The name of this module is a play on its function. Put simply, it generates a series of semi-random or themed stepped control voltages, or if you prefer, white and pink control voltages. The pink function is probably better known as 1/f, and thus the name of "1l/lfinite melody".

Assuming a 1 v/oct VCO is being driven by the module, in the 1/f mode, the size of a pitch step is inversely related to the frequency at which it will occur. In other words, the smallest steps are most common, steps twice their size occur half as often, and so on.

As well as the 1/f mode, it has a pure random mode, where no weighting is put on the selection of notes. A large step is as likely as a small one.

As well as having these two modes of operation, there are 4 CV outputs available. With the exception of the fourth, each successive output is a shifted version of the previous. In random mode, this is very much the same effect as produced by an Analog Shift Register (ASR), but in 1/f mode, it takes on a whole new feel, as each of the six shift registers involved shifts at a different rate, according to the 1/f weighting. As such, the second and third outputs are related to the first, but not identical to it. The fourth output is subject to the same shifting weighting according to the mode of operation, but instead of driving a D/A converter, it drives a bank of knobs, allowing harmonic relationships to be set up.

And if that wasn't enough, the randomness used by this module is obtained externally, from a regular white noise generator, or other varying voltage source, allowing more structured themes to be created. The random values are loaded in series at a rate determined by an external fast clock. This can simply be a spare output of a VCO (even one in use playing melodies etc.) or it can be deliberately controlled, or slowed right down to gain even more effects. The sensitivity of the random input can also be controlled via CV, again giving more possibilities.

See Diatonic Converter for an alternate output for this module.

A little on how it works:

The schematic of the Infinite Melody.

First, a very simple theory lesson - 1/f implies a weighted ratio between a level and its corresponding rate. A change in level of 1 will occur at the clock rate. A change in level of 2 will occur at half the clock rate, a change in level of 4 will occur at a quarter of the clock rate, etc., thus giving rise to the 1/f ratio. This means the larger the step in output level, the less frequently it will occur. On the other hand, random implies exactly that - there is no determination of how great a step will occur - any step size is as likely as any other.

As is shown in the schematic, the Infinite Melody consists of several distinct sections.

The first section is the clock and random or 1/f selector. The clock and reset signals are processed by the LM358. When the 1f/Ra input is LOW (below about 2V), the reset line of the 4024 will be held LOW, allowing it to count, and clock signal connected to the second input of each EXOR gate via the 10k resistor will be blocked by the forward biased diode. The output of the counter is then fed via the EXOR gates (inverted in the process) to the clock inputs of each of the 4015 shift registers. Each successive shift register receives it's clock at half the rate of the one above it.

When the 1f/Ra input is HIGH (above about 2V), the reset line of the 4024 will be held HIGH, forcing all of the outputs of the 4024 to LOW. The clock signal connected to the second input of each EXOR gate via the 10k resistor will pass as the associated diode is now reverse-biased, and no longer blocking the signal. The clock signal is now fed via the EXOR gates (inverted in the process) to the clock inputs of each of the 4015 shift registers. All shift registers are now being clocked at the same rate.

The second section of the schematic is the random number processor. Each shift register in the chain requires a random level to be present at its data input when it receives its clock signal. Traditionally this would be achieved by having a series of six independent random number generators. Instead it was decided to clock random data in serially at a rate significantly higher than needed. This section is essentially identical to the Gated Comparator, so I suggest you investigate that article for more information on how it works.

The block diagram of the Infinite Melody.