It was a normal troubleshooting weekend. Much time spent and little forward progress. However, I have two blinky lights on the FPGA board. One at 5pps to know the board is working, and a second to light when Ethernet data comes into the mapped universes. Tested the serial output at both 57.6K/115.2K and it is working, also varied the number of bytes (channels) and the output matched. Also verified that the test function outputs correctly for turning on all of the lights to 1/4 intensity (only the first 510 to support RGB), then all 512 off. So maybe this week I will work on the Renard. I will keep the two byte outputs in as long as I have the resources.
Found that Vixen 2.5 with the Ethernet DLL will not output to the universe if there is no change in the data. I had in the design a function to clear the outputs if no Ethernet was received for 2.5 seconds. Is this desirable?
ronp5683 wrote:Found that Vixen 2.5 with the Ethernet DLL will not output to the universe if there is no change in the data. I had in the design a function to clear the outputs if no Ethernet was received for 2.5 seconds. Is this desirable?
The most appropriate idle period would depend on the sequence and the props. Some songs could easily have a section of lights with a static pattern for more than 2.5 sec.
I think Vixen 2.x works that way for other (non-ethernet) plug-ins as well.
There are some easy work-arounds to force output to be sent each time. For example:
- use a dummy channel and set a slow ramp (say, increase by 1 each frame, and then wrap)
- the FxGen plug-in has a "timing track" generator function that will set 3 channels to a 24-bit value that changes for each frame in the sequence; you could add it to a sequence, run it once, then remove it again, leaving the timing track channel values there
- save a Routine with a changing channel value and then just load that as needed
- just change a channel value by, say, 1 every so often to force Vixen to resend the data
Or, make the 2.5 sec idle period configurable so the user can choose a value appropriate for the song and sequence.
don
The specification for E1.31 requires that a packet be sent at least once a second to keep alive the output lights. I am going to leave in the off function at 2.5 seconds as required for the specification. Seems like there are work-arounds for this issue.
Tested Renard tonight and it is working. Leading characters are 0x7E 0x80, followed by the data. The 0x7D, 0x7E, 0x7F values are supported with the two byte outputs.
DMX is the next output to be worked.
Quoting from the design document:
If a packet is received before the current one is sent out. The resultant output will be partially the old packet and partially the new packet.
I don't think that .this is a good thing. The reason that I say this is that it is a triplet of bytes that determines the color of a pixel, not just a single byte. This means that a pixel might end up with a color that is neither the old color nor the new color, but something that is potentially entirely different. This might end up persisting for some time (long enough for the odd color to be conspicuous) if there is a scheduling hiccup in the controlling computer for some reason.
Preventing this from happening is probably not easy to implement, but would be a good thing to consider.
P. Short wrote:Quoting from the design document:I don't think that .this is a good thing. The reason that I say this is that it is a triplet of bytes that determines the color of a pixel, not just a single byte. This means that a pixel might end up with a color that is neither the old color nor the new color, but something that is potentially entirely different. This might end up persisting for some time (long enough for the odd color to be conspicuous) if there is a scheduling hiccup in the controlling computer for some reason.
You are right that this is a bad thing. I questioned this earlier on the forum and received no comments. Ignoring new messages while you output the universe data is closer to what is in the specification. They really want you to look at the priority and use the higher priority. The problem is that I start outputting a message as soon as I start to receive it. So for me I can easily ignore messages until the output is done.
I have completed the code for Renard at 57.6K/115.2K, generic serial at 57.6K/115.2K and DMX. I really can't test the DMX, so I checked it looking at the output on an o'scope. So I am done with coding for a little while. I have an 2801 string on order and I am thinking of buying a DMX controller.
So for now, I will update my motherboard PCB and order some soon. So I am asking if anyone is interested in a copy of the the board, and is willing to try out the design. They will have to buy the FPGA, Ethernet board, and programmer.
Also I left the buffers off the motherboard, is there a favorite buffer for output to universes?
