Video output mode

[ronp]

While I am dreaming, eventually you will have to support HDMI/DVI/Display port to Ethernet to really handle pixel output properly. This is how the large matrix displays are driven, and don't tell me to use multiple pi's.

[Gilrock]

If you just want to drive a big led wall like the 36' million pixel wall in my church you won't be using xLights.  There is plenty of software to create a video to drive those displays.

[ronp]

It is not really about driving lots of pixels, it is more about accepting mediocre. I want stale bread, a grisly steak, flat beer. What we have works, but it is just OK. So if I went and bought a monitor I would accept a 20/40 hertz refresh rate? If you put a scope on the Ethernet port of a PI or computer you will see jitter on the Ethernet data. Why would we not move to a better output device that would have exceptionally low jitter, accurate timing and support higher refresh rates like our monitors. Say I have 120 pixels on my roof line, a curtain effect at 20 hertz over one second would turn on 6 lights at a time, and 40 hertz would be 3 at a time. If I cut the time in half for a fast beat, it just gets worse. Why wouldn't I want an output to support 60/80/100 hertz refresh rates?

If you look at the professional software "Madrix", it supports video output, why would we not move toward the same thing professionals use? 

[Gilrock]

xLights mainly just creates the data anyways so it would be xScheduler where an output would be added.   I'm not sure exactly how you are asking it to work.  You want some data driven out the HDMI port on PC graphics card?  You mentioned ethernet in the first post but then in your reply you said ethernet as jitter.  What hardware would I use in my yard to drive my pixels strings with this data?  I just haven't heard of anyone using hardware that could benefit from this.  And I've seen someone demo Madrix and I wasn't that impressed.  It says something to me that the person owned Madrix but still mainly uses xLights.

[babybear]

It all sounds pretty. And what about those of us who really don't have much money. The first time I looked into building a Mega tree the only site I could find was LOR. By the time I priced it all out it was almost 4 grand. I walked away for a year. Then I stumbled onto Xlights/nutcracker. My first tree was under 500 dollars. What your talking about does not sound cheap. I work with glass all day. People will pick out a speck in the glass. You see glass is made to look through not at. Just like these shows. People are not looking at the LEDs and how many are lighting at a time, 6, 4, or 3 they are standing back and looking at the show, all the Lights at once. At my house I have people stop and watch a show in the middle of a snowstorm, do I really need a higher resolution? I get the feeling you have yet to put up a show yourself. There is a special meaning/feeling to doing a show. It has nothing to do with bigger brighter better. Its to make people smile

[ronp]

Good discussion. These can be built for around $40, and I already have built two. It would be a HDMI input port with 1-6 Ethernet 100Mbit output ports. The card would show up as an additional monitor to your computer. You would just play a video and the hardware would strip the data from the computer display data and output as Ethernet multicast pixels. The one I did supported 640x480 at 60 hertz, but the chip supports up to 1080i.

The format can be whatever the forum decides, for example from each pixel on the display:

0x00 00 00 - null
0xFE FA 20 - frame sync
0xAA AA AA - Ethernet multicast address
0x01 01 01 - RGB pixel data #1
0x02 02 02 - RGB pixel data #2
. . .
0x03 03 03 - RGB pixel data #170 (repeat)

[Jguenther]

You then have to create your content as all video which seems a lot harder than just dropping an effect on a model right?

[JonB256]

Good discussion. These can be built for around $40, and I already have built two. It would be a HDMI input port with 1-6 Ethernet 100Mbit output ports. The card would show up as an additional monitor to your computer. You would just play a video and the hardware would strip the data from the computer display data and output as Ethernet multicast pixels. The one I did supported 640x480 at 60 hertz, but the chip supports up to 1080i.

The format can be whatever the forum decides, for example from each pixel on the display:

0x00 00 00 - null
0xFE FA 20 - frame sync
0xAA AA AA - Ethernet multicast address
0x01 01 01 - RGB pixel data #1
0x02 02 02 - RGB pixel data #2
. . .
0x03 03 03 - RGB pixel data #170 (repeat)

Ron, how is what you're describing different than a ColorLight Sending board with outputs to multiple ColorLight receivers?  Or is it pretty much the same but your new design? (and possible cheaper )

[Gilrock]

So xLights already has an export to video feature for the previews so just play that video in your favorite video player.  I'm pretty sure xSchedule can be told to play a video out the computer also.

