16b tilt/pan #1 - chaser->play resets LSB values

[GGGss]

As promised in numerous threads, I'm digging deep into the 16b MSB/LSB issues.

Issue #1: chaser -> play resets the LSB set values !: 0

Project is attached,
* 1 scene with a mover which has fine tile/pan capabilities. Set fine value different than the course value.
*clone this scene
*add these to a chaser
in the preview, the beam stays dead center
*go produce
*plays with the 2 scenes (inside a solo frame) -> beam still dead center
*press play to chaser -> You instantly see the beam deviate from the set dead-center
-> DMX monitor reveals the 0 value for the LSB
-> Watch the video which will show you the outcome

https://youtu.be/3KnryBfQm_0

[mcallegari]

Hi Fredje, thanks for reporting.
Your findings have been tracked here
https://github.com/mcallegari/qlcplus/issues/1484

I think I have now a solution that properly implements 16bit fade (not just for LTP channels)
Can you please check if all your issues are resolved with the automated build on GitHub?
https://github.com/mcallegari/qlcplus/a ... 7452723908
If you have troubles in downloading the right version please let me know and I'll upload it somewhere.
Thanks

[GGGss]

That these issues were going to need an overhaul on the engine, I could have assumed. Checking the repro, I only now see what that means. Ow boy (!)
I found the artifacts from the build. Will try and report.

[GGGss]

Hi Massimo,
I already detected an issue with the latest build: QLC+-qt5-4.12.8-20240108-e43706e.exe.

The two scenes are almost identical except for 1 LSB channel. (I did not choose to use zero and 255 values - these tests are around the point where LSB+1 becomes MSB + 1, and LSB != 0 is for the next run).
Please take a look at the project attached.
When the chaser is started, the MSB values are copied to the LSB values. One would expect these to stay as the initial value. The fade-in kicks in. The LSB value drops to the scene-set value in step 1. In step 2, the value rises from the copied value to the set value in the scene.
This behaviour can be monitored with the DMX monitor.

[mcallegari]

Oh yes, thanks
I've pushed a fix for that. please test again
https://github.com/mcallegari/qlcplus/a ... 7466298396

[kenact]

I downloaded QLC+_4.12.8-20231110.exe, which I assume doesn't have the latest fixes for this issue. If you post here when the latest Windows test version is available, I'll be happy to test it and report back.

[Yestalgia]

I downloaded QLC+_4.12.8-20231110.exe, which I assume doesn't have the latest fixes for this issue. If you post here when the latest Windows test version is available, I'll be happy to test it and report back.

You can find the latest Windows build in the github actions link that Massimo sent. See image.

[GGGss]

Oh yes, thanks
I've pushed a fix for that. please test again
https://github.com/mcallegari/qlcplus/a ... 7466298396

Brilliant!
Issue: when you start the chaser, the LSB values are reset to zero. See the example attached PTissuecontrol2.
I now set the LSB value for the tilt to 90 in scene one. If you start the chaser you immediately see that the LSB values (both LSB tilt and pan) are reset to zero.
Once the chaser passes its first run-through, the values normalize and it works. Hooray.
So the only problem rests when the chaser runs for the first time.

Let me have a look when LSB at 255 + 1 should add MSB+1 and reset the LSB to zero:
Except for the ^^issue this works A-Okey+. (added the + for my appreciation of your work) - @followers, have a look at PTissuecontrol3 attached.

Let's step up and try something different: EFX
Now the Issue with the EFX has come to our attention... I'll start a new thread

[mcallegari]

I found and fixed this one too
https://github.com/mcallegari/qlcplus/a ... 7482129456

[GGGss]

Tested and confirmed on LTP channels,
It is now tested and confirmed with non-conflicting HTP channels also.

There is an issue with dual HTP channels, where HTP-logic should kick in over on combined scenes.

The LSB value of an HTP channel stays in place, albeit the calculated value of MSB+LSB is lower than the set scene.
See the project attached.

Scene 1 and 2 play nicely; HTP-logic does its job as already stated.
Scene1= 0,0 = 0000h = 0
Scene2= 1,60 = 013Ch = 316
Scene-preset= 0,180 = 00B4h = 180

Now select the 'scene-preset' on top of 'scene2' or vice versa.
180 < 316, The HTP-logic in the engine does an MSB (01h) to MSB (00h) compare and an LSB (3Ch) to LSB (B4h) compare.
With the latest build, QLC+-qt5-4.12.8-20240110-6611b14, HTP-logic lets the MSB and LSB comparison win separately (B4h wins).
Result: Even though the MSB+LSB is higher (316 vs 180), the LSB to LSB comparison lets the LSB (B4h) of the lower MSB+LSB-total (180) win.

If this can be solved, not only the theatre guys will express ultimate gratitude.
QLC+ becomes a fully 16-bit enabled engine - a fact to be very proud of.
When I deeply remember this, it was one of the first issues I posted on the forum back in 20??-something.
Now, what about an HSI color model? Just kidding!! (I will now hide in a corner for the next 30 minutes, resting solely on my knees)

[Yestalgia]

QLC+ becomes a fully 16-bit enabled engine - a fact to be very proud of.

Massimo is making some big moves this year!

[mcallegari]

Hey Fredje I'm not sure if you received a notification from GitHub but I left a message for you.
I worked some more on 16bit and I think I found the final solution.
Can you please test either the GitHub build or the ones I loaded in the QLC+ website (test area)?

Thanks!

[Deece]

Hi, sry for jumping in, but it seems the issue I was seeing with the X-Y Pad Presets - Scenes, is related to the work you are doing on 16 bit Pan/Tilt.
I have tried the test, Linux dated 23-02-24, and it fixes -some- of the issues, but when using a scene for presets in X-Y Pad the handle and the side bars on the pad produce unexpected results.
I attach a WorkSpace demonstrating the issue.
4.12.7 : some of the fixtures don't respond, some respond incorrectly - (using the DMX monitor)
4.12.8 test 23-02-24 : All the fixtures now respond, but with incorrect results.

[GGGss]

Hey Fredje I'm not sure if you received a notification from GitHub but I left a message for you.
I worked some more on 16bit and I think I found the final solution.
Can you please test either the GitHub build or the ones I loaded in the QLC+ website (test area)?

Thanks!

I got the notification and saw it
I'm grabbing some gear together and take it home to regression test the changed code.

[mcallegari]

Great, thanks!

[GGGss]

HTP channel behavior is now correct for 16bit values.
LTP logic follows its behavior in conflicting situations.
I reported this to github as well.
https://youtube.com/shorts/LzlSQrVsLVE?feature=share

I'll do some more testing: XY-pad, EFX, ...