Texas Avenger "TA" Driver series - Triple channel + Bistro or Narsil + Clicky or E-switch - The Ultimate open source driver!

Just a amateur hack here, so please don’t take my statements to follow as fact. I would think the LDO will drastically change the situation. Check out the data sheet for your LDO. I’m fond of using the LM2936 and it recommends a 10uF Cap on the output pin. Check out this diagram from it’s data sheet:

The LM2936 data sheet has no limitation on the cap size, just a minimum. It also recommends a certain ESR. A .5 to 1 ohm resistor in series with a typical SMD Cap will accomplish that for the LM2936. Here you can see I have 20uF in caps a 4.7uF resistor (Way to high, was just trying it out) on DD driver for 4S cells. I used 47K (R1) and 4.7K ohm (R2) resistors on the voltage divider.

The LM2936 is way too big for our drivers (can handle 40v on input). Just giving a for example case for you to think about.

I figured it would effect things a fair amount but I am not sure how exactly. I guess one thing I didn’t take into account is the 5V output will change things a bit.

I figure a 10uf cap in the place of C2 will accomplish the cap needed for the LDO along with the C2. I have no facts to back that up though.

I obviously have plenty of room to add a resistor to the C2 cap if needed, I can add the pads easy enough and test later of course.

Any other suggestions? I just tossed this together real quick cause I was bored and didn’t want to spend a lot of time on it.

TA, keep the small resistor in front of the first capacitor. I would guess a 10 ohm or so would be ideal. You may even go higher since you have voltage to spare with a multi-S setup. The active regulator does change things, but I expect the ringing to be still there without this R.

You are OK with only the Cout of the LDO, depending on how close to the MCU it ends up. If you do have the space, a 0.1 to 1 uF C2 would be a nice to have.

You may need to bump up Cin to at least 2.2 uF, to prevent the LDO from seeing the FET turn-off spikes. This is where a resistor larger than 10 ohms can help.

I’m not recommending you put a resistor in front of C2. That will depend on your LDO. For the LM2936, it is critical. What LDO do you plan on using?

Whoah!!! I just saw this build and link on OSHPark…

HOLY SMOKES that’s awesome!!!

17 (covered), 20, 22mm variants are “the norm”
Man…so awesome!

It actually hit me this morning a FAR simpler and more efficient way of doing a 2s or 4s setup in a SRK. Kinda hard to explain but I will knock out a prototype setup and see what people think.

Thanks! been wanting to do this for a long time but just never had the time.

If TomE decides to make a Narsil for this type of driver that would truly complete my goal for these save for adding a 4th channel to the reset pin if that is even possible.

Yep, got all those sizes take care of along with a few variations of them as well. Also a 26mm that can be sanded down to ~23mm or so if needed.

I might measure the S70 and L6 drivers along with the DO1 and knock out some in those sizes as well just in case someone wants to swap those out as well, but those would need new firmware to be able to use the e-switch.

This leaves me with the second “issue” for a 4S SRK driver. I would really prefer NOT use an FET with a multi-cell setup. You loose so much efficiency doing that I just find it a waste, plus it is real hard to control current to what you want.

What I really want is a buck driver. I have done exactly zero research into buck drivers for flashlight use.

I know the RMM has a nice buck driver with good specs, a single coil would not hack it but using them either in parallel or multiples of them with each feeding a single LED would work.

Anyone have any good suggestions on components for a 2S & 4S (would prefer to use components that can do either with resistor swaps ect) driver?

Maybe pull the power for the MCU at 4.2v before the series connection?

Yeah, controlling current is a down side if you want less than the cells can deliver to the emitters, but I would think a FET driver would be pretty efficient. Well, other than driving emitters at a high level that is inefficient (due to the emitter’s design) when the cells are fully charged. Maybe you’re thinking about the inefficiency of the 7135 linear regulators when the cells are full?

A typical buck regulator like we use is about 80% efficient when driven hard. Almost all our buck drivers use the same Chinese buck converter (forget name at the moment). These buck converters are fine driving with 2S, 3S, and 4S cells. They will work with 1S cells, but not deliver much current due to lack of voltage overhead. Same with driving 6v of emitters with 2S cells or a 12v of emitters with 4S. Probably will not get much if any regulated run time.

A buck will work nicely for driving a 6v of emitters with 4S or 3v of emitters with 2S or 4S. Then you enough voltage overhead for decent regulated run time.

The HX-1175b driver is cheap enough to be a good donner for components. It has terrible parasitic drain though.

Ding Ding Ding. We have a winner!

This seems like the easiest method for a side by side cell setup by far. No worries about any of the LDO’s, zeners ect.

Here is a quick proto-type I knocked out to visualize it.

The ring on the outside is connected to the 1st cell to give you 4.2V to anything connected to that ground plane.

The FET / 7135 / Buck driver grounds to the 4S negative terminal. Thus giving the LED 16.8V.


More I think about, it seems quite easy to update Narsil to support this triple output. Between what I know, and code, in bistro_tripledown to enable the 3 PWM channels, and TK's level_calc utility for generating the ramping table, that seems to support 3 outputs, I should be able to come up with a pretty good smooth ramping, then the fixed modes look pretty easy to update.

