How To Understand the NE-602 and the Gilbert Cell Mixer

 

I think the key to understanding the Gilbert Cell Double Balanced mixer is to separate out the three tasks that this device completes, and consider them one at a time, using different diagrams: 

1) It mixes two signals to produce sum and difference outputs. 

2) It balances out the RF input. 

3) It balances out the LO input. 

                                                                  Task 1 -- Mixing

The Gilbert cell is like the diode ring mixer in that it switches the polarity of the input signal at a rate set by the Local Oscillator. Another way of saying this is that the mixer multiplies the input signal by 1 and by -1. 

Steve Long of the University of California described the essence of this mixing this way (using the diagram above): 

 

An ideal double balanced mixer simply consists of a switch driven by the local oscillator that reverses the polarity of the RF input at the LO frequency.  http://literature.cdn.keysight.com/litweb/pdf/5989-9103EN.pdf

In an effort to see this for myself, I drew (noodled!) this diagram: 

There are four transistors -- two differential pairs with RF coming into the bases of the pairs. 

The LO is a square wave.  The LO alternately turns on transistors 1 and 4, then 2 and 3.  When 1 and 4 are on, we are in period 1 -- here there is no switching of polarity.  Portions of the RF waveform are passed to the outputs.  But when the LO turns on transistors 2 and 3, portions of the RF wave form are "crossed over" to the opposite output.  Polarity is reversed.  We see this in period number 2. 

Take a look at the resulting output waveforms.  This is the same waveform we see coming out of a diode ring mixer.  I really like this drawing because in that complex waveform you can actually see the sum and difference frequencies: 

I could see this diode ring waveform myself on my oscilloscope: 

https://soldersmoke.blogspot.com/2020/11/diode-ring-magic.html 

TASK 2 -- Balancing Out the RF Input 

In a diode ring, and in other diode mixers, the balancing out of the input signals really takes place in the trifilar toroidal coils that are part of the circuit.  Barrie Gilbert needed an integrated circuit mixer that did not use coils.  

Again referring to the above diagram, Steve Long of the University of California put it this way: 

The ideal balanced structure above cancels any output at the RF input frequency since it will average to zero.

To fully understand this I find it helps to look at the Gilbert cell circuit drawn in a different way.  Here is a drawing from Alan Wolke W2AEW that I found very helpful. It comes from his excellent YouTube video: https://www.youtube.com/watch?v=7nmmb0pqTU0

Suppose the RF waveform at I1 is causing the current through R1 and R2 to increase.  At the same time, the opposite phase current through I2 will be causing the current through R1 and R2 to DECREASE.  So there is no net effect of the RF signal at the output.  The RF is balanced out. 

TASK 3 - Balancing Out the Local Oscillator Signal 

Here too I used my own drawing, and was guided by the words of Steve Long: 

It also cancels out any LO frequency component since we are taking the IF output as a differential signal and the LO shows up as common mode.  

The important thing to realize here is which transistors are being turned on and off by the local oscillator signal.  On one half cycle of the LO, transistors 1 and 4 are on.  So  the LO signal at the LO frequency are both pulling the same amount of LO frequency current through the resistors. So you have the same change in voltage at the output terminals.  And the output terminals are differential.  The LO signal results in no voltage difference between the terminals.  So the LO frequency is balanced out. 

The same thing happens on the following half of the LO cycle.  Here, transistors 2 and 3 are turned on. Again, both transistors pull the same amount of LO frequency current through the resistors. There is no differential voltage.  So no LO frequency energy passes to the output.  LO frequency is balanced out. 

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

I am surrounded by Gilbert Cell Mixers and I have been using them in my homebrew rigs for many years. I use them in up-converters for my RTL-SDR receivers.  I have one in the downconverter for my 17 meter receiver and had one as the mixer in my first SSB transmitter. I built a 40 meter SSB transceiver with NE602s on either end of the crystal filter. Years ago, I built a DSB transceiver with several NE602s.  My SST QRP CW transceiver is made with NE602s. I have on my bookshelf Rutledge's book "The Electronics of Radio" that is all about the NORCAL 40 transceiver, built using NE602 chips.  But until now I really didn't know how these chips worked.  Truth be told, for me they were mysterious little black boxes, and that bothered me.  Now I feel a lot better about using these clever devices.  I plan on stocking up on the old style (non-SMD) NE602s.  

Apparently Barrie Gilbert rejected the idea that he invented the circuit that bears his name.  It seems that Howard Jones first used this circuit in 1963, with Gilbert developing it independently (in an improved form) in 1967. 

 Barrie Gilbert was quite a guy, with electronic roots in the world of tinkering: 

https://www.microwavejournal.com/articles/33425-living-and-learning-with-barrie-gilbert 

https://vintagetek.org/barrie-gilbert/