I have been thinking a lot about the idea of range switching since reading Thierry's posting on this matter.

Making a 'standard' Theremin, using an equalizing inductor, it is, as Thierry indicates, quite easy to obtain 5 linear-enough playable octaves.. Problems (with this "simple" equalization method) really become noticable when one wants more than 5 octaves.

Until I played the EW-Pro recently, I was not aware that its linear range was limited to 5 octaves.. What Thierry has said regarding Bob Moog overcoming the problem of linearity by limiting the playing field to 5 octaves, makes good sense..

I do not particularly like the idea of having to switch octaves - and much prefer the Tvox approach.. But I have not managed to get better than 5 linear octaves without a lot of complex circuitry, and this translates to an expensive product.

So - I have been trying to work out how octave switching can be done, with the idea that I could make a lower cost version of my Theremin ( I am still doing my Theremin with full linear coverage of > 7 octaves).

I share my thoughts:

!) Assuming the fixed oscillator is at 455kHz, and the variable (pitch)oscillator ranges from 455khz-32hz (454.968kHz) to 455k-1050Hz(453.950kHz) (Giving a difference frequency range of 32Hz to 1.050kHz, which is 5 octaves)..

1.) I cannot see a way to change the register by simply changing the reference oscillator frequency.. for example, shifting the frequency up to 455.032kHz (455kHz+32Hz) would increase the lowest note to 64Hz (one octave higher) - but this would not shift the other difference frequencies proportionally (with the reference at 455.032kHz and the variable at 453.950kHz, the difference frequency would be 1.082kHz, not the 2.100kHz required.. one would simply be adding 32Hz to all the difference frequencies).

2.)I do not see that Multiplying the reference frequency will achieve what is required.. if the reference frequency is doubled to 910kHz, the difference frequency will range from 455.032kHz to 456.050kHz.. even if one was able to get rid of the 455kHz component, octave shifting would not occur.

3.) It seems to me that one must multiply both the reference and variable oscillator frequencies.. And that this will achieve the desired result:

~~~~~Reference~~~Var max~~~~~~Var Min~~~~~Difference Freq

x1~~~455 kHz~~~454.968 kHz~~~453.95 kHz~~~32Hz to 1.05 kHz

x2~~~910 kHz~~~909.936 kHz~~~907.9 kHz~~~~64Hz to 2.1 kHz

x4~~~1820 kHz~~1819.872 kHz ~~~1815.8 kHz~~128Hz to 4.2 kHz

x8~~~3640 kHz~~3639.744 kHz~~~3631.6 kHz~~ 256Hz to 8.4 kHz

x16~~7280 kHz~~7279.488 kHz~~~7263.2 kHz~~ 512Hz to 16.8 kHz

The above shows how multiplying both the reference and variable oscillator waveforms can give switchable ranges, each covering 5 octaves.

It can be seen that for each of the above ranges, the frequency of the variable (pitch) oscillator remains fixed in the range 453.95kHz to 454.968kHz, making the job of linearizing [b]far[/b] easier than if it was required to span a larger range.

Each multiplier circuit could consist of an analogue multiplier with both inputs tied together - the oscillator is fed into these inputs, and the output is double the frequency of the input.. This is the same type of multiplier as used to extract the difference frequency.. except that, as both inputs are the same, there is no difference frequency, there is only the sum of the frequencies - therefore the input (and all its harmonics) are doubled in frequency.

The output of the first multiplier is fed to the input of the second multiplier.. for as many multipliers as are required - and the range switch selects which signals are taken to the final analogue multiplier for extraction of the difference frequency.

So.. For each switchable range (except the lowest 'original' range) two multipliers are required.. to cover a reasonable range (32Hz to 8.4kHz) a total of 6 mult