Abstract
When the input data change, D/A converter using 2-1, 2-2…, 2-m-weights suffers from large impulsive noises (Glitch) owing to short-time data errors resulting from delay differences among registors and switches, which must be rejected by expensive and power consuming resample-and-hold circuit. And also it demands very precise weights only to maintain monotonicity.
To solve these problems, it is pointed out that we can use Parallel Expansion method. The upper m input bits are expanded into 2m-1 Thermometer Godes which have as many High Levels as the input binary number and each of which has equal weight of 2-m. And they are strobed into registors together with the lower bits to switch the weights simultaneously. Then, the glitch is suppressed to less than 2-m and the allowable weight error for monotonic conversion is expanded to ±2- (n-m) . To design TV signal D/A converters by this method, the allowable limit of the glitch is experimentally found and the increases in DG and DP owing to weight errors are calculated.
