[sdiy] discrete 3-bit AD converter??

[Magnus Danielson]

From: media at mail1.nai.net
Subject: [sdiy] discrete 3-bit AD converter??
Date: Sun, 30 Dec 2001 10:44:56 -0500 (EST)

> 
> This might be a stupid question, but I'm willing to follow it up with an
> even stupider question :)
> 
> I'd like to be able to convert a CV going from 0 to 5 volts into a six
> 3-bit binary numbers.  Well, just to illustrate my question, let's just say
> I want four 2-bit numbers to represent one to four volts, such that:
> 
> 1V = 00
> 2V = 01
> 3V = 10
> 4V = 11
> 
> I'm thinking I could do this with a stack of comparators each set to the
> next highest voltage (off of a "totem pole" voltage divider run off of the
> incoming CV), but I'm thinking there must be an easier way that would use
> less chips.  If that is the best way, is there a single CMOS chip (like the
> opposite of an LED segment driver) that will encode it into a binary
> number??  Are there chips with multiple comparators IC's without the extra
> pins for such a purpose??  Basically, I'm trying to devise a way to enable
> a number of logic states with single control voltage.

You need 3 op-amps, so you could use a quad op-amp chip if you want
to.

John Simonton (The PAIA guy) made a 4 bit digitizer. The schematics
seems to have fallen of the web.

Jörgen Bergfors Octave Quantizer is a deluxe variant:

http://www.idg.se/personal/bergfors/bergfotron/Oct%20quantizer.gif

If you look at Jörgen's schematic you will see that there is 4 input
comparators, each generating one bit. The top comparator refers
directly to the ground voltage. The second comparator compares to the
result of the first and ground. The third comparator refers to the
first and second comparators and ground, and naturally will the fourth
refer to all the previous three and ground. You will see that the
resistors feeding the - input will weigth the different references so
that you get the right voltage.

What is neat by Jörgen's design is that it uses CMOS switches and a
buffered reference voltage network to get the output voltages insead
of the traditional resistors from output mixing. By this he avoids
depending on the output resistance and actual rail-to-rail properties
of the op-amp outputs. The internal resistance of the 4066 do vary
with the voltage, but are small compared to the 100k resistors, so it
is only a minor fault-contributors, the matching between the 100k
resistors is a bigger issue really.

Cheers,
Magnus