I would like to use an Analog Devices MLT04 to multiply a complex
subaudio signal by 1KHz.
The result will be viewed on a spectrum analyzer.
What I am hoping to see is relative amplitudes of the major frequency
components within the subaudio, transposed by a factor of 1,000.
Can anyone tell me if it matters whether or not I use a 1KHz square or
sinewave for this purpose, and, if so, why?
Many thanks,
Terry Osbourne
What's normally done in a dynamic signal analyzer (aka FFT spectrum
analyzer) is to digitize the signal, and then multiply digitally by
sine and cosine to shift the frequency. Do a search for "image
rejection mixer." Of course, since you want to look at LOW
frequencies, you wouldn't need to do that. If you can simply digitize
the signal in its frequency range, you can use an FFT to look at the
spectral components. You may wish to use windowing to keep artifacts
from the abrupt edges of the sampled block from spreading throughout
the FFT'd spectrum.
If you want to use a multiplier to move the spectrum up to a range
where you can look at it with a spectrum analyzer that won't go to low
enough frequency, you can do that, but be aware that just multiplying
gives you both sum and difference frequencies. If you use a square
wave, if it's a perfect square wave, you'll have images of the spectrum
at the (say 1kHz) fundamental, and at all odd harmonics (3kHz, 5kHz,
7kHz, ...) If your signal of interest is narrow-band enough, that
won't matter. You can insure it is narrow band enough if you
low-pass-filter it before feeding it to the multiplier. You need to do
that anyway, so that other frequency components don't alias into places
you don't want them. If you use a square wave, you might as well use a
frequency mixer instead of a multiplier. They tend to be cheaper. If
you decide to filter a square wave to make it sinusoidal, use a low
pass filter that has steeper cutoff than Phil's suggested 18dB/octave;
that won't attenuate the 3rd harmonic much at all. It can be easier to
just use a bandpass filter for a single fixed frequency. You can also
just filter the OUTPUT of the mixer/multiplier. But again, if the
signals of interest occupy a narrow bandwidth, it shouldn't matter.
You can just let the spectrum analyzer get rid of the unwanted bands,
since they are far away from the interesting stuff.
Cheers,
Tom
