Komby Pixel Optimiz...
 
Notifications
Clear all

Komby Pixel Optimization

14 Posts
7 Users
0 Reactions
5,596 Views
Posts: 204
Topic starter
(@libby2cm)
Estimable Member
Joined: 14 years ago
[#34790]

Perhaps this has been discussed elsewhere, but I had a thought regarding tweaking the approach to pixel adressing to allow addressing a pixel as a unit rather than the individual subchannels, thus allowing 512 pixels on a universe?

To be compliant with the DMX protocol I would imagine this would be achieved via abstraction somewhere.. Maybe v 2.0 of the rfshowcontrol protocol..?

I ask because as I add more pixels each year, maybe even RGB signs, I'll eat up free RF channel space in no time with only 170 pixels in a universe.. Plus id need like 30 transmitters.

Another idea might be to port the xmit code to run on a more powerful device like a r-pi and modify the code to allow running multiple logical controllers (maybe each with own VIP) mapped to multiple Nordic modules.

Probably would have to bump up receivers as well since I doubt a 328 will be fast enough to drive 512 pixels. I'd imagine rf speed would have to increase as well to account for increased addressing of 512 pixels.

Hopefully this will spark a discussion.


13 Replies
sporadic
Posts: 1199
(@sporadic)
Noble Member
Joined: 14 years ago

First thing that needs to happen is that E1.31 and the notion of universes needs to go away. It's just a hold over from the original DMX protocol and stuffed into the E1.31 layered burrito. A simpler UDP protocol that allows addressing of contiguous channel blocks would allow the endpoint devices to process data more efficiently.

On the RF protocol side, I started a discussion here a while back, but it never took off - http://diychristmas.org/vb1/showthread.php?2365-Expanding-the-nRF24L01-protocol-and-getting-past-the-512-channel-barrier .

Driving pixels on the receiving end is more of a memory issue when it comes to clockless pixels like WS2811's. You need enough RAM to buffer the entire pixel stream. On the transmitter side, you're really only limited by bandwidth of the signal chain once you remove the heavy lifting (like E1.31 processing). It's fairly quick and easy to continuously pump out data via the nRF if everything is nicely formatted to begin with (serial Renard stream in my case). In my current setup, I'm pushing ~1400 channels per transmitter from 480kbps Renards feeds (via E1.31 bridge), going over the air at 1Mbps. However, there's no Arduino stuff. It's done from scratch and running on AVR Xmega E5 micros at 32Mhz.


Reply
Posts: 352
(@pmscientist)
Reputable Member
Joined: 16 years ago

Running off an RPi with a nrf24 hooked up is a good idea. Assuming a 40Hz refresh rate, the pixel limit at the default 250kbs data rate is 260 pixels, though it''ll be less in practice. You can use a higher bandwidth mode, up to 2Mbs, but there's some tradeoff in range.

To some extent though, seems like 30 universes is a scale the nrf24 radios aren't really intended for. Sure, it's likely possible to get there using just those, but there's probably a more efficient way.

Personally, I use WiFi to get data to a somewhat central point that has some sandevices controllers, which can handle up to 16 pixel strings. From there, I use a komby sandwich off the same wifi router to run some elements that are hard to reach and/or are long distances from the wired controllers. This allows handling of the overall high bandwidth needs, while still leveraging the flexibility the komby system has.


Reply
jchuchla
Posts: 8698
(@jchuchla)
Illustrious Member
Joined: 15 years ago

I'm not sure what you're getting at here. Changing the protocol or addressing scheme wouldn't change the amount of pixels that would be supported. In the end, bits are bits. It wouldn't help anything by changing from three 8 bit channel values to one 24 bit pixel value.

Rereading your first statement above, you're suggesting that we address all three lighting channels within one DMX channel? That would result in a HUGE DECREASE in dim levels and color options. You'd end up taking 8 bits, and cutting into 3. Which you can't do evenly. So that would mean that you'd have 7 dim levels for 2 of the colors, and only 3 for the 3rd. That's not at all useful. That reduces your total possible number of colors to only 147 as opposed to the 16.77 million we have today. The other way you can do it in the same channel space is to time multiplex it. Send the 3 8 bit values in sequence in consecutive frames. This would result in 1/3 the refresh rate overall. So that's not particularly desirable. This wouldn't work so well though because we already get a lot of packet loss. We'd need a way to make sure the receivers get all 3 packets, and only update the lights if they successfully received all three in sequence. I'd bet that that would give us drastically reduced performance because the chance for 1 lost packet in every 3 is pretty high.

