Having all kinds of problems with gamma.

I finally got things built and I am testing and attempting to do a rough calibration with the pixels on the floor. Here is the problem:

I am simply trying to get a good greyscale. I have made a PowerPoint slide that is a gradient from white on the left to black in the center and white on the right.

What I am seeing is little to no grey. The entire strip is bright white until the middle where about 20 pixels are off. Bordering the black pixels it is very slightly less than full white.

If I increase the gamma the amount of black pixels goes up, but there is still no gradient. These pixels supposedly have 24 bit color capability, so I should have 256 shades of grey available. I have played with as many setting as I can find to no avail.

Here is my setup:

• 300 Pixels (SK9822) when installed it will be about 475, but testing with 300.
  
  
  • Settings - all 300 at the top of the image, 0 on the sides and bottom. I just want to capture and show the top of my PowerPoint slide.

  
  

• 60A MeanWell Power supply - Inserting power at the beginning and the end of the strip
• RPi4b - slightly overclocked, 8GB version
• Logic level conversion using 74HCT245
• Video capture from Hyperion Windows capture
  
  
  • Settings: Scaling factor - 16

  
  

Thoughts?

Zitat von jeroen warmerdam

why 16? :eek:

with a high setting the colours become pale in PROTObuffer in Hyperion, your network can provide probably the highest quality with DirectX11 so you can set it on 1,
60 FPS
protocol; Protocol Buffers
port; 19445
priority; 110

then alter setting Image Processing > increase up to 2.5 RGB on each color and test again.
Turn on Blackbar Detection
Set protocol ledstrip to SK9822 OR APA102 ( sometimes you can switch those for better results because the SK9822 is in fact the (cheaper) APA102
then set to your liking ofcourse

Alles anzeigen

My thinking was that I only have 300 horizontal right now, and I am beginning with 3840 in the source. It has to get scaled somewhere, right? If so, my I9-9990 / GTX2080 can probably do it better than the RPi. Anyway, I changed it to 1.

I got the best results with the gamma set to 1 for all three colors. I will explain in more detail below.

I am using the SK9822 protocol. It adds two more configuration settings over the APA102... Do you know what they are?

• Max Current - Defaults to 31
• Adaptive Current Threshold - Defaults to 255

I would love an explanation somewhere to these...

Zitat von jeroen warmerdam

thats weird, are you sure you set everything correct in LED layout and saved your setup?

did you use the hyperion free app to test with different effects and colours? > you can download it from the Playstore and then connect to localIP:19444

It isn't that weird. The gradient in PPT only has full black for a very thin vertical strip. When you see what I discovered (again, I will detail it below) it will make more sense.

I have the Hyperion app. I am going to email myself the same PPT file and test it from my phone.

Zitat von jeroen warmerdam

as last, did you connect the GND to your whole setup? use more GND pins to your PSU on the RPI, don't forget the ledstrip and levelshifter

That's the one thing my background in AV (been in the industry since 1974) has taught me. A good ground can save the day. The Mean Well supply is providing the 5V and Ground for the whole system. I will add as many more to the Pi as there are pins for. According to the pinouts on the level shifter IC (74HCT245N - The Philips Semiconductor version) I tie the ground to two pins, 10 and 19.

Here is a question about the level shifter. It is an eight channel device. I am only using two, data and clock. I read somewhere that best practices suggests that the unused channels should also be tied to ground. But I couldn't find details. Do I tie both in and out sides of the unused channels?

Anyway, I did some more testing and here is what I found:

The system has extremely high bit resolution as it approaches black. I made a slide that was full black and a strip in the middle that had an RGB value of 16.16,16. No problem. I then started to reduce the value of the strip. It disappeared at 1,1,1, but it was visible at 2,2,2. Against a background of 0,0,0. Hell, unless my room is pitch black, I can't even see that in my monitor. This was using a gamma value of 1.0 for all colors.

BUT (you knew there had to be one, right?)

When I did the inverse (background of full white [255,255,255]), I had to bring the stripe in the middle down to 16,16,16 before I could detect a slight difference in brightness. Putting this in terms of video, it is clipping whites BIG TIME. If I made the middle stripe 192,192,192 I could barely detect a slight difference in level of those pixels at a gamma of 7 (!) for all colors.

So how do I fix this? If I use the brightness setting it drops the whole strip of pixels. The same goes for the max current.

Do I need to do the color calibration for intensity as well as the hue? I am fine with the colors, so I didn't do the calibration. If I set white for 16,16,16 (or wherever it tops out in luminance) will that solve things? And then should I do all of the other colors? Or will the brightness compensation handle that. AND, if I take the time to do the luminance for all 7 colors, then can I turn brightness compensation off altogether?

I am attaching the powerpoint in case anyone wants to play.