Hotwire LDO driver! 13mv drop with 1185

[andrewwynn]

been wondering about the startup resistance of bulbs.. 12V/.25ohm = 48A.. i think i really like the idea of max current.. might not be necessary since i'm going to have a ramped startup anyhow.. maybe like 40-50 mSec startup.. possibly as much as 100-200 msec.. i picked an FET that can handle full power for short term specifically so i can hold back any amount of current or voltage in a 'soft start' incan solution.. as well as constant loads of up to 7-9W in the case of battery/lamp voltage mismatches of up to several volts (i.e. running a bulb at 12.5V from 4xLIon.. where startup Vbat is as high as 16.8V!)..

The FET i picked is the IRL3716.. 4 mohm resistance! 110 to 200W max power dissipation.. and i plan to heatsink directly to the body of the light so not too worried about the likes of 3.35A x 12.7V (absolute dead short on the load.. full power dissipation on the FET)... unless i made some major oversight the FET looks dead on perfect for this job.. 13mV dropout at 3.35A (1185).. meaning i could see as much as 10.787V get to the bulb! (if i blow off current regulation).

I have the parts all ordered so i can breadboard the driver next week.. should have an operational light or two by next weekend. I will do some 1msec peak readings of voltage and current to see exactly what happens with the 1185 and other bulbs. My goal of course it to get constant output of the max i can push the 1160 in a 7.2V host.. i believe i can push the bulb to at least 800+lumen (unless i got some bum data from my hotwire guru).

The other 'gems' that can be pushed if current and voltage are limited at startup... the 1166.. i actually got that bulb to operate at over 14V, but only by pre-heating it at 10.8... it blew the first time i tried to light it up cold with 14+V (which i expected.. it was a 'tester sample'.. no pain no gain.. RnD)..

I believe that bulb can be run at over 12.5V.. that is only 12% overdrive.. and that is over 900 lumen! (bulb lumen)... i have a host lined up to run it... can't wait to see if i have to dial it back.. even if i have to run the lamp at 800 lumen.. it will be 800.0 lumen from 100% to 0% of battery.. no dimming whatsoever... my least favorite thing about hotwire bulbs.. dimming.

-awr

 

[evan9162]

That's a very nice FET.

One thing to note: Look at page 4 of the datasheet at the Maximum Safe Operating Area graph. That's what you'll want to watch in terms of startup (and if you want to use current limiting). It'd probably be a good thing to graph the startup current of your bulb if possible, to see how much current it pulls and how long it takes to settle down to steady state levels, just to make sure you won't be damaging the FET in any way.

 

[andrewwynn]

thanks.. yeah i agree i think i found a kick butt FET.. i will double check the page you say 'cause i do want to keep within the limits.. i don't think the startup will be more than about 50-100ms even with the 4.7µF cap i plan to use vs 1µF for the startup.. i should have all the parts by monday to breadboard one of these drivers.. i will post all the exact details for people that want to copycat, but hope to offer a complete drop-in sometime very soon.

-awr

 

[NikolaTesla]

The head of the KIU kit will be a good heatsink. The ULTIMATE Mag85 will materialize soon. 12 Cell Modamag to 3D coming on line soon. Plenty of spare voltage to regulate down. Add 1 FM 3D bored tube, Install KIU/AWR regulated Mag switch. Add 1 Super reflector. 1 dozen CPB 1650's to fuel it. 1 new WA1185. 1 coated glass lens. As soon as Andrew whips up the regulator, between both of us, (I have the rest of the parts) we should have a prototype for show & tell. I can't wait :naughty:

 

[NewBie]

If you need a source for the IRL3716, Arrow has 507 in stock at 1.74 ea at the higher online pricing.

Newark has them for 1.55 ea, 756 in stock.


MIC5158BM is 3.03 ea, 190 in stock at Arrow.
MIC5158BM is 2.6105 ea at Future

Looks as though when you choose the right combo there, you are looking at 4.17 total, and the rest of the parts are insignificant in price, easily 5 dollars. Adding the 1 dollar for the board, and 1 dollar for assembly, would put you nicely within your 7 dollar target.


When setting everything up such that the dead battery voltage is higher or equal to the bulb voltage, this means the regulator will be burning up power (Vbatt-Vbulb)*Current= Lost Watts turned into heat, the whole time, with the amount of heat it is making being alot higher when the cells are fresh.

All the LDO is doing is buring up the difference between the Vbatt and Vbulb.

Any idea on the losses in the beginning, middle, end?

How do you figure you'll pull the heat out, pot it to the flashlight body?

The soft start is a really big deal, most especially when overdriving, as I found a year ago, on the Thor's Hammer regulator that I've been fiddling with when I get a chance. Saves a ton of bulbs. You won't need the current limit portion of the regulator if you get the soft start capacitor selected such that the soft start time is longer than the filament warm-up time. By implementing the soft-start I mentioned on Thor's Hammer, I've gone from flashing 2/3rds of the bulbs at an extreme 14% overdrive, to zero, even after using the same bulb for a long time. Even to the point of passing the Tc hours point on several bulbs, and well past the B3 point.

It would be interesting to see just how far a person can push a bulb, when using soft start and with regulation.

I know there are a few folks that have waned on eloquently about the soft start making no difference at all. Sorry, but bench tests say otherwise. If a fella was running a bulb within spec, and in the specific bulb design, the filament wasn't pushed very hard at all (kinda like underdriven- cooler filament temps like you find in auto bulbs or signal flashers), as is with certain bulbs, in order to get a very long life out of them, then there is a good chance a person might not see a difference. But in high performance bulbs, where the filament is driven considerably harder, ran at a higher Kelvin temp for increased efficiency, I see a considerably improved lifetime with soft-start, even when ran in spec.

Insta-flashing is a real pain on alot of these high performance lights, some folks even do a partial discharge to help limit the surge current. And some of these bulbs are just a tad bit more expensive than the 10 cent bulbs in cheap lights...

The availability of various regulators with soft-start circuits is a real nice thing and it will be great to see them in more and more lights, quickly saving folks considerable money that can be used to purchase more lights. Increasing the reliability of the hotwire is a wonderful thing, but choosing the right parts for the job is important for relability in the electronics, and it appears you did a great job there.

The drawback, for the bulb sellers, significantly reduced revenue from decreased insta-flashed bulbs.

 

[andrewwynn]

oh i love logging in and seeing detailed replies!

I ordered the FETs from digikey.. i get things 'next day' from them for like $3 shipping.. i live like a 6 hr drive from them so UPS and USPS are 'next day'.

I think i ordered the driver chips from arrow, they are enroute!

The only other part with any price at all is if i put in a nice big output cap it could cost 30-50¢.. and a nice reliable potentiometer i usually pay 40¢. (anybody wondering how to type a ¢ on a mac you type option-4).

I have a spreadsheet that shows me the power dissipation on the FET with totally fresh batteries.. on some configurations it gets as high as 7+W but just for a minute or two and quickly down to 3-4W. (as much as 10% or so).

pot to flashlight body=yes.. actually i'm hoping to get away with this concept.. arctic alumina from the FET to the body and using a spring of some sort (possibly just some foam rubber).. to push the FET from the PR post against the body of the light.

I need regulation for some of the bulbs ... i want to put an 1160 (5V bulb) into a 7.2V solution.. which of course is 8.4V no-load fresh batteries! I also want to put the 11.6V 1166 into a 14.4V solution.. i got it to work once with bats that aren't fully charged.. I agree that soft start will be enough for any of the combos we use that work already (like 1331 in 10.8v solution).. which is likely to blow with fresh cells but if it took 1/3sec to turn on would probably be fine.

I will be finding out 'just how far' i can push a bulb.. those two bulb options have the possibility of being over 900 lumen if i can push them as hard as i hope to.

empirical evidence in the 'real world' shows absolutely that soft start will help from blowing bulbs.. how often in your life have you had a bulb blow just while on vs when you first turn it on.. stress breaks things, period... less stress less breaking, does not take rocket science.

when i bought my hotwire light (KIU socket mag3d).. it came with two 1331 bulbs.. one never worked the other i blew within 1 week... the add'l 'afterburner' from the fact that LiON or NiMH batteries have 16% extra voltage from 'rated' when fresh off the charger makes matching bulbs pretty tricky.

It took me like three allnight sessions to find the match of parts, but major kudos to JohnMuchow for finding a different make of the controller (the one i had found and already had in my order did not output enough voltage to run the higher voltage ckts.. was virtually identical in operation for the lower voltage solutions but had a voltage 'doubler' vs. 'tripler' and wouldn't boost the voltage high enough to run the 14V solutions...

I had found a number of 'decent' FETs that were like 20mohm and 17mohm but when i found the one with the combo of FOUR and with 100-200W instantaneous power capability...

