AnsweredAssumed Answered

Strange results when measuring mixer NF

Question asked by coil on Sep 22, 2015
Latest reply on Oct 18, 2015 by Dr_joel
Hello all, I am preparing to characterize a CMOS mixer I've fabricated using my CXA series SA, but first I am trying to get a handle on frequency converter NF measurements using a simple off the shelf mixer, the ADE-1L from minicircuits. The RF frequency for this will be 65MHz, IF will be 15MHz, and LO will be 80MHz. I've measured the conversion gain under these conditions to be about 5.35 dB (using synthesizers to give LO power pf +4dB and RF power of -20dB), not the noise figure measurement's calculated gain), which is very close to the gain specified in the device datasheet.

My setup right now is as follows:
The LO is driven from a CN0285 synthesizer dev board from analog devices. The LO passes through a 35MHz HPF and then a 3dB attenuator to give +4dBm to the mixer.
The IF port passes to a 7th order 20MHz lowpass filter (to kill the LO feedthrough). I don't have the SA's preamp option, so I use my own external preamp, a HD26316 followed by 10-15dB of attenuation.
On the RF port, I connect the noise source (346A) directly when doing DSB measurements. For LSB measurements I can use a 65MHz BPF to eliminate the upper sideband.

The issue I'm getting somewhat contradictory measurements depending on whether I use LSB or DSB measurements. If I use DSB, I get G=-1.5dB and NF=1.5dB. It's good that G and NF are about equal, but they should be about -2.2 dB (3dB above the measured conversion gain).

If I change the measurement to LSB, without adding the filter on the RF port, I get the exact same results, which is surprising. If I then add the 65MHz BPF on the RF port (and input its attenuation of 2.5dB into the loss compensation) then I get G=-6.0dB and NF=6.9dB. So now the gain is too low, and the NF is much greater than expected. If I back the LO power down by 3dB, I can get G=-5.8dB and NF=-6.3dB. 

So neither measurement gives me a convincing result, even for just the gain. The fact that lowering the LO power increased G (which doesn't happen when I measure gain with my synthesizers) suggests to me that something is corrupting the noise measurements. I've estimated my NF uncertainty to be about 0.4dB, so that's not enough to account for the discrepancy. But I'm running out of ideas on how to track down the source of the error. Every part of the system is shielded, and so far I haven't been able to see any significant changes when shutting off other nearby instruments. 

Any recommendations?