the tuner impedance when you find the gamma optimum.

Thanks RC & vfone. Now I recalled why I switched major and didn't want to be an RF engineer way back then, when things was done mostly manually.
Look like I have to setup a coax switch to facilitate my measurements.
Is there a quick way to home in the gamma optimum so I don't have to shoot blank in the dark because theoretically it could be anywhere in the smith chart?

PS: My tuner loss is very low (like 0.05dB) at 50 Ohm (the probe is fully retracted) and I still can't get anything lower than 2.8dB.

If the tuner loss is low means you get the extra loss due to the Missmatch Loss:
Missmatch Loss[dB] = -10*LOG[1 ? Γ2] where Γ = (VSWR ? 1) / (VSWR + 1)

Most of the cases input return loss (or VSWR) will be sacrificed to get the best noise match.
Tune the stub tuner until get the best noise figure, do if is necessarily a compromise for better gain, measure the resulting impedance of the tuner using a VNA, and design the matching network using a Smith chart software.

Try putting an additional broadband low noise amp following your DUT.

RC,
I don't think I need a pre-amp since my NF from 1st cal sweep is considered exceptional by HP-AN 57-2.

----------------------------

I guess my problem probably caused by the test fixture affecting the tuner's tuning range and depending on gamma opt of the DUT, some can be tuned and some cannot unless some sort of impedance transformations are added between the tuner & test fixture to move the tuner back to the tunable range. This has been verified by tuning an AT4270 with known gamma opt at 1GHz. I got it down to 1.7dB (1.5 typical) considered the cal path doesn't include anything in front of the DUT (tuner loss, N to SMA adapter, test fixture loss at input)

I still have a hard time to tune the unknown gamma opt transistor above. I did manage to lower its NF somewhat with some impedance transformation but still no cigars.

It works after I add an ARRA phase shifter after the tuner and in in front of the DUT.