I found a different solution that i also bought the parts for.. and would be easier for the avg. joe to wire up.. only has the FET and a single 8-pin chip that i think is available in 'DIP' format.. so easy to solder it's crazy.. I will make a schematic for that design and post it as well because it would be easier to 'copy cat' me.. i think it would be humorous if somebody copy-catted me before i even get the chips and build one i think that'd be funny.

I have a PR 1166 and plan to build a ckt to run it at like 12.5V in a 14.4V solution.. but not one that would be recommended.. it's an envelope-pusher... (4) 14500s in a FiveMega 1D light... and the other model i'm workin on is the 1160 in a 7.2V (6x17500 2D)... both of those should be about 900 regulated lumen. I should have the parts by tuesday, so hope to have those built and running before wed when my life becomes 'nanoworld'.

Some of my motivation is to have more reliability, but to be able to push bulbs harder to make them more efficient and brighter.. but i can't stand 'dimming' so probably that's the initial motivation.. to make a hotwire light that outputs say 700, 800 or 900 lumen and every time you turn it on.. that's what comes out.. just like what i'm used to with LED lights.. constant output start to finish!

-awr

 

[evan9162]

Andrew,

I'll bet you're going to need something better for thermal transfer from the FET to the light body (especially for 7W). You're trying to mate a flat surface (tab of a TO-220 package) to a round surface (inner wall of Mag body tube) - You won't get much thermal transfer, even potting with AA or something similar. I'd suggest a rounded chunk of aluminium or copper if possible.

Another thing to try would be to use some 1/2" copper pipe, cut off a 1/2" long piece, drill a hole, and use a screw/nut to compress the pipe between the transistor and body tube of the mag - the pipe will flatten out and should conform to the shape of both the transistor tab, and mag body tube.

 

[andrewwynn]

i should only need to transfer about 4-5W, but i agree on the issue of mating the surfaces.. i might cut a piece of copper flat on one side and curved on the other side to mate the FET nice to the curve on the inside of the body. the transistor i have picked out is not the T0-220 with the screw hole i need to use AA glue to hold the FET to my heat conductor.. which might just be a strip of copper (like 12 ga wire pounded flat). I won't know what i come up with 'til i make the prototype... i'll just have a small heatsink attached on the breadboard and get a sense of how much heat i have to deal with. Thanks for the idea though... i'll def. post what i end up using, it'd be very cool if it ends up being 'collaborative'.

-awr

 

[StoneDog]

Is there a safe "upper limit" with respect to the battery voltage? Or perhaps it might be better asked, will this ckt allow one to increase Vbat for longer runtimes (sort of like a DownBoy buck regulator with a Lux3)?

Jon

 

[NewBie]

i should only need to transfer about 4-5W, but i agree on the issue of mating the surfaces.. i might cut a piece of copper flat on one side and curved on the other side to mate the FET nice to the curve on the inside of the body. the transistor i have picked out is not the T0-220 with the screw hole i need to use AA glue to hold the FET to my heat conductor.. which might just be a strip of copper (like 12 ga wire pounded flat). I won't know what i come up with 'til i make the prototype... i'll just have a small heatsink attached on the breadboard and get a sense of how much heat i have to deal with. Thanks for the idea though... i'll def. post what i end up using, it'd be very cool if it ends up being 'collaborative'.

-awr

For 3 dollars, or 1 dollar at production levels (too high for you), you could do a spiral heatpipe. Not that you need it, but nice way to spread all that heat out over a flashlight body, and it would do it quickly. 4-5W is alot in a contained space. Copper/Al works also, just another way of snazzing it up.

One place:
http://www.acktechnology.com/Heat Pipe.htm

 

[andrewwynn]

stone: the biggest differential i have in the spreadsheet so far is running an 1185 at 11.0V from a 14.4V nominal battery pack.. that puts 3.35A (probably a little more).. through with a 3.33 to 5.73 voltage drop on the FET. This puts between 11W nominal and 20W peak power loss on the FET, but with enough heat sinking it should work it's well within the rated specs.. the efficiency electrically speaking drops like a rock.. that works out to 76% nominal efficiency and 66% to start with... however that is an estimated output of 1300 lumen... an 1154 bulb running at 14.0V will output about exactly the same amount of light, but in that same circuit will be 97.2% efficient nom. and 83% starting... 'electrically speaking'.. here's where it gets werid..

at those drive levels it's about 47W to run the 1154 and about 38W to run the 1185.. but with the efficienty loss on the converter it actually takes 49W out of the battery to run the 1185 in that solution and actually 1W less to run the 1154 in that solution.

