No it doesn’t have to, here is the SC64LE tested by Zak, cooled so that it doesn’t thermally step-down, it has a 3A buck driver :
Yet it still outputs 70% lumens just before LVP step-down and the decrease only starts at ~80% of the runtime before step-down. Consider that the LH351D doesn’t have the lowest Vf, 219C and 519A are lower for example and would be even less affected.
Another thing to consider is that the drivers I showcased have lower dropout voltages, the TPS62867 one have a resistance less than 40mΩ, I haven’t measured the SC64LE but It could be double this amount (TLV62085, 31mΩ HS FET, 15mΩ RPP PFET, 15mΩ inductor, and they use fairly high sense resistor usually).
Buck-boost converters have much higher switch resistance and low switch current limit, in the end they don’t do much better at low Vin while being less efficient.
Looking to replace the driver in my S2+ (currently 5A driver, actually 5.4A driver). Looking for something with less current (2-3A ideally) and more efficiency (used as a bike light with 18350 battery).
Does anyone have any info on this BD39 buck driver? Anyone tried it and liked it?
Hooked it up to my bench PS to test and also used a clamp meter to test the LED current:
It’s rated 3V - 9V input.
It is indeed a buck driver but doesn’t hold a steady output current until 4.5V input (not fully regulated across the rated voltage range). For example, the medium (30%) mode is 680mA out at 4.5V input and around half that current (340mA) at 3.5V input. LVP kicks in at 3V on low and high but 3.2V on medium (strange). I measured the input voltage with a multimeter right at the driver rather than rely on the power supply (because of the voltage drop when at high current) [and because my PS doesn’t have remote voltage sensing].
Also, it seems to have some dreaded next mode memory madness. If on for < 4s then next mode is saved. If on for >= 4s then current mode is saved. Also means that if you try to change modes while on awhile (>= 4s) then you get the same mode, then have to click again to get next mode. I think I’ve seen this referred to as “quick-on mode memory” or something like that.
I’m now wishing I went with the Convoy T2/T3 driver.
Texas_Ace and I designed a buck for the Q8 long ago and we made a simple PWM output filter to convert PWM to a voltage for that. We never actually tried it though. Of course if you're trying to cram things in 15mm, two extra resistors and a cap aren't trivial. Edit: Oh and we had an LDO to work with as I recall. It was based on a tailcap mod to get 2S or 4S power.
rhd, most drivers exactly do save the present mode in eeprprom and use wear levelling. The reason for the OTC and OTSM is to allow more than one type of action. Short medium and long presses can all achieve different actions.
I started to see the light. It sounds like a way to measure ON time, for use in a forward clicky, as opposed to measuring off time in a reverse clickly? So to advance modes twice you click off, then short press twice (turning off before it can udo the mode advance) then click full on? I mean that could definitely work (sounds like it has), but it has some clear downsides.
You’re probably talking about the LM3409 buck LED driver, which DEL used for the BLF GT and Lexel in all his buck drivers and I think I recall seeing the TA Q8 driver you mention, which simplifies things a lot since being a LED driver it has a constant current output, set by a sense resistor and an analog voltage for analog dimming, voltage which can be a filtered PWM signal like you mentionned, or from a DAC, or just from a voltage divider/potentiometer if no MCU.
High Power LED drivers are not common though, this one is asynchronous (not super efficient) and not suited for 1S input, so the Q8 for example needs to be indeed modified to 2S+ to work with it.
Analog device has several of them, also asynchronous and 4.5V+ input, they do have a synchronous one, LT8613, but again not for 1S.
My schematic example allows the use of general purpose DC-DC converters/controllers which there are thousands of, from low current buck to multiphase multi kilowatt output boost converters. And just the same it takes an analog voltage, the constant regulation just needs an Op-Amp and a few passives.In my driver I use the DAC of the T1616 which saves on a significant amount of components compared to filtered PWM.
Anyway this isn’t PWM dimming, this is just the control signal. Often LED drivers support PWM dimming (like the LM3409) but they have special circuitry for that.
But it seemed to not be a real common thing in BLF/modder flashlight stuff maybe. Then mountain drivers were using PWM output not PWM-filtered control, and it seemed that's actually what most LED applications needed. Both have their use. I think in the end we put both on the Texas buck. PWM-filtered control is far better for efficiency. In fact I don't see a ton of point in a buck without it unless you just need to get from 12V to 6V. There's a little point otherwise, but not a ton. However, especially for something like a Q8, PWM-output is still useful for moon modes, and that's why we discussed including both, and I think did.
It seems the LM3409 has been around lights longer than the Texas Buck thread too...
Although again, I didn't see it mentioned much before the Texas Buck, but I'm not sure. Richard at MTN recommended it though, so he was likely using it, and we did look at some other things, including synchronous drivers. In fact I now see we did design calculations for the texas buck for 1S:1S (Battery:LED), even at 4.2V, as well as every other possible combination. We just didn't advertise them all in the end, or I think we didn't, because, well 1S:1S for example seemed kind of pointless for that light. But we even had designs for a 4.2V powered board with an LDO just to control PWM filter.
I modelled performance in detail for different switching frequencies, voltages (battery state as well as batt/LED configuration), and power outputs, all with inputs for specs (Rds on, inductor resistance, etc etc) so we could compare efficiency and ripple for different parts. Multiple types of loss were calculated for every component. It was all in a configurable spreadsheet so I could try specs for different components and see performance instantly across a range of configurations. The images of the results of that are gone.
I don't recall the exact reasons we went with nonsynchronous. Simplicity, using a recommended part, not trying too much at all once, practical details of implementing various drivers that existed at the time, may all have been part but not all of it (there's probably some discussion there). As I recall specs for n vs p fets for different drivers were part of the discussion based on drivers at the time. But there can also be improvements at light loads/high duty cycles, and we certainly selected a good diode. Looking at my old spreadsheet it seems it did quite well on paper, and even not terrible in 1S to 1S but I'd have to review that and my old comments more carefully. The version I have in front of me says it could do 4.2V to 3.8V at 20 watts with 12% loss, 8% at low output. It's over twice better at 2S:2S though.
Anyway, 1S to 1S bucks are of course possible. I know Sofirn uses some small ones in some of their lights, two-board designs though as far as the ones I've seen, and they even sell their drivers, at least on aliexpress. I guess they aren't the only ones. One reason BLFers maybe haven't raced towards them for taht, and maybe manufacturers too, and it's true in those lights, is they don't do as well at reaching direct drive. So hitting those lumen records (and advertised lumens) isn't possible with them. I'm with rhd, for my uses, I prefer regulation. In 2S to 1S that's not an issue, and that, or 4S to 2S, were the main idea for the Texas Buck I think. There's also the fact that though really optimizing buck designs is a lot more work. There are a lot of types of losses to consider (even gate drive losses are quite significant, but are a tradeoff, like everything) and they change with operating mode and frequency choice, and in the end, 7135s are usually good enough to move on.
But some of the best drivers chips aren't so old, and maybe it's just waiting for someone like rhd to update us with one that keeps up the times.
Oh, and indeed the 1616DAC is a nice solution. 1616 was only barely coming into use in custom drivers then (2016) and it's quite possible that still just nobody has fully taken advantage of this avenue for reducing parts for a small hobbyist buck driver.