EMRFD Message Archive 780

Message Date From Subject
780 2007-05-21 16:18:11 Kevin Purcell Giving the Gilbert cell mixer a low input impedance
I find Google patents to be an interesting place to browse. The
collection seems complete but (Fessenden's AM patent from 1906) but
beware the complete text indexing only seems to go back to 1990s.
Before that the meta data is searchable but sometimes incomplete or

For Gilbert cell mixer you can find Gilbert's patent


and some modifications.

One of these US patent 5515014 shows some interesting 1496 hacking in
the patent literature


The trick is in changing the bias of the constant current source
transistors by taking their bases to ground and reverse biasing their
emitters to +3v so they're fully switched off.

The high (medium) input impedance common emitter amplifiers are
turned into low input impedance common base amplifiers by taking the
bases of the lower pair of transistors to signal ground (through a
decoupling capacitor) and adding external emitter resistors (pin 2
and 3) to ground. These resistors also set the standing current for
the mixer tree. The differential input signal to the mixer is fed
into the emitters at the "gain adjust" pins 2 and 3 on the 1496.

In the particular version he shows he uses 2kohm emitter resistors
with a standing current of 1mA (per emitter -- 2mA total) giving an
differential input impedance of 50 ohms (2 x 25 ohm). Of course to be
interesting you would need to increase the standing current (to get a
decent IIP3) and so perhaps need some impedance matching from 50 ohms
to very low Z.

This would be useful for matching transistor tree/Gilbert cell mixers
to filters that terminate in a low Z impedance e.g. SAW filters in
the example but one could imagine using this with Cohn filters and
other low Z crystal filters.

With a ground base input amplifier I suspect it also might improve
the mixer to antenna isolation making use of the grounded base's good
reverse isolation. And it improves the bandwidth of the circuit too
(by removing Miller capacitance coupling the input and output in the
common emitter topology).

One of those interesting tricks (which I don't think is worth a
patent -- it worth a Technical Topic or an note in EDN) that I
believe I might have come up with (but probably wouldn't have!).

But the patent is well written and a nice intro to the Gilbert cell.
Kevin Purcell