The bulb efficacy is much higher in the 1185 (35 lum/w) compared to the 1154 (28 lum/w).. but it's bizarre how similar that averages out to a very simliar output in lumen/battery-watt.

To summarize your question to a point-blank answer..

The primary motivation for designing the circuit was to allow the use of higher voltage batteries matched to lower voltage bulbs.. this gives two specific results i was aiming for:

1) brighter output (naturally).
2) longer regulated runtime.

The bulb-battery combinations best suited to this so far:

5V 1160 bulb run from 7.2V nom. battery.. about 90% efficiency and 100% regulation for full battery life.

11.6V 1166 bulb run from 14.4 nom battery... about 86-90% efficiency and also 100% regulation for full battery life..

It would be possible in certain circumstances to increase the runtime by adding a cell.. for example we are planning to use the regulator to run the 100W osram bulb in regulation from 12 cells vs the 11 max it can handle with direct-drive... but we don't really expect much increase in runtime, just that the output will be constant... however, especially with lower current bulbs, it would be quite possible, for example to run the likes of a 10W bulb that is 12V from a 14V source for quite a bit longer if not over-driven to the max, so there is not much current so not much power loss on the FET even though there is a high voltage differential.

This circuit is a linear circuit, no switching.. so the efficiency is directly disproportional to the voltage difference of the bulb and the battery.. doesn't mean there aren't swell trade-off options to get longer runtimes though.

I did not optimize my search for best runtime but for most output, so i'm sure there are some neat lower wattage bulbs that would work great along the lines you describe.. the 1316 comes to mind for 7.2V operation.. 10W bulb, 236 rateed but 6V bulb.. running at 7.0V you will have regulation for nearly 100% of battery life and get about 80 minutes runtime from a six-pack of NiMH high mAH models.. that's double the runtime of 1111 in the same host.

it is not practical to 'double' the cells to double the runtime since it's a linear device, but adding a cell or two in a NiMH solution is possible.. but it won't greatly extend the runtime it will mostly just flatten the output (make it more constant).

========

newbie.. that heat pipe is neat stuff, thanks for pointing it out.. 4-5W is a lot in a contained space, but not too much to have a 'relatively sloppy' heat junction to the case of the light.. the FET can handle heat temperatures waaay too hot to touch, so if there was a thermal gradient from say 70C to 45C across the thermal junction to the case that might actually be nice.. won't have as hot of a spot on the body, just runs the FET hotter. The important things are that 1) FET not hotter than max and 2) SOME thermal junction to the real world. The IC will be seprate from the heat of the FET since it is affected up to 2% by temperature.. it doesn't matter than the FET is affected by temperature becuase the controller will modify drive level to counteract that.

I should have the parts tonight or tomorrow.. getting excited to build up one of these drivers! I'll most likely start with the 1160 in a 7.2V host i will try to 'hold back' and keep the voltage set to a bulb saving level at first just to make sure everything is working nice.

Oh yeah.. i'll probably also make one for the 14.4V host to run the 1166.. that one should be on the order of 'nutso'.

-awr

 

[StoneDog]

stone: the biggest differential i have in the spreadsheet so far is running an 1185 at 11.0V from a 14.4V nominal battery pack.. that puts 3.35A (probably a little more).. through with a 3.33 to 5.73 voltage drop on the FET. ....

...The primary motivation for designing the circuit was to allow the use of higher voltage batteries matched to lower voltage bulbs.. this gives two specific results i was aiming for:

1) brighter output (naturally).
2) longer regulated runtime. ...

-awr

Thank you for taking the time to give such a complete response. I can't say I understand everything completely, but that is because I know very little about hotwire mods right now. I will study this thread a lot more and try to get a better handle.

I'd really like a nice, bright regualted incan. Something with the SF A2's amazing regulation that everyone raves about but with more light and longer runtime if that's possible. I'm not too proud to go with a lower current bulb if it still beats out something like a Streamlight TL-3 AND will last at least an hour on decent batteries. If that means a Mag 2D bored to accept 8 NiMH, or maybe just a Mag 2C with 2x18650 then so be it.

Thank you again for your response and your research into this ckt. I will definitely be watching this thread and consider me in line for the final product.