Upping the microcontroller wouldn't help us transmit more or faster with a given radio. Even with the 328 we already have to pause between transmissions because the radio requires it. A faster micro would just help us drive more radios, or process more data. But we're still limited by the radio pipe in any given radio. You can only ask so much of a $1 radio module.

A better approach would be to reduce the realtime bandwidth requirement. One approach to this would use distributed players, where each "receiver" holds the sequence data, maybe it's uploaded in advance (slowly) via the wireless link. Then during showtime, the wireless link would just be used for sync. Think FPP multisync but with an nRF based network.

Theres only two ways to get more channels: faster bitrates, or more bandwidth.


Reply
Posts: 350
(@kmalsom)
Reputable Member
Joined: 14 years ago

I don't think the RFShow protocol limits to 512 channels does it? I thought the limit was 4584 ( http://learn.komby.com/wiki/58/configuration-settings#HARDCODED_NUM_CHANNELS ) and that it depended on the speed that you were transmitting. And that if it was transmitting RENARD data instead of DMX, it could handle a lot more channels than 512.


Reply
Posts: 2541
(@lightup)
Famed Member
Joined: 17 years ago
jchuchla wrote:...Theres only two ways to get more channels: faster bitrates, or more bandwidth.

Interesting statement.
Wouldn't "faster bitrates" imply greater bandwidth? I suppose you can have wide bandwidth and a very low bitrate. I would see your statement as one way. 🙂


Reply
jchuchla
Posts: 8698
(@jchuchla)
Illustrious Member
Joined: 15 years ago

It's important to remember that we're not pioneers in this technology. There's professional systems that have been doing this long before us. Good ol fashioned standard definition video was able to very efficiently address and control just over a million pixels. And it's been doing it wirelessly, over the air for over 75 years in the form of television broadcasts!

There ARE higher bandwidth protocols out there that are better suited for higher channel counts. Most of those big digital billboards get their data from a computer like device via DVI. In these applications it's usually then converted into a balanced serial digital signal (very high bandwidth, 6Gbps+) for transmission to the display itself. Yep I'm talking plain old DVI like monitors use. Why? Because it's designed for millions of RGB pixels! There's a certain threshold where lights are no longer individual lights, they become video.
The ACN protocol itself is envisioned for very flexible communications, including high channel counts. Streaming ACN is just one possible implementation. A full blown ACN implementation is probably a better match for the pixel needs.


Reply
jchuchla
Posts: 8698
(@jchuchla)
Illustrious Member
Joined: 15 years ago
LightUp wrote:Interesting statement.
Wouldn't "faster bitrates" imply greater bandwidth? I suppose you can have wide bandwidth and a very low bitrate. I would see your statement as one way. 🙂

Not necessarily. Sure, faster bitrates in the same frequency space would allow for greater bandwidth. But what I meant was overall communication bandwidth. With the Nordic radios, raising the bitrate also increses the width of the RF channel, therefore reducing available channels. It's pretty much a wash. faster bitrate = less usable channels & less distance. Lower bitrate = more available channels & more distance.


Reply
kingofkya
Posts: 2165
(@kingofkya)
Noble Member
Joined: 15 years ago

And we could also add simplistic compression of some kinda but again more power required on the already taxed pixel controllers. So i guess if you want to run way too many pixels bit the bullet and run the ethernert. Copper will always beat rf for speed.

Still it all boiles down to the fact some one has to start it to get done. (and possible take it too far like komby did on the wm2999...) I mean we really talked about about every idea i have seen in the past year, but we ignored the ideas because it needed to get done.

For example: one idea i had early one that will likely never get done... Why cant we upload a show to the controllers and then just send out a wireless sync pulse to keep everything in line. In theory you could do that with just one channel because you don't care how long it takes to upload as long as it runs smooth and in sync. This method would also be much kinder to the rf spectrum rather than just blanketing the 2.4ghz spectrum when your show is running.


Reply
jchuchla
Posts: 8698
(@jchuchla)
Illustrious Member
Joined: 15 years ago

Since we like lights anyhow, there's always wireless communication using light. These work very well:
http://www.digitalairwireless.com/outdoor-wireless-networks/point-to-point-wireless/laser-fso-links/geodesy-fso-laser-links.html

--Jon Chuchla--

Sent from my iPhone using Tapatalk


Reply
Page 1 / 2
Share: