in an RF mixer chip to create a Dalek sound effect. Later on, I discovered that I had a couple of them
[in the old DIL package, too], so I though I'd do a quick bit of hardware hacking and see how they manage
at audio frequencies. The mixer is based on a Gilbert cell, which will, of itself, work right down to dc,
but there is internal biasing of the transistors, which they don't show in any detail on the
datasheet, so I'll have to ac-couple the signals in.
Here's the equivalent circuit from the datasheet:
Pins 6 and 7 connect to the base and emitter of an internal transistor. In a radio, that would
connect to an external LC network to form a local oscillator. They show several example circuits on
the datasheet. The internal buffer then takes the signal, coming in on pin 6, to the mixing cell
without loading the external circuit too much. The transistor doesn't have to be used, instead it
would be possible to simply couple a signal straight into pin 6.
The signal input is on pins 1 and 2, to a differential pair. Above that the mixing with the signal
from the oscillator takes place and then the output is taken from pins 4 and 5. Although both input
and output are differential, there's no reason why I can't use it single-ended [differential would
be electrically quieter].
What to do for the oscillator? For a Dalek, I need a frequency of around 30Hz. Using an LC circuit,
the components get a little on the large side for a frequency that low, so I thought instead I'd
try a phase-shift oscillator. That will need some external gain - the internal follower doesn't
give us any voltage gain - so I'll need an external transistor as well.
Here's the circuit I ended up with after a bit of fiddling around. It wasn't designed properly:
instead I chose reasonable-seeming values and hacked it about until it worked.
The signal comes from a signal generator [50 ohm]. I've terminated it with a 56R and coupled it to
the input pin with a 10uF electrolytic. At the other input I've placed a 10uF and 27R to ground, so
that the inputs roughly match. For the output, I'm just looking at the pin with the x10
Here it is, built on my wonderful, whizzy, element14 breadboard [it used to advertise some video
channel, but I rebranded it]:
Here are the signals at the LO input pin [blue] and one of the output pins [yellow]. The oscillator
is oscillating just a fraction lower than the 30Hz I was trying for. The signal input [not
shown] from the generator is about 300Hz.
The oscillator is producing something a long way from a sinewave. Unfortunately, because of the ac
coupling, that then ends up off-centre and the modulation is lop-sided. But the multiplier is
working and performing the modulation. It's all rather noisy. The breadboard and my sloppy ground
layout probably don't help.
Here are some FFT plots to show the modulation better.
This is the LO ('local oscillator') at pin 6. The misshapen sinewave leads to a lot of harmonic
distortion. There's also some contamination from the input signal.
This is the signal going in. It's a much better sinewave, but still slightly messy [cheap
This is the resulting output signal. The LO is fairly well suppressed. The sidebands above and
below the signal, as a result of the modulation, are now evident.
If the LO were a perfect sinewave, we'd just see the pair 30Hz away from the signal, but here we
see the LO harmonics contributing as well. For a more complex waveform [voice, for instance] the
output would contain the original signal, the original shifted up in frequency by 30Hz, and the
original shifted down in frequency by 30Hz, all mixed together. That would sound like an ugly, very
extreme chorus effect, but quite inharmonic - any musical intervals would be by accident rather
than design: just what you want when you're an evil cyborg intent on galactic domination [or for
scaring young children hiding behind a sofa].
Update: 6th March
I was curious about the lop-sided modulation and wondered if what I said about being
able to use it single-ended, rather than differentially, was actually true. The
datasheet suggests it can be used like that, but at the same time they do seem to steer
you to differential coupling of the input and output.
It occured to me that one way to test that out would be to feed the same signal into the
LO input as the signal input. The multiplier should then give the square of the input.
For a sinewave going in, that would result in a sinewave of twice the frequency coming
Here's the result [blue input, yellow output]
You can see we have the frequency doubling, but the waveform is, again, lopsided, so the
chip certainly doesn't balance very well internally.
Here are the two, differential, outputs on the 'scope together.
If I ask the 'scope to calculate the difference, like this
we see a sinewave, so it all works fairly well if we take the output
As a further experiment, I added an external 100k pot to pull the biasing of the
transistor around. With a bit of adjustment, that then allowed me to get a reasonable
modulated signal out, but only on one or other of the outputs. This trace shows the
input (yellow) and an output (blue), with that output adjusted for best sinewave.
But if I look at both outputs together I see this. Pulling the biasing of the transistor
never results in both coming right simultaneously.
Although this was just a bit of hacking for fun, it was interesting to see how easy it was to
explore rf techniques, like modulation, at low frequencies using an inexpensive radio chip and a
modest oscilloscope with an FFT maths function.
For an actual effects unit, a variable, external oscillator might be a better choice than trying to
build something around the internal transistor. Some sort of gain control at the input and output
would also be useful (essential?): the differential amplifiers that make up the cell are only good
for a few hundred millivolts before signal limiting sets in but, at the same time, the levels need
to be kept reasonably high because of noise.
If you found this interesting and would like to see more blogs I've written, a list can be found
here: jc2048 Blog Index