Actually it was a variation of V2 which was 20mm in diameter. In order to get the board down to 17mm, I had to lose the through holes and add a resistor/capacitor pair on one of the MCU pins that I had for off time switching. So this 20mm version was basically a one off.
Yeah the LT1761 regulators are not cheap. The reason I think they are expensive is because of their wide input voltage range. There are not a lot of regulators (that I could find at least) that support our potential 3-18v input range. If anyone finds a cheaper one that is just as capable, let me know. It has to be a LINEAR voltage regulator though - no way I can fit a switching IC on there. This regulator also has reverse polarity protection which is not always a feature of these devices.
To be honest I went with what was recommended in the datasheet because I wanted to get it right the first time. The 1uF input cap is actually shared with the MCU input as well. I'm not really sure if it's required seen as the IC is supplying a 'clean' voltage already, but it's become habit to filter the input to the MCU to avoid any hiccups. This all said, removing 1 capacitor is going to have an almost negligible impact on the total cost of the boards. I think the best way we can reduce cost is to buy in bulk. I know it's been mentioned previously, but if RMM sells kits he could get the cost down in $5 or so depending on how many he is willing to stock. I'd do it myself, but my location makes it pointless because shipping is a killer.
[quote=wight]
That's all speculation and not recommendation on my part. I don't want to play armchair general too much with Mattaus's hard work. It's an expensive driver to build though, and if we can get the cost down without too many compromises it'll make all of us happy I think!
[/quote]
Meh. Any input is valid input provided it's not just to stoke flames. If it improves the design or provokes good discussion then I'm all for it :)
I’m thinking of re-doing my M3 on the L2P with a double or triple cell configuration and using this working Knucklehead. I’ve also asked Buck if he’d make a 2 cell tube for the Convoy C8 so I can try it in a bigger light with a more decent reflector, might go with an MT-G2 in that one as it’ll have more mass to handle the big boy.
Will let ya know how this M3 works out. Should be pretty good, maybe not as potent but will last longer. Which is really what this one’s about, right?
I've got boards and some components on the way which should save some $$$ for those who only want to build a few of them. May be a week or so until I get around to listing everything.
I did note the reverse polarity protection as I was looking at the datasheet, this regulator definitely looks like a great choice. If someone finds a better choice for this application I’ll be astounded.
Of course I don’t think that eliminating 1 cap is going to help the bottom line :-). I was thinking of anyone who doesn’t have cheap access to something like Digikey or Mouser and has to rely on buying components from China in 100pcs for $3 increments or similar. The original BOM calls for every capacitor on the board to be different! Eliminating the 0.01μF, changing the 10μF 0603 to 1μF 0805, and changing the 1μF 0603 to a 0805 reduces the number of capacitors and the number of varieties (3 varieties and 5 total pieces). That’s assuming it actually works with those changes…
That's actually a very good point. When I get time (or if anyone beats me to it) it wouldn't hurt to try bridging the bypass capacitor pads, and changing the input and output caps to match, and see what happens. It might not give the best performance on paper or in simulations, but if it works then it would surely be worth it. Nearly every time I post questions in electronics forums I get 3 or 4 people telling me that I'm doing bad things to LEDs, that it won't work, or its a really horrible way to go about it. Practice shows us what we can do. Fiddling with the voltage regulator cannot damage the buck IC. Worst case is you'll lose the voltage and MCU ICs.
EDIT: I should point out that the components on the buck side are NOT interchangeable. Those capacitors were chosen for a reason. The set resistor can be changed to alter maximum output, and even though people are experimenting with different inductors, parts that are as similar as possible to the specified Coilcraft model are HIGHLY recommended.
Yesterday I had it working reliably on the bench. Easy repeatable results. Measureable results. Tonight I put it in a light. Can’t get normal patterns out of it to save my life! Lo doesn’t want to work, the modes are catatonic, very iffy, oddball behavior.
I have the ground pad soldered to the pill, made sure the inner step was very narrow so it wouldn’t touch anything on the inductor side of the board, even notched that step to make absolutely certain that the positive lead is plenty clear.
Going to take it apart, again, and see if it performs on the bench in an attempt to figure this out.
Turns out that in this light the ground was an issue. I got a decent ground and had it working fairly well. HAD. The low was flickering still so I reflashed it with the moon mode enabled as per WarHawks > instead of == and bumped moon from 1 to 8, bumped the 6 of low to 12 and left everything else alone. Put it together, came on nice and low, stepped up to another low mode then quit.
The 8HSOP is fried, again. Very similar little wet melted dots on top, to the right of pin one and sort of below it.
I flashed the Qlite with this firmware but moved the fuse back to 75 from 65 and put the 16 chip Qlite back in this light. It works smoothly and flawlessly with 5 modes. Notably, the exact same firmware is working fine with the 7135 chips, with settings of 8, 12, 25, 120 and 255. This yields .01A for 15 lumens, .10A for 33 lumens, .37A for 88 lumens, 2.47A for 483 lumens, and 5.77A for 1028 lumens, and of course this is on one cell, a rested Samsung 20R.
So I’m wondering why the firmware works fine with the ATiny13A in the Qlite, flashed with the same firmware but fused for a higher PWM, yielding similar results in power on a relative basis with no apparent grounding issues. This is the 3rd 8HSOP I’ve fried. Not to worry, I’ve got more coming so I can fry a bunch of em.
The AMC7135 chip and LED2001 IC are two VERY DIFFERENT beasts. Comparing them is not a good idea. It's like comparing a bicycle and an LMP1 race car. One is insanely more complicated than the other. The AMC chips are very robust and work under varying conditions because they are simple. The trade off is that they are very inefficient. The LED2001 chip is a lot more temperamental, but the advantage is it is insanely more efficient at converting power into actual light. We're talking 30%+ gains.
AMC chips are easy to work with because they are on or off, there is no difference. They are effectively direct drive, burning off any extra current as heat. When you PWM then, they just turn on and off - you get max power for every on period but the effect is a dimmed output for any PWM levels under 100%.
The LED2001 regulator is a switch mode device. It's constantly switching itself between buck and boost to maintain the correct output (even though the overall effect here is bucked output). If you then PWM the IC to change the brightness, you are turning the device on and off....while it's busy turning its internal processes on and off. It's an on/off change over the top of another on/off change. If power to the IC is momentarily cut, even for a fraction of a second, it could throw the whole process off kilter. Because of the simplicity of operation of the AMC chip, a small cut in power may have an unnoticeable effect on the output.
Basically put, make sure your power path is flawless and once again the firmware could be the culprit. If I'm reading the results properly.
I wonder if the solution would be to have the buck/boost circuit run independent of the gate feeding the emitter, as in put one of those small FET’s (from the 15DD) between the buck/boost ckt controlled by ATtiny. A bit more complex but this way you aren’t shutting down the component doing all that power conversion
