This 4 bit digitizer originally appeared in my Lab Notes column in Polyphony magazine. This type of analog to digital converter is usually called a flash converter because conversion happens asynchronously, with no clocks or strobes. Input a voltage and the digital output appears in a flash :-) This design is fast enough to digitize signals into the low audio frequency range, less than 1kHz. or so.
A typical n-bit flash converter uses the brute force approach of having 2^n comparators arranged so that increasing input voltages cause successive comparators to change state. A 4 bit flash converter usually has 16 comparators and an encoding matrix to convert these 16 outputs to four bits. This circuit finesses the operation with 4 comparators and a bunch of resistors.
Briefly, each comparator stage represents a bit. Higher order bits contribute to the switching threshold of lower order bits though a weighted network of resistors. For example, the threshold of the comparator for bit D0 is set by the states of D1, D2 and D3 as summed by R21, R22 and R23 respectively..
Another way to view the digitizer is as a neural network that has been trained to map a continuously variable input onto a binary 4 space, which is a classical problem in neural network theory. The resistor values in the summing network that sets the threshold of each stage represents the strength of connections of synapses.
Many of the resistors have two designations, R3/R28 for instance, because the circuit board that was offered for the circuit had two digitizers. For the same reason the outputs also have dual designations. The reprint of the original Polyphony article goes into agonizing detail of the operation of the circuit and ways to use it.