Jon

 

[andrewwynn]

I have a huge spreadsheet i've been working on for about a month.. it includes over a dozen bulbs and about 8-9 different hosts and battery solutions.. since some hosts can handle multple battery solutions.. probably like 20 different combinations of hosts.

based on the master spreadsheet i made a smaller one that is just for the batteries.. i posted over here a very pretty chart showing all kinds of battery solutions, and their watt-hour ratings.

I have lights in-the works that will blow away the A2 (well anything in the size-class of the ones i'm working on basically).. I have one light in-the-works that is 1600 lumen from a 2 1/2D (can do it in a 2D but it's pushing the batteries a bit past their recommendation)

The 'favored' hosts will likely be 2D 800-900 lumen regulated output.. 40 min runtime.. same light in 2 1/2D host is 75min runtime..

I Really love the 2D-3 host.. there are a lot of possibilities.. 6AA.. 8AA..R123x9 17500x6... with losing some output capacity you can put 4x17600s in series for 14.4V.. amazing possibilities.

-awr

 

[StoneDog]

Hmmm... the 1160 is rated at 618 lumens at 6.0v. That may not be much in the world of hotwire modding, but it's definitely up there and I would bet it beats an SF M4 HOLA and comes close to an M6 HOLA.

Given that specific application, the 1160 in a Mag 2C driven by 2x18650 would make for a bright little light. Runtime wouldn't be spectacular but might be close to the A2. And if the ckt does indeed cut out when Vbat drops below the desired voltage, then it would be safe to use unprotected cells.

I would be surprised if you and the other hotwire modders would be satisfied with this configuration, but to a newbie it sounds like a great DIY starting point.

Jon

 

[andrewwynn]

the 1160 is an awesome bulb to put in a magcharger.. you'll get more like 500 to 550 lumen once you factor in switch resistances.. but it's a real winner and a total 'gimme'

2C 2x 18650 would need regulation to get the voltage down or you'll melt the bulb.. but it's a total winner.. 1.6C and like 37 minutes runtime.. the bulb can be turned up to 'ludicrous speed' or even up to 'plaid'.. i would bet 750-850+ lumens is not out of reach with that 2C 2x18650 solution..

the normal modders haven't done that combo because it requires the regulator.. in-general DD is the only thing people do in hotwire.. very rare to find a regulated ckt, what i'm trying to change.

-awr

 

[NewBie]

Andrew:

Take note of that MOSFET's datasheet, where the on resistance is specified at an internal die junction temperature of 25 degrees C.

It doesn't take much to raise the on resistance by a factor of 50%, resulting in adddional losses, especially if you have a high thermal resistance from junction to ambient. See page 4.

 

[JohnMuchow]

Andrew,
I'm glad the Micrel chip looks like it's going to work!

I haven't given this a lot of thought but do you even need to worry about the inrush current? Since the regulator will see the big voltage jump across the sense resistor, won't it instantly bring down the FET's gate voltage level to keep the sense resistor voltage drop at 35mV or less? Just size the sense resistor to ensure a max current level that is OK...especially for prolonging bulb life by minimizing turn-on shock.

You've picked a darn good FET there (my favorites are the IRL1404 and IRF1405). Since the FETs are being used as voltage-dropping pass elements, they probably won't ever get even close to being turned on fully. I think that the Rds-on value won't matter too much since they'll be operated in their linear region and the Rds-on will be much higher. You might be able to save some $$ by going to a higher Rds-on FET with equal power-handling specs. Try www.arrow.com, they have some great prices on quantities over 50pcs.

More stuff I haven't thought about too much...
I'd try to find the FET with the lowest junction-case thermal resistance to minimize the need for a heat sink. The IRF1405 has a j-c resistance of 0.45 so it is a lot more efficient at moving the heat away from the junction and lowering its temperature for a give heat sink size. I'm not sure if it's worth too much of your time at the lower power levels we're talking about here, but it sure made a big difference for an application of mine where each FET was dissipating 45W continuous!

Good luck with your project!

 

[andrewwynn]

actually it does take 'much'... to rase the resistance 50%.. up to a whole 0.006ohm! is 140C.. or 286F! that's close to melting solder (tangent: i did actually heat up a soldered joint hot enough to melt itslef in an early test of one of my hotdriver circuits!).. i couldn't figure out how the heck this wire kept falling off my transistor but apparently it was heating up hot enough to desolder itself!)

I have the FETs now... i also have the p-ch versions... they are 'rice-size' no wonder they can only handle 2.5W.. but the N-ch 1302 has a case-size that is the size of an T0-220 but w/o the tab and the feet