FPGA based pixel co...
 
Notifications
Clear all

FPGA based pixel controller

85 Posts
18 Users
0 Reactions
33.5 K Views
timon
Posts: 2149
(@timon)
Noble Member
Joined: 16 years ago
ronp5683 wrote:The E1.31 definition requires that the receiver support IGMP messages. There is no requirement for supporting ARP messages. Address Resolution Protocol is basically a way for a computer to find out the MAC address of a device at a known IP address. So the computer sends out an ARP request to an know IP address (example 192.168.1.100), and the device will respond with an acknowledge that contains it's MAC address. Once the computer has the MAC address, it can open communication with the device. This is not required for multi-cast, where a generic message is sent out onto the network for any one to receive.

The internet group management protocol (IGMP) is simular, except it is used by routers and switches. The routers and switch will send out an IGMP query. Each device will respond with what internet addresses that they are interested in receiving. The router/switch will build a table of where to send these packets as they arrive.

The Vixen software does not implement these messages, and I have a dedicated computer for my light show, so there is no router switch involved. I have not implemented these messages, because they take up logic and space. I would rather spend my limited resources on supporting as many universes and pixel types as possible. However, if there is enough interest, I am willing to implement these features.

I understand saving space for more pixel types but I think in the long run it would be good to add both ARP and IGMP.

What would be interesting is a way to not have to include all of the pixel types into array but be able to download only the ones you need. Not sure if that can be done without doing a full rebuild but it's something to think about.


Reply
Posts: 238
Topic starter
(@ronp5683)
Reputable Member
Joined: 14 years ago
timon wrote:I understand saving space for more pixel types but I think in the long run it would be good to add both ARP and IGMP.

What would be interesting is a way to not have to include all of the pixel types into array but be able to download only the ones you need. Not sure if that can be done without doing a full rebuild but it's something to think about.

The ARP is only required for uni-cast, which I do not support. I was being honest when I said that I did not support IGMP, which is my only deviation from the specification. If you buy my controller, I will support your needs. As it is the controller works fine hooked up direct to the computer or through a hub. FPGA's work fine for pixel controllers, but for doing messaging over Ethernet, that is better left to a processor.

The limitation on my controller for more universes is memory, and I am tapped out at 32. So removing pixel types will gain very little. Also, if I comment out the other pixel types the compiler will remove the code, which is a 5 minute compile. If you want to play with a FPGA's driving pixels, buy the 8 universe controller. You can update the configuration without changing the program in the flash. Or basically you can play with code through RAM and not lose the original programmed code.


Reply
Posts: 238
Topic starter
(@ronp5683)
Reputable Member
Joined: 14 years ago
BufferBoard

Have some buffer boards on order, so you can drive pixels directly from the board. With four of these buffer boards you can control a sign up to 32x170.


Reply
timon
Posts: 2149
(@timon)
Noble Member
Joined: 16 years ago

I really do like what you done using the FPGA however I've been thinking that there are things that get sacrificed with a pure FPGC configuration. I'm thinking that the combination of a FPGA and a standard micro-controller such as an ARM Cortex M3 I think is an even better combination.

The FPGA is far superior to a micro-controller in handling pixel data especially the three wire ones that really need to be spot on when it comes to timing. Micro-controllers on the other hand are better in handling the protocol need with the Ethernet traffic associated with E1.31, E1.33 (RDMnet), ARP, DHCP, IGMP and all the additional network protocol who's timing is not near as critical. The cost difference adding a processor such as the LPC1768 is minimal and the gain is great. A single LPC1768 should be able to front end at least two FPGAs handling 32 universes each.

Just a thought and my humble opinion.


Reply
Posts: 238
Topic starter
(@ronp5683)
Reputable Member
Joined: 14 years ago

I had thoughts of doing a daughter card for Raspberry PI, this would basically give you what you want for a little more money. However, you get the same thing with a PI and using the Ethernet output. I have created a product that supports more lights, 32 universes versus 12. It also has lower latency, time from computer sending data till the data reaches the lights. A higher update rate, I can support 32 universes with an update rate of 200 hertz or 5mS. All this for about the same price as my competitors products. There hasn't been an overwhelming demand.

I also can do 64 universes on my same motherboard (3 wire only) with a different FPGA card from waveshare. I also could do a product that would support 1920 universes for about $100 (this is basically VGA 640x480), but until I figure out what the market really wants, I am content with my current product.

If you want help doing a processor/FPGA card I am willing to review the schematic and do the FPGA code.


Reply
Page 17 / 17
Share: