Repository navigation
Description
Wire index 3 on a cab is a pan control. I read it off the screen on my own unit
today, on CorOS 4.0.1, and the numbers the catalog gives for it are not the
numbers the unit displays.
I wrote wire values by index and read the screen:
| wire | screen |
|---|---|
| 0.0 | 50 L |
| 0.5 | C |
| 0.75 | 25 R |
| 1.0 | 50 R |
So the display is bipolar, linear in the wire value, centred at wire 0.5, and it
runs 50 on the left to 50 on the right with the side as a letter.
I read the same four values on a mono cab and on a stereo cab. The display is
identical on both. The only thing that differs is the label: a mono cab says
PAN and a stereo cab says BALANCE. The catalog gets the label right.
What the catalog gets wrong is the span. It declares mono PAN as 0..10 and
stereo BALANCE as -1..1. Neither one produces 50 L at wire 0.0.
How far this reaches
36 parameters carry a min_string / mid_string / max_string triple, which is
how the device marks a control with labelled ends. Their declared spans do not
agree with each other:
| declared span | count | examples |
|---|---|---|
PAN over -1..1 |
14 | |
PAN over 0..10 |
8 | Default Cabsim 12000 idx 3 and 11, Mixer 11000 idx 1 and 3, Parallax 3008 idx 17 and 25 |
BALANCE over -1..1 |
8 | Default Cabsim (ST) 32000 idx 3 and 11 |
| over 0..1 | 3 | |
A/B PITCH MIX over -50..50 |
1 | Micro Processor (ST) 7039 idx 14 |
Only the last one matches what I read on the screen. That is one model out of 36
describing the control the way the unit draws it.
So this is not only a cab problem. The Mixer pan and Parallax pan are in the
same family, and both convert Real against 0..10 today.
Why it matters
ADR-0016 says a Real is the screen's number line. For these parameters it is
not, so a caller who asks for a number they read off the unit gets a different
position.
The symptom that shows it clearly: to reach hard left you write Real(0.0) on a
mono cab and Real(-1.0) on a stereo one. Same physical control, same wire
value at the end stop, two different numbers, only because the catalog declares
two different spans. On the wire, 0.0 is hard left in both cases.
What I think we do
Two options, and I have not picked one.
- Refuse
Realon a parameter whose declared span does not reproduce the
display, and say why in the refusal. This follows whatunits.UNMEASURED_BOUNDS
already does for a bound nobody has measured. - Record the -50..+50 display mapping for the labelled-end family and convert
against that, with the readings above as the evidence.
Option 2 is more useful to a caller and needs more measuring first: I have read
one parameter on two layouts, not the family.
Either way the readings belong in tests/test_scales.py, where a screen reading
has to reproduce the display exactly.
Not from the clone work
I found this while checking PR #45
on the unit, but it is older than that PR and that PR neither causes it nor fixes
it. Its parser change is correct and measured. I am filing this separately so it
does not get folded into a review.
Also confirmed while reading, both of which were sitting in the "cannot yet
explain" appendix in docs/domain-model.md: displayPos
is the order the unit lays controls out on screen, and mid_string is the label
at the middle of the wire range.
Activity
fix: a pan's drawn span is measured, not declared