This correspondence presents a test data reduction technique called Accumulator Compression Testing (ACT). In brief, the accumulator syndrome is the sum of the partial syndromes of test results over time. This accumulator syndrome and the syndrome for the test are treated together as the signature of the test data output. ACT requires an adder and a register, in addition to the counter required for syndrome computation.

The drawback of any data compression technique is that faulty signatures can be mapped into good signatures, masking the fault effect. The authors compare ACT to signature analysis both analytically and experimentally. For values of the accumulator syndrome close to (m**2)/8, where m is the number of ones in the output data, ACT will do worse than signature analysis. This will not happen in a worst-case output stream, where the number of ones is close to half the number of bits. It is not clear how often this will happen in real cases. The experimental results are good, and it is possible to adjust the input stimulus to provide an output with proper characteristics.

The main disadvantage of ACT is the extra hardware required. However, as the authors point out, for microprocessor testing this hardware is already available. It is not clear how well other techniques with equal amounts of hardware would compare to ACT. Still, ACT seems worth investigating. I also appreciate the effort of the authors in presenting fault simulation results with their analysis.