budude wrote:It would be neat to integrate the BAM/MIBAM dimming into a DC version of the code for the Renards. It produces much smoother dimming and requires less updates/power from the PIC.
+1
Josh
One of my unstated goals was to leave the powerline as clean as possible. This means making the load appear as close as possible to a resistive load to the power line, i.e. a power factor close to 1.0. This would probably be difficult and expensive to attain in a hobbyist situation, but I think that this is something to keep in the back of ones mind. It should be noted that none of the techniques (except the sine-wave dimming discussed a few years ago) come close, neither the forward or reverse-phase dimming, nor using standard rectifiers for converting to DC.
As far as BAM goes in DC application, I don't see any big reason to adopt it in the designs that we are using. The current PICs are plenty fast to get by without it, and there would not be any significant power savings in the PICs themselves to write home about. And I remain far from convinced that dimming would be any smoother using that, unless there is something that I'm missing. The current Renard DC refresh rates are 120 Hz, more than fast enough. The only possible benefit I can see is if one could easily linearize the light output by going to 10-bit or 12-bit PWM. Anyway, there is nothing stopping someone from trying to implement such a thing, either as a test or for real.
P. Short wrote:Anyway, there is nothing stopping someone from trying to implement such a thing, either as a test or for real.
Already done actually - the code has been integrated into the PropController DC dimmer code - all I can say is the dimming was subjectively smoother compared to standard PWM code. How much? I don't know - perhaps not enough to worry about but even if it's a small amount - why not? A PIC-BASIC version exists but unfortunately, I don't have that so the algorithm would need to be reverse engineered. Also - if I spend less time updating the outputs, I can spend more time doing other functions within the code (whatever that may be).
I also think it may be worth pursuing.. From a cost perspective, it will be hard to be the basic Triac Design though. I have been thinking about some possible Reverse Phase / Sine Wave Dimming Designs also. For AC you usually need two MOSFETS or IGBTS for each channel unless you full wave rectify the incoming AC and run all of your loads off that instead. Doing this, you should be able to get by with one MOSFET/IGBT per channel.
The interesting thing about either of these two methods, is that they both allow for a BAM algorithm to be used. Sine dimming also has some other goodies, like reacting to shorts, custom output voltages, etc.
P. Short wrote:The current triac-based approach generates a lot of garbage on the AC power lines, which I think is the cause of some of the current 'flicker' problems, and is likely to get worse as the number of channels increases. Some of the flickering problems may be helped by using snubbers, snubberless triacs, and MT1-Gate resistors on the SSRs, but these solutions don't reduce the amount of noise going back into the AC lines.
Is the cause of the noise that the AC voltage is turned off (or on) suddenly in mid-cycle? If so, would a slower fall or rise time reduce the noise? (at the expense of making the dimming logic more complex)
It seems like 120 VAC strings are still a preferred/convenient form factor even for LED strings that are not individually addressable, in which case the need for AC dimming may not go away any time soon. But the amount of current to be dimmed per channel would drop (multiples of 20 mA for LEDs instead of multiples of ~ 150 mA for incandescent). Would lower current open up new possibilities that were not much more expensive or complicated than the triacs?
don