For ramping, it will be different tables, depending on 350 or 380 7135's. I prefer a 350 for the single because of it's lower moon mode, but for the 6 7135 bank, it doesn't matter, but will need to know (I think) what they are when the table is generated. You can also easily tell it how big you want the ramping table assuming a 16 msec update rate. So if you want a full ramp to take 1.5 secs, or 2.5 secs, you can control that via how many entries to spec when you run level_calc, then it's easy to confogure Narsil to match it. I experimented a lot with different table generations til I got, what I felt, looked real good -- this would take a lot of time again, I'm sure, but might go quicker for me now, knowing it a bit better.

I think it’s worth your time, Tom. As long as you don’t mind giving up your indicator LED.

Having a Medium/High/whatever you call it mode that is a consistent pwm-less level no matter the state of charge is really nice. It always annoyed me that my medium modes (the ones I use most) would “droop” so much as my battery wears down when using the FET.

Current control is very hard for sure, particularly at 4s voltages where you have enough overhead voltage to drive a single emitter at almost 200 watts if you let it flow freely!

That can be “controlled” though with PWM, this leads to the actual issue I was talking about when I said efficiency. The FET is indeed quite efficient, no worries there.

My issue is the LED efficiency when using PWM to adjust the current. You loose a LOT of output doing this along with a LOT more heat. For example Dale was able to get 2200 lumens from an XHP35 using what I assume was an FET driver and he said that was all it had.

Acebeam on the other hand are able to get 2600 lumens in the K70 in what I am sure is a setup that leaves a bit of wiggle room for more. I am positive they are using a Buck driver and this explains the extra output. They are also not using the latest XHP35 that cutter is selling I am sure as well and these numbers fall better in line with what I would expect from a properly driven XHP.

I know I am preaching to the choir about this but for those that do not understand, basically with a 4s setup an FET starts dropping the LED’s efficiency so you don’t get as many lumens per watt and it runs hotter as well. The change can be quite drastic as well.

I am not worried about regulated output, I just want the best output for a 6V or 12V LED possible.

Now, far as components, does anyone know the parts list for : http://www.mtnelectronics.com/index.php?route=product/product&path=67_115&product_id=525

Or is that a closed source driver? Those parts/specs x4 on a Q8 would make for a killer driver IMO. Might need to rework the MCPCB to allow connecting each LED individually but that might even be able to be added to the design before it is finalized.

Are there any buck converters that would allow for ~4A of driver current at 4s voltages and work work in a setup like this?

PD68 - Yea, that's definitely true. For single LED lights, the 2.1A-22.A is a nice #, where you figure 5A or so is 100% FET, so top medium mode of 42% is excellent, and like you say, it's a stable PWM-less medium mode. From what I see in the output level table in bistro-tripledown, almost all the modes are on the single 7135 or bank of 7135's, so for ramping, you get the same benefit, accept you don't know where the 100% of 7135 bank is, unfortunately. But still w/ramping, you get the benefit of not drooping output as the cell depletes for all levels accpet the brightest ones that use the FET.

I suppose though the ramping, if tweaked for cells at full charge, and as the cells deplete, the top levels won't change much - but still not a totally bad thing.

Scaling it for multiple parallel LED's is another thing - if max amps is 16A on the FET for example, then 2.1A is only 12% or so.

FOr Mtn MAX Buck, it's totally open - you can buy the parts and PCB separately.

That would be great if you got that done. Is it possible to get the 4th PWM channel on the reset pin? And/or move the voltage divider to the reset pin?

If so then the freed output could be used to control an indicator LED, which would be quite cool itself. Or it could control another channel of 7135’s. I am thinking 2x single 7135’s would give you moon, low and med modes, then all of them together could give you a nice high mode.

Course really you could make one of the single 7135’s non-PWM if needed as well, the key is if the reset pin can be used for anything. I doubt removing the voltage divider is possible with more then 1s setups.

Most of us can’t reprogram if we use the reset pin though.

Here is the link for parts on the MTN-MAX, though I noticed he doesn’t list values. Might have to PM Richard

Ahh, supposedly you can use the reset pin if you got the full development kit for the Atmel parts. Not a lot of money, can be bought at Mouser, think'n $50 or $60 or so? Now, I'm not sure what the limiting factor is - can you program it just once witha the USB dongle, or can't program it al all? Once the reset pin is used for I/O though, you can't program with our standard cheap dongles. Richard of course has the full dev kit, not sure if anyone else around has it - I've been debating buying it for a long time.

It's this: http://www.atmel.com/tools/avrdragon.aspx, purchase here: http://www.mouser.com/ProductDetail/Atmel/ATAVRDRAGON/?qs=%2fha2pyFaduhmHOAilKIdN%252bhVgiKoKD8waM5I7zDMzC0%3d

I ditto PD’s statements, that is the main reason I am building these drivers.

Far as the Q8, whats to say you are limited to 7-9x 7135’s in a Q8? I started playing around with a SRK driver based on this same setup but stopped since there was no firmware for it and I could not improve on the Q8 design by PD if not adding the 7135’s.

If the 7135’s were added in my early testing you could fit upwards of 16-20 without much hassle and still not be crowding the board, if you really wanted to maximize space I bet you could fit 24-26 but things would get cramped.

So that is up to around 7-8A of regulated current you could use in a Q8, which I honestly think is more then you need for a med mode but you also don’t have to install all the 7135’s from the factory. From my own testing I think around 5A is a good regulated number for a SRK light, the light itself can handle that amount of heat long term without getting too hot and it is still ~1800+ lumens (have not actually tested it at this level since getting my sphere working but good guess, might be more).

Plus turbo would seem way brighter if jumping from 5A to 16A. Can always add another mode between as well, say around 8-9A.

I will look into the MTN MAX buck parts and see what I find.