[minima] Kiss Mixer Analysis

  • From: "Joe Rocci" <joe@xxxxxxxxxx>
  • To: <minima@xxxxxxxxxxxxx>
  • Date: Fri, 7 Feb 2014 14:45:37 -0500

I've been running some LTSpice analyses on the mixer and I have a few 
observations that might be useful. 
  a.. The as-designed mixer seems to have just a little over 6dB conversion 
loss. It seems like 6.0dB is the best you can get using ideal switches, and the 
finite on-resistance of the J301's produces about an extra dB.
  b.. It would seem that best performance can be had when the FET's are 
selected for highest and best matched Idss. 
  c.. It also seems that the conversion loss is lowest when the source bias is 
adjusted so that the positive peaks of the LO gate drive signal are just a bit 
above the source voltage, but not enough to drive the gate-drain diode into 
hard conduction.
  d.. The adjustable source bias circuit can possibly be eliminated completely 
by disconnecting the ground at the center-tap of the LO drive transformer and 
replacing it with a  combination of about 200K resistor in parallel with a 
0.1uf cap, both to ground, grounding both FET sources.This produces a 
self-adjusting gate bias that varies with LO level so that the peaks just 
barely turn on the gate-drain diodes in the FET. A bonus is that you can 
measure the DC voltage at the junction of the R & C to determine that the LO 
drive level is where you want it.
  e.. Initial analysis indicates that the mixer 1dB compression point is over 
+10dBm, so IP3 will probably be somewhere around +25dBm. I will be looking at 
this more closely.
I made these analyses by simulating with an LO freq of 10MHz, an RF freq of 
9.9MHz, and an RC-filtered IF output freq of 100 kHz. As Farhan mentions, the 
use of FET's instead of diodes for mixer switches would seem to have a bonus of 
better mixer termination immunity. In a diode mixer, the on-state voltage drop 
across the diodes is a non-linear function of the signal current. This 
non-linearity is where distortion produces are born. The FET's on resistance is 
pretty linear. This should help keep it predictable when driving wildly 
reactive loads like a crystal filter.

Until I get around to building and testing myself, I'd appreciate anyone else's 


Other related posts:

  • » [minima] Kiss Mixer Analysis - Joe Rocci