The panel_pidp1 simuation error

[Bill E]

It''s actually an error in the pdp-1 emulator itself.

The -1 has a number of 'microcycles' within the main 5 usec cycle. During that time bits in various registers, especially the AC, can change.

The emulator samples the state of all the light bits at one random subcycle in the main cycle.

I think Angelo's intent was to try to duplicate what the real lights would do, reflecting the microcycle changes. BUT, sampling at random points ends up biasing the light updates way too much towards the microstate values, smearing the light pattern. You can think of it as stretching microstates to look like full 5usec states, but that's a simplification. This is the cause of the complaint from CHS about the unrealistic appearance.

It's even worse. An analysis determines that the microstates, which are fractions of a cycle and actually contribute nothing to the real visible lights because they are of such a short interval, end up having a major contribution to the light intensity.

With the old single-sample approach, that one random-subcycle sample was the displayed lamp state — so a rare 0.2us transient with a 4% chance of being sampled meant the lamp flickered fully on 4% of the time and fully off 96% of the time. A stark, visible binary flicker from something that should be imperceptible.

The new algorithm used completely avoids this because of its new sampling algorithm.

You can wake back up now. :)

Bill

[AB]

I laughed and woke the wife up with that one