[dkulp]

Good discussion. These can be built for around $40, and I already have built two. It would be a HDMI input port with 1-6 Ethernet 100Mbit output ports. T

6 - 100Mbit is still less than one GBit (kind of).   If using xSchedule on a PC or similar, you could even bond multiple ethernets if your network switches can do that.  (mine can)

FPP does have a Virtual Matrix function that can take a range of data from the fseq file and output it via the HDMI port.  I haven't played with it much though.  The bigger problem with the Pi/BBB is that you will hit limitations on loading the data off the SD card/USB drive before much or this gets to be a serious problem.  On the latest Pi3B+, you can only get 20MB/sec off the SD card, 30MB/sec off USB thumb drives.  (BBB is 21MB/sec SD and 10MB/sec USB) Until we change over to a fully compressed fseq file (or default to btrfs), those will drive some limits anyway. 

[Gilrock]

For full disclosure to all the other users aren't you the Ron from ronsholidaylights.com?  If so then is this a push for a product you already have or something your testing?

[ronp]

Yeah, it's the same Ron as ronsholidaylights. I developed HDMI to Ethernet converter a couple years ago, but it didn't go anywhere. Now Falcon is having problems with his P5 panel with both a beagle bone and one of my panel controllers. It seemed like it was time to bring this up again, since the P5 panels will make channel counts pretty large. This is basically a product I want, so I am ok with it all being open source. The idea is to move you guys toward how the large video displays are done.

I really like the idea of a single high performance output converter, kinda like the color light stuff. But with the ability to drive both the pixel controllers as well as the panels. I am pretty much offering free engineering to design a open source standard/platform for the community. Selling stuff does really interest me too much, that's why my stuff is sold by the renard-shop.

[ronp]

So I really don't consider myself a vendor. I am only involved with one item for sale on the Renard-Shop, and it is basically to help build the 4 output controller. The rest is all done by the store, with no profit for me and a couple dollars for the store. I have done quite a few things over the last couple of year

1 - First with a 32 output pixel controller, and probably the first guy to figure out it was a dumb idea. Please use smaller controllers.
2 - First with pixels extenders as part of the controller
3 - Only guy doing power over Ethernet to drive pixels
4 - Completed my first led matrix controller supporting E131, while everyone was fiddling with the beagle bone.
5 - First guy to support 16to1 (or P5) LED matrix's.
6 - First guy to design pixel board for the PI, two other vendors are selling variations, one is the exact board.
7 - First guy with a PI led matrix board.
8 - Had gamma correction in my controllers before it showed up in xlights. A must for LED panels and nice fades, hint 2.22.
9 - First guy to do daisy chain-able Ethernet for controllers.
10 - My controllers will support 100 hertz update rates, will yours?

Not sure all of my claims are correct, but I have am clearly pushing designs by all that I have supported. I do this as a hobby. My goal is to push the forum towards the future.

[dkulp]

Yeah, it's the same Ron as ronsholidaylights. I developed HDMI to Ethernet converter a couple years ago, but it didn't go anywhere. Now Falcon is having problems with his P5 panel with both a beagle bone and one of my panel controllers.

He seems to have network issues or cat5 cable issues or something completely unrelated.  There are people using Beaglebones running in bridge mode with matrices 10 times the size of his that are not having any issues.  150+ universes on a wired Beaglebone with error rates under 0.1% should be normal.  His error rate is signficantly higher with only about 12 universes.   As channel counts get higher, the quality of the actual NETWORK becomes a significant factor.   10 year old routers, old wifi, cheap switches, etc... will all impact things in very unpredictable and very hard to diagnose ways.

IMO, the best way to deal with high channel counts is to not transfer the data at all.   Local storage and a remote protocol to keep in sync.   

[gpoduska]

Kind of along the same lines....  I am a Google Chromecast fan..  It would be great if a virtual matrix could be run in a chrome browser tab, then we could cast the tab to a Chromecast dongle attached to a projector !!  This would make for wireless projectors without having to use a Pi setup! 

Is this feasible?