What is not clear to me is how DMX could be involved in say 10, or more, bit dimming. Does one use more than one channel for the additional bits? I couldn't find any DMX info to do more than 255 bits for dimming, except by using an additional channel.
For controllers, there are active developments going on for DMX (Chilloutdoc, myself, and others) that are doing the traditional DMX
versus "modified data stream" that the REN series does.
As Budude notes, it's a fairly simple barnacle to any board to convert it. Though I think you would have to cut traces as well, or risk Tx collisions.
budude wrote:For Renard - simple enough to "fix" - just jumper the in/out connector D+/D- lines and you have THRU to all downline controllers. The Renard/DMX code runs at 250Kbps as well - it's as DMX-ey as anything else...There are a few other controllers - most from RPM that act as standard DMX controllers but they aren't very common - take a look at the Wiki.
LabRat wrote:For controllers, there are active developments going on for DMX (Chilloutdoc, myself, and others) that are doing the traditional DMX
versus "modified data stream" that the REN series does.As Budude notes, it's a fairly simple barnacle to any board to convert it. Though I think you would have to cut traces as well, or risk Tx collisions.
or just pull the second 485 chip since it's not being used... :wink2:
Bah... you are so correct. I was still thinking the single RS485 chip solution.
LightUp wrote:What is not clear to me is how DMX could be involved in say 10, or more, bit dimming. Does one use more than one channel for the additional bits? I couldn't find any DMX info to do more than 255 bits for dimming, except by using an additional channel.
That could be possible, but it would then affect the upstream sequencer and reduce compatibility; as you say, it's not part of DMX and not in any way expected.
It's worth looking at the goal tho. There may be another way.
Others have pointed out that the eye is not going to be able to resolve even 256 discernably different levels of brightness, so we don't inherently need more than 8 bits of control.
The problem is the non-linear natures of the eye, incandescent bulbs, and LEDs. To compensate for this one can use arbitarily complex light curves, or simple gamma correction. The difference between 0 and 5 may be huge (or 1 and 6), but 250 vs 255 is imperceptible. And lights respond non-linearly in ways that may come closer to compensating for the eye (incans) or not (leds).
Some software (eg: Vixen) tries to improve this using 8 bit to 8 bit lookup tables - but you wind up with too coarse granularity at the low end, and unneeded resolution at the high end, at the output end of the table (8 bits resolution). Sending out more than 8 bits per channel would be one way to improve this as you suggest - but it's non standard and incompatible.
The other way is to have an 8 bit to 10 bit (or 12 bit or whatever) lookup table in the DMX controller itself (in nonvolatile memory). Two existing examples would be the Lynx Express AC controller (which I believe has 1024 time slots per half-cycle of AC rather than 256), and the e681 pixel interface (which can use the 3001/3005 chip based pixels which have 12 bit internal resolution). The former uses custom 8->10 bit color curve lookup tables which RJ calibrated by testing some actual common light strings (incan and LED) to give consistent results; you configure the Lynx Express with which table to use, picking the canned table which works best for your lights (or he can add new ones to the code if he's willing).
In the e681 pixel interface (which uses e1.31 ethernet as input - which is basically an encapsulated DMX512 protocol with support for multiple universes, ie: more than 512 channels), an 8 bit to 12 bit lookup table based on a configurable gamma factor is used. While not as customized as the Lynx Express tables (which being built empiracaly can have arbitrary shape), in practice the gamma correction is reported seems to work pretty well (if you pick the right gamma value). In that case you pick one gamma per four outputs (of 16). I forget how many different curves the Lynx Express supports for mixed strings.
Your photo editor probably has a gamma corrections slider if you want to see the effect.
The cool thing is - you configure the controller in each of these cases based on the lights you have connected to it, then it responds more "as you would prefer" to standard DMX 8 bit values. Vixen or LSP have no need to know which curve/gamma is used or what kind of lights are connected. The 256 levels of each DMX channel are more than enough, once the needed non-linear scaling has been done. You have essentially made the lights appear closer to "ideal" to the software, to the DMX protocol, and even to a DMX console (with sliders like a DJ would use).
(You obviously disable the software 8 bit to 8 bit look up table gamma in your sequencer software for such channels).
I'm hoping the new controller will support something like the Lynx Express on the physical level (ie: more than 255 time slots) with a configurable onboard lookup table from 8 bits to the number of slots available. If you can download the lookup table, you can choose to calculate it (on your PC) as a gamma based curve, or any other arbitrary curve. It's all software for configuration, but the actual low level lookup has been pushed closer to the lights being calibrated.
