Design of a Gilbert Cell mixer for low frequencies

Thread Starter

Félix Thouin

Joined Aug 14, 2015
9
Hello all!

I am trying to design a Gilbert cell for a low frequency mixing application. I have inspired myself from w2aew's excellent video on the matter but have carried on a few modifications to allow for larger signal and LO amplitudes. I have copied the current version of the schematic below and will describe in this post how I got from w2aew's circuit to this one. The LO oscillator I would like to use is a 1Vpp sine wave at 160Hz and the signal would be around 20 to 40Hz with 300mVpp.

Despite my efforts, I cannot get the circuit to produce a clean output. Hopefully, someone might be able to point me in the right direction.

1621206772933.png

First, I biased the bottom differential pair with an active current source instead of a resistor (Q7 and its biasing resistor friends). This had the effect of balancing the mixer's output between the positive and negative cycles of the signal. Here is what the voltage at OUT looked like before
1621207396577.png
and now with the current biasing
1621207431081.png
I then noticed the envelope still seemed to clip so I added the emitter resistors at Q2 and Q5. This reduced the gain and kept the collector currents nice and sinusoidal giving the following output
1621207580012.png
Now that the enveloppe doesn't clip anymore (while it remains somewhat asymmetrical), it seemed like I just had to do the same trick on the upper differential pairs to get the waveform right. I then added R17 to R20 at the emitters of the upper transistor pairs and expected to get the mixer right. However, I was very disappointed when I saw that the output looked like this when zooming on one envelope cycle.
1621207779174.png

The emitter resistors definitely did reduce the gain and the waveform no longer clips on the envelope, however, significant distortion is observed when the signal crosses the axis. I don't understand what could be causing this effect.

I could reduce the amplitude of the LO to get my circuit to work, but I would like to better understand what is going on here before continuing. Does anyone have a clue what is going on here?
 

Thread Starter

Félix Thouin

Joined Aug 14, 2015
9
I haven't figure out where this distortion comes from and for the sake of time I will simply reduce the amplitude of the local oscillator to 50mV using a voltage divider. The oscillator circuit will be built right next to the entrance of the mixer to minimize the impact of interference on the output. I will also provide good shielding using a metal case.

I am still curious to the nature of the distortion. I'll ask around and if I find the answer I will share it here. I wonder if it has to do with non-linearities in the topmost differential gain.
 

Papabravo

Joined Feb 24, 2006
21,158
It seems like the frequency is low enough and the drive levels high enough to allow the driving signal to look less like a sine wave an more like a square wave. As the circuit is performing a multiply operation, moving the result too close to saturation or cutoff is likely to have an effect on the output. At higher frequencies it is much tougher to get good square waves. This may be part of the reason they are not often used. Mixers want sine waves to do their thing.
 
Top