search

hzeller / rpi-rgb-led-matrix

Controlling up to three chains of 64x64, 32x32, 16x32 or similar RGB LED displays using Raspberry Pi GPIO
GNU General Public License v2.0
3.66k stars 1.16k forks source link

Multiplexing for non-standard panel: 104*52 res, 1/13 scan rate, 320mm * 160mm LED Panel #1587

Closed andytheengineerguy closed 1 year ago

andytheengineerguy commented 1 year ago

Hi,

TLDR: I think I've reached the point where I need to make my own multiplexing mapping for this non-standard panel, and was wondering if anyone could lend me a helping hand / point me in the right direction? How do I do this?

Context: I would consider myself a medium-level user of this library - I currently have two 128x64 panel pairs running in parallel (basically, two 128*128 duplicate squares) using rpi-rgb-led-matrix, using both C and python bindings, with gifs I have made myself. I wanted a better scan rate so the images would appear better in photos, so I purchased these 1/13 scan rate panels from Aliexpress (much better than the 1/32 scan rate panels I have now), and am trying to bring them to life. So far.... no luck. All 18 multiplexer mappings, with different row address types, and even different considerations for how they might be 'chained' internally (i.e, 52 cols with a chain of 2, 26 cols with a chain of 4, etc) does not work..... I have tried everything :( So I have accepted I need to make my own multiplexer mapping for this. But immediately, I feel out of my depth.

Additional Info:

  1. The panel I'm bringing up is this one: https://www.aliexpress.us/item/3256805351469362.html?spm=a2g0o.order_list.order_list_main.5.7b391802I9Ggqy&gatewayAdapt=glo2usa

  2. The multiplexing config that's worked the best so far is:

    image

  3. The rotating square in this case looks like this. It looks like the screen is divided into 4 sections height-wise, where the first 2 and second 2 duplicate each other - anyone know what this could mean?

    image
andytheengineerguy commented 1 year ago

Update: Fixed it. Turns out the mapper required for this panel was a form of the ZnMirrorZStripe multiplexing pattern, just with a larger tile_height.

Marking this as closed.

bsparacino commented 1 year ago

@andytheengineerguy Can you share what you did to make this panel work? I believe I have the same panel and am struggling to get it to display properly.

andytheengineerguy commented 1 year ago

hi @bsparacino, sure, happy to help.

I found out (with no help from the suppliers or the non-existant datasheet.. 😂!) that the panel almost worked with a ZStripeMultiplexMapper configuration, with even- and odd- block offsets set to 4. But the height was still funky (as shown in the screenshot of my original message).

To fix this, the tile_height needs to be 13 (not 4, which is the original value), as shown:

image

So...

  1. Change height to 13
  2. Register a new multiplexer mapper image
  3. Re-make the codebase
  4. Run the rotating square demo using --led-multiplexing=19

Once you've confirmed this solution actually works, I'd recommend modifying the ZStripeMultiplexMapper class in order to preserve the existing mappings that depend on it. I achieved this by adding the width and height as variables that you pass into the constructor. So then we can pass our new height of 13 while still utilizing the existing ZStripeMultiplexMapper class.

image

When you do this, don't forget to update all the other registrations with their default tile sizes (8 and 4)

image

Come to think of it I will probably do a PR for this change.

Good luck, lmk how you go.

bsparacino commented 1 year ago

What parameters are you using to run the demo? When I run it with multiplexing 5 it is similar to your initial image. But with your code added I am not ending up with the corrected image.

./examples-api-use/demo -D 0 --led-gpio-mapping=adafruit-hat --led-cols=104 --led-rows=52 --led-row-addr-type=0 --led-multiplexing=19 image

./examples-api-use/demo -D 0 --led-gpio-mapping=adafruit-hat --led-cols=104 --led-rows=52 --led-row-addr-type=0 --led-multiplexing=5 image

andytheengineerguy commented 1 year ago

Hmm. Strange. Your panel looks the same in terms of specs (resolution, scan rate, size) but the physical construction is definitely different than mine. Different locations of the IDC connector (maybe different LED driver chips?) etc.

We are using different electronics. I'm using an RPi4 with an active-3 hat from electrodragon.

I just ran the demo with the following and it worked ./demo -D 0 --led-gpio-mapping=regular --led-cols=104 --led-rows=52 --led-slowdown-gpio=4 --led-row-addr-type=0 --led-multiplexing=19 375021560_779879780574895_5363056622137742043_n

Just re-ran the 5 pattern, we are definitely getting similar results for 5 ./demo -D 0 --led-gpio-mapping=regular --led-cols=104 --led-rows=52 --led-row-addr-type=0 --led-multiplexing=5 --led-slowdown-gpio=4 375019634_252345181098337_3195114032820448922_n

andytheengineerguy commented 1 year ago

I have not used the adafruit hat before. The only other thing I can think of right now is the connector type - is it HUB75 or HUB75E? If it's E, you might need to enable the E line with some soldering/jumpers on the hat (I needed to do that with different 128*64 panels I brought up using this library)

But for these 104*52s specifically, they are regular HUB75s, so I didn't need to worry about it.

bsparacino commented 1 year ago

About 20min ago I actually soldered the E line on the hat which resulted in the photos I posted, before I did that it was a total mess.

bsparacino commented 1 year ago

Well I must have different modules, this function solved it for me https://github.com/hzeller/rpi-rgb-led-matrix/issues/1499#issuecomment-1385945046

andytheengineerguy commented 1 year ago

Out of curiosity I just tried the mapper from #1499 (comment) and got a similar vertically-jumbled pattern to what you started with. So, yep, different panels indeed. At least its consistent: image

Well, glad you got yours working, and good to know that if I ever need an alternative for these 1/13, 104*52 panels, I know which mapping to use 💪