OK so you are talking about a pin to pin compatible swap out for the PIC
Having a controller that has 12 bit intensity capability does nothing if the tool can't produce it.
Doing a math calculation on the output value to linearize 8 bit to 12 bit does nothing. Vixen 3 has already calculated the intensity at that instant in time. Mapping from a 8 bit value to an equivilent 12 bit value will produce the same output intensity.
What would be GREAT, is if Vixen 3 would allow us to state the number of bits to use for any given channel. If Vixen could - say dimming over 1 sec - calculate its output value based on a 12 bit value instead of 8 - then you WOULD see a difference in the illumination output --- it would be smoother by a factor of 4.
So --- dimming levels greater than 256 won't produce any meaningful lighting change unless Vixen 3 can increase its output range.
Time and the future will tell.
P. Short wrote:First, there are 256 levels of dimming in the Vixen design, not 100. When you say, for example, 75% in Vixen, that is stored as 255*3/4, or 191 in the Vixen output data file. That 191 (or is it 192??) is what becomes transmitted to the DMX or Renard boards.
I understand the conversion, but at 256 levels of dimming, isn't each percentage point is worth 2.56 "level" units? So there really isn't a 1, 2, 3, 4, 5.... 256, it's 0, 2.56, 5.12, 7.68, 10.24, etc. There is no 1, or 4, or 6, or 8 or 9, 11, etc. 50% is 50% whether there are 256 dimming levels or 4096, the relativity between 0 and 100% is the same.
So a 5 second ramp-up at 50ms timing will use 100 cells from start to finish, each cell value would (theoretically) be 1% and at 256 levels of dimming, each cell increment will be equivalent to 2.56 "levels." In this example, the only way one could actually utilize all 256 levels of dimming were if the ramp operation was 256 cells in duration or nearly 13 seconds long at 50ms timing.
In the same 100-cell ramp operation if there were 4096 "dimming levels" each cell increment would be almost 41 "levels", not 1-2-3-4-5-6-7, etc. and one would need a ramp operation of nearly 205 seconds (3 minutes, 25 seconds) to actually utilize all 4096 levels at 50ms timing.
What it does mean, of course, is that the faster the update timing is, the more of these available dimming levels one can utilize which ought to make for a smoother ramp effect. This might create wonderfully smooth effects with pure DC current, but A/C would still be limited in what could be usable insofar as the ZC occurs 120 times per second. The other issues affecting it are, of course, the PC's CPU speed and the number of channels in the sequence.
I guess the bottom line of my point (I think anyway...) is that the number of dimming levels may not be as important as it sounds. 4096 sounds impressive until you realize there's really no practical way to utilize that many. Dimming curves, on the other hand, could be a very useful tool to spread out the critical portion of the effective change in illumination for LEDs, and in this context, it seems to me that having more illumination levels would be a benefit.
It actually can make a difference - since the steps can be smaller internally to the lights. There's a post somewhere by jstjohnz (Jim) with 12-bit pixels and he compares them to 8-bit pixels yet data is coming from an 8-bit DMX stream.
edit - - here is the video: http://www.youtube.com/watch?v=XlmVjcnP3IM
My concern is not so much about the smoothness as linearizing the output. For example, with the current Renard setup and incandescent lights most of the action is at the low end. As you increase the intensity from 0 to 100%, it starts out very dim, and slowly becomes brighter. At some middle range the brightness increases rapidly, and at 75% the brightness has reached maximum value, and (apparently) changes very little as the intensity goes from 75% to 100%. My objective with the 10- or 12-bit dimming would be that the apparent brightness increases 'linearly' as the input values ramp smoothly from 0% to 100%.
