Hat’s off for trying this ambitious project. To be clear, led’s will be direct drive, one die at a time, no color mixing?
I’m guessing that the fet’s will end up on the backside again to allow for more vias from the micro; like for a 5 pin (0.05” O.C.) SIP programming connecter. Also a smaller B+ pad for a post or upside down spring. There is a 14 pin Pic with 4 hardware PWM’s (16f1503).
The channels are individually controllable, just one or any number can be on at any time and each can be at any level, with enough programming it’ll be able to make about any color.
I really want to keep the FET’s on the front, I like the clean back look but will be adding the pad’s/via’s for programming, instead of a premade header/connector I plan to use another identical non-built board with header pins soldered to it (with wires direct from the programmer) to make the connection for ISP without having to even remove the driver from the pill (that’s one of my main goals anyway)
Thanks man, this is actually the easy part, it’s the C coding that’s gonna be the real test for me… I need to use this specific MCU cause at first I’ll be running FW provided by 3ttereve till I can write my own. His UI uses a momentary switch however I plan to write mine to work with just a rear tail clicky cutting power for both on/off and mode switching so I have more host options. (I’m actually building this 17mm baby FET version specifically for a supfire A6 with a quad MCPCB and optic and W,R,G and B emitters so it has to be able to run without a momentary switch).
Another application for the rear clicky only FW is to eventually make a triple S series light with no white, only RGB, I want to make that my EDC!
People suggesting heatsinking for whatever method you use to drop the excess voltage for the red channel are on the right track. You’re talking about quite a lot of heat. It seems like dropping it using a power diode next to the edge may be the way to go.
On covering vias (“tented” vias), see post #426 in the Knucklehead Buck Driver Discussion Thread. I generally crank the limit value up higher than what I’d use and then manually set the Stop anywhere I want it.
The power supply cap should be on the micro side of the reverse polarity diode. SW+ should have a 0.1uf cap in parallel to the switch for debouncing (assuming the MCLR pin is made a digital input, and the internal weak pullup is enabled).
By way of explanation, please allow me to point you here: FETs and gate resistors - scope images.
Debouncing the momentary switch pin may not be a bad idea, but I don’t think we’ve been doing it on other designs.
1) I once built a board for a linear driver and had my supply cap on the led side, instead of the micro side, of the reverse polarity diode. That driver didn’t work properly until the cap was placed on the micro side. I can only assume that the micro was being starved of power in that configuration, causing a brownout or reset? I realize that this was not a fet driver, but any micro manufacturer usually suggests that a bypass cap be placed as near to the micro supply pins as possible.
Can’t argue with the comfychairs’s scope pics. There was no reverse polarity diode on my 2014 contest DD fet driver, so no boost voltage to worry about. No gate drive resistor either, and seems to work just fine.
2) The DD fet driver of my 2014 contest build had the problem of skipping modes, because of the noisy momentary switch. Using a 0.1uf cap in parallel (along with the internal weak pullup) took care of that problem as a “hardware” solution, pretty common practice, highly recommended. It can also be done in software too, but I’m lazy in that way.
I don’t disagree with anything you are saying. FWIW I’ve been advocating moving the decoupling cap as close as possible to the MCU while keeping it on the BAT+ side of the protection diode. OTOH if we left the decoupling cap in it’s normal place (right next to the MCU with nothing in between) and just used a really large cap across BAT+/GND we could maybe smooth the problem out that way too. Bottom line is that right now comfychair seems to be getting away with just using that single cap as described, at least with a single FET / LED running at moderate currents (AFAIK).
Got some time this evening to get on some stuff, here it is now, still need the CLK and DAT via’s
Now that I can see it let me point out the issues-
no bstop on the Batt+ pad then in the 3 large traces from the center pad to the outer arc for the LED+ via’s there will be a spot at 8mm out that’s exposed for an 8mm spring to make contact with the 3 traces, the pad is 5mm for either a 105c spring or a brass button.
I rearranged stuff to leave room at the top for the possibility of adding a bank of limit resistors. remember the old version of the 15dd has a limiting resistor on the FET’s GND leg that you had to jumper if you wanted full current, is that not the same principal I’m needing here?
No, not really the same I think. You’re looking at a tiny difference in voltage (like 0.1v) vs a big difference in voltage. (like 1.0v)
- I'd remove any extraneous traces from your PCB like the end of the semi-circular VIN ring (top right of underside of the PCB).
- Remove any right angle bends in traces (trace going to gate of Q1).
- Thicken the crap out of the trace from the source of each FET to ground (remember the 'drain' is LED-, the 'source' is GND on an N-Channel FET). They're thick already but they can be thicker...so make them thicker. These are VERY high powered LEDs.
- You're probably going to want to make the LED- solder holes a fair bit bigger as well because you'll need thick gauge wire.
- Ditto for the B+ holes.
- I cannot comment on the technical aspects of capacitor and reverse polarity diode placement, but I can say that from the day I first started designing LED drivers I have always placed my supply capacitor between V+ and GND on the MCU side of the protection diode.
- I have not had a chance to read through and digest comfy's FET/Gate scope thread, but from a safety perspective you should always include an inline and pull-down gate resistor when controlling a FET. The inline resistor will protect your PIC if something goes belly up in the FET, and the pull down resistor prevents the gate from floating; which can also cause problems. If I'm completely wrong on this please don't start a discussion in this thread about it, just tell me I need to read comfy's thread :)
That's all I have for now :)
Design updated while I was commenting. That's what I get for living in the future lol.
- Matt
Ok one more for now, the 8mm ring is not solid copper so its not connected to the GND programming via like it appears, it just displays like it is.
Well that’s kinda what my driver was, 0.1uf and a 10uf across the Bat+, except no protection diode. Had to go to a 1kHz PWM to reconcile the PWM register value to a percentage of max amperage. That could have been an argument for a larger cap than 10uf.
I’ve got the “everything looks good” from tterev3 and went ahead and ordered a set of the 1.1.4 boards below.
kinda got an XMEN look to it 
4 FETs, 4 claws…Sabertooth 
Depeding upon the Vf of the read LED and the battery voltage, you might be able to use two red LEDs in series. That way you get light out instead of burning off the excess voltage as heat.
If the goal is to use Luminus LEDs and do this in a Shadow SL3 then this will not be possible.
Where can one find one of these shadow sl3’s as a host or for under $50? This is a neat little light.
Cereal_killer, you’ve got some silkscreen text over copper. That isn’t going to print and OSH Park warns of unpredictable results. In my experience all they mean is that they can’t tell you exactly how far the fab will “auto clearance” silkscreen near the copper.
Also (I haven’t checked this) don’t forget to make sure that you make the lines thick enough or they’ll be auto-fattened. OSH Park claims that the silkscreen is done by what is effectively a 200DPI printer, so I treat the minimum thickness as 5mil.
Hey guys got some progress made, instead of piggybacking a driver on the stripped factory one which the switch mounts to I’m making a driver specific for this build which the switch mounts directly to for a direct fit, no need to piggyback onto the stripped factory board anymore. (BTW I already have the 17m PIC RGBW driver ordered to play with)
How do I get the “wings” so they’re part of the outline?
I decided to switch to the little baby fet because the AOD3400A Comfy found is capable of 5.7A, I figure 15A+ will be plenty, no need to use the big ones and since I’m using the smaller components (and I dont need the voltage divider circuit either cause the 85-20SU can do it internally) I went ahead and added a white channel to so this driver can fit a wider range of lights.
Do what I do…use the SOT-123 diode instead of the SOT-323, but that is cool driver, oh and de-stop those vias off the attiny Tools>DRC>Masks>Limit (set to something high like 50mm or whatever measurement metric you use) but make sure to go back and stop those VIAs you want to solder to by clicking the via and checking the stop box.
But that is a wicked cool driver man…like how you fanned the small FETs out to the side
To make the wings you will have to draw a milling ARC and attach it to the milling layer, then take the wing from one side, copy, mirror and attach to the other side
I find putting up a grid in view helps make sure I put stuff on the same level thru the driver
wight knows better than me…still an amateur on Eagle