Relation of Mixer compression point and LO power?

mixer compression

An very senior Engineer (20 years in RF design)told me the mixer compression point often 10dB below LO power, so the higher the LO power the higher mixer
compressin point. I am wondering what I realize before is the LO power is prohibited to be too high due to harmonics, this also causes mixer linearity degraded. I guess his experience may come from some histroical reason, but
I dont' know the detail of it? anybody has the idea?

To my experience is true, I have used and I am using many +7 dBm mixer and I have found the compression point around - 1 dBm on the RF port.
If you need a couple of dB more you can increase the LO drive up to +10 dBm, not more othrwise you have to change mixer.
If you need Higher compression point you have to go to + 13 dBm, + 17 dBm or more LO drive mixers.

Mandi

What kind of mixer configuration do you use? Active or Passive?
Do you have any idea of the reason behind this phenomenon?

A passive mixer is the basic building block of a mixer. In it, the LO power is dominant and is responsible for switching the devices (Diodes or MOS's typically). As long as the RF power is weaker then the LO, the performance does not change in that the LO signal fully switches the devices, causing the RF signal to be chopped in some manner.

When the RF signal starts approaching the LO power, it starts to affect the device behavior. Instead of the RF passing through switches that are set by the LO polarity, the RF now starts to change the on-resistance of the switch. This starts happening around when the RF power is 10dB away from the LO. As the RF power rises the harmonics increase until we get compression.

Now lets move into the active mixers such as a Gilbert cell. The active mixer can be split up into 3 sections: A passive mixer, an RF amplifier and an LO amplifier. In a Gilber cell, the RF amplifier is the transconductor feeding the commutator (typically the 2 transistors at the bottom that turn the RF power into a current). The LO amplifier is the commutator itself because those devices (MOS or BJT typically) are biased up, hence the gate/base junction amlpifies the signal feeding into the emitter/source. The cummutator is the group of 4 transistors that switch the polarity back-and-forth comming from the RF amplifier. Finally, the passive mixer is also the commutator but as seen on the source/emitter side, offering no gain but simply a through switch behavior. So the passive mixer side behaves just like described before, but the active sides need to be additionally considered.

Now if the RF amplifier gives similar gain as the LO amplifier, then the rough 10dB rule can still be preserved. If one amplifier gives much more gain then the other, then you can change this rule of thumb, as seen at the input of the entire system. For a Gilbert cell, the current seen through the RF amplifier is really the same as the current seen through the commutator because only 2 of the 4 commutator transistors are on at any given instance of time. For BJT's, gm=IC/VT, since Ic is the same for all on transistors, we can expect a very similar gain for both the LO and RF ports. Therefore, the mixer's 10dB rule should roughly be preserved. Notice that the transconductor does NOT affect the gain of the commutator (to a 1st degree) because the LO sees a virtual GND where the commutator transistors meet (emitter/source side).

Hope this makes sense without having diagrams. On modern integrated mixers, the LO goes through many levels of amplification and limitting in order to minimize the LO's amplitude effects on the mixer's performance. The 10dB rule therefore is not observed except by pure coincidence. Typical IC's have a 10dB LO input range, yet the compression point is not affected by the LO amplitude.

Greg