Today I received the book for my CSC 376 Computer Organization class, Computer Systems Organization & Architecture by John D. Carpinelli (Addison Wesley Longman, 2001).
It has a full-page sidebar on page 32: “PRACTICAL PERSPECTIVE: Why LEDs Are Usually Active Low”. They get points for the attempt, but they fall short:
This is one reason to use active low LEDs: It is preferable to source the current to light the LEDs directly from the power supply than from a logic component.
Ignoring the grammatical error in the sentence structure, they fail to explain why it was preferable in the past, and why it rarely matters in modern designs.
Bipolar TTL devices, which are all but obsolete, usually have much more capacity to sink current than to source current. Modern CMOS devices commonly have equal source and sink ratings, leaving no particular justification on the basis of drive capability.
They attempt to proide a second justification:
A second reason to use active low LEDs has to do with speed. The logic signal to light the LED often comes from the output of a gate, usually an AND or OR gate. For active low signals, the faster NAND or NOR gate is used instead. Thus, active low LED signals are often generated faster than their active high counterparts.
This is not universally true, and is completely irrelevant even in the cases where it is true.