Lessons Learned in Colorlight, LEDVision, and Sender Boxes

[Kensington Graves]

Forgot the pix.  I'll get to Chris' configuration tomorrow.  I've been going since 3:30 this morning and I'm pretty wiped out.  I may also need to flip to master branch if Chris has pushed an update that increases the number of outputs on the Colorlight.

[CaptainMurdoch]

Forgot the pix.  I'll get to Chris' configuration tomorrow.  I've been going since 3:30 this morning and I'm pretty wiped out.  I may also need to flip to master branch if Chris has pushed an update that increases the number of outputs on the Colorlight.

Is your issue that you have 8 panels in width but only 7 configured in the UI?  After re-thinking the layout, I think you may just need to number them in-order from left to right and top to bottom when viewed from the front.  The ColorLights should be handling reversing the chains.  I think I have to do it myself because my panels are in 3 different orientations in my matrix due to the window size and cable routing.

This is one of the reasons I always install the MatrixTools plugin, because I can draw on the canvas with the mouse in the UI and see if the pixels show up correctly on the actual matrix.

[Kensington Graves]

The matrix is only 7 panels wide right now.  I'll try ordering them left to right and see if that changes anything.  Worst case is that I disconnect the daisy chained cables and start fresh with smaller matrices and build up to the full size.

[CaptainMurdoch]

The matrix is only 7 panels wide right now.  I'll try ordering them left to right and see if that changes anything.  Worst case is that I disconnect the daisy chained cables and start fresh with smaller matrices and build up to the full size.

OK, 4-3-4-3 are how the ColorLights are setup as far as how many panels are on each chain?  left most (from the front) is 4-panels per chain, next is 3 panels per chain, next is 4, and far right is 3?

[Kensington Graves]

From the front, the top left quadrant (CL 2) is 4x7 (256x224), top right (CL 1) is 3x7 (192x224), bottom left (CL 3) is 4x7 (256x224), and bottom right (CL 4) is 3x7 (192x224).  Looking at it from the front, there are 28 panels in the top left quadrant, 21 panels in the top right, 28 panels in the bottom left, and 21 panels in the bottom right.  Does that answer your question?

[CaptainMurdoch]

From the front, the top left quadrant (CL 2) is 4x7 (256x224), top right (CL 1) is 3x7 (192x224), bottom left (CL 3) is 4x7 (256x224), and bottom right (CL 4) is 3x7 (192x224).  Looking at it from the front, there are 28 panels in the top left quadrant, 21 panels in the top right, 28 panels in the bottom left, and 21 panels in the bottom right.  Does that answer your question?

I was referring to the whole matrix while your 4 quadrants above are only half of it.  I'm assuming the 4 quadrants you refer to are going to be duplicated to make the other half of the matrix.  I think I'm following now, I was wondering why you kept referring to quadrants when you mentioned using 8 receivers.

I added an 'Auto Layout' button to the master branch which may help testing if you switch to master.  The auto layout button assumes each panel row is on a separate output, and panels have their arrows facing up so they are fed from the left side of the back.  Set the panel size, # of panels wide, # of panels high, and click the Auto Layout button and it will set the output numbers and panel numbers for you.

[Kensington Graves]

Yes, the whole matrix will be 4-3-4-3.  Although I am considering 4-4-4-2 just to make things interesting.

I'm flipping to master now, and I'll test tonight.

The Maybe a Good Idea Fairy visited me today.  Would there be any use for FPP to access the controller layout file produced by LEDVision?

[Kensington Graves]

You, sir, are a steely-eyed missile man.  That auto-layout is slick!  That's definitely a feature to implement in v.6.  It'll eliminate the "how do I..." questions on panels for new panel users--provided they "up orient" the panels.

The only issues I found were in test mode.  On the R-G-B cycle, FPP displayed B-G-R rather than RGB.  I checked to see if the auto-layout had changed the color order, but it hadn't.  I flipped it to B-G-R on the test page and it displayed correctly.  Also, on the R-G-B-None chase, I couldn't really see a black break.  But the panels displayed the proper colors on my Xlights matrix test sequence (butterfly to fan to plasma; the latter two pictures are attached).

[gergmchairy]

I added an 'Auto Layout' button to the master branch

You sir are a legend..  I asked for this a year or more ago and it went nowhere....

[CaptainMurdoch]

The Maybe a Good Idea Fairy visited me today.  Would there be any use for FPP to access the controller layout file produced by LEDVision?

I don't know how much the controller layout file helps since the data protocol is so simple.  FPP doesn't even need to know about controllers, we just send a block of data in a simple format.  One or more packets per ColorLight row, 497 pixels per packet.  So if your ColorLight layout matches your FPP layout and is 14x14 of 64x32 panels, that would be 896x448.  The 896 pixels per row will fit into two data packets since it is less than 994 (497*2) and we'd send two for each row so a total of 896 packets (448 rows times 2).  FPP could push the data out a lot faster if we didn't have to rearrange or apply a gamma or color order adjustment, so I'm wondering if there should be an optimized code path for those who lay out their matrices in the simplest manner such as yours, who also leave the color ordering up to xLights, and leave gamma at 1.0.

What would be nice is if we could program the receivers ourselves, but that would be a big project and would be very low priority for me.  See my other response at the bottom of this message. 

From your other post, the Auto Layout button sets the panels back to the "C - Def" setting which means they are all using the default color order at the top of the screen, but it does not change the default itself.  The channel test page assumes RGB color order and the panels are normally BGR so that's why you saw a BGR pattern.  I normally use R-G-B-All-None when testing because it makes it easier to see the order.  It is harder to differentiate between R-G-B-None and None-B-G-R if the pixel colors are BGR vs RGB.  Having the All in there makes it easier to see the pattern of 5.

I added an 'Auto Layout' button to the master branch

You sir are a legend..  I asked for this a year or more ago and it went nowhere....

If you saw how long my ToDo list is, you'd get a sense of why.  It rarely gets shorter, new ideas pop up all the time.

[Kensington Graves]

Well, I certainly appreciate you taking the time to add it.  I have the last Unistrut fittings on order, so with some luck I can finish building the frame by next weekend and start adding panels to the second half.

[Kensington Graves]

So now that the IMBN Matrix is built and functioning, I started playing around with the sender and scaling videos and what not.  The main issue I was trying to address was producing a full screen, properly scaled image on the matrix.  Heretofore, I would run video on HDMI1, which in turn was connected to my sender box, and that pushed the video to the receivers and panels.  However, I was getting "letterboxing" at the top and bottom of the screen.

I attempted to remedy the situation by experimenting with different RPi output resolutions and sender input/output resolutions.  The closest resolution to the IMBN Matrix aspect ratio is 800x600, but that resolution did not really do what I wanted.  And going lower or customizing the sender output resolution to 896x448 (which it can do) did not produce the desired result.  Lower than 800x600 made the last three panels go all wonky--display green lines, etc.

I decided to resize a test video yesterday and see how that worked.  When I went to add the media to the playlist, I happened to click on the "Video Out" dropdown box (normally "Default" and I've never messed with it) and found both my virtual matrix and LED Panels in that dropdown.  I chose "LED Panels" for fun, saved the playlist, ran a 1920x1080 video, and it played perfectly and filled the whole screen.  Except it was red shifted.  Martian conifer forests and Christmas trees are cool and all that, but I sort of wanted my original coloring.  I went to the Channel Outputs -> LED Panels tab and switched the color order to "BGR" from "RGB", and the video displayed the correct coloring.

It also appears that "Sequence + Media" also has the option to push the sequence directly to the panels, so I may not need the existing virtual matrix model at all.

Given that a 30 second video is pushing 300MB when I drop it on the Xlights model for the matrix, this solution appears to be the best alternative I've found.  I need to test it in Remote mode, but I'm assuming it will work.  The RPi running the matrix will be on Remote anyway.

Once again, FPP has come through.  

[dkulp]

When playing directly to the Pixel Overlay Model like that, it doesn't use the hardware decoder or scaler or anything.  Thus, if you can, re-encode the video to the exact size needed to reduce the CPU load a bit.    4K videos and such likely aren't usable without the hardware decoder, but you don't need that high of a resolution for your panels.

[Kensington Graves]

Thanks for the info, Dan.  I figured there was some magic happening under the hood.  I was pleasantly surprised to find the feature.  I did re-encode the video for a lower resolution, and it worked as well.