Dual 8-bit ADCs provide pseudo high resolution

Douglas Butler, Imetrix Inc, Cataumet, MA -- EDN, 10/14/1999

Many inexpensive µC s now come with several 8-bit A/D-converter channels. However, many high-dynamic-range sensors need more than 8 bits of resolution. The simple technique in Figure 1 provides a pseudo-high-resolution result for any application that has a mixture of large low-resolution signals and small signals near zero.

For example, the turning-rate sensor of a robot autopilot operates in two realms: Either the robot is rapidly turning to a new heading, or it is trying to travel a straight line. When the robot is turning, the least significant bits (LSBs) are meaningless because they change too fast for the system to track them. When the robot is trying to travel a straight line, the LSBs are important, but the most significant bits (MSBs) are all zero.

By using two 8-bit ADC channels for the same signal, the technique in Figure 1 can accommodate both conditions. One channel, which the controller uses for fast motion, reads the input directly and adds four 0 LSBs for a 12-bit result. The second channel reads the signal amplified by 16 with four 0 MSBs. The amplification gives this channel the resolution necessary for small-error signals while the robot is traveling a nearly straight line.

The controller first reads the second channel and checks for overflow. If no overflow occurs, the result is the reading plus four zeros for the MSB. If an overflow occurs, the controller discards this first reading and proceeds to read the first channel.

The controller can perform amplifier-offset correction any time the signal is in the small-signal realm by reading both channels, shifting 4 bits, and subtracting. In most cases, this correction is unnecessary. (DI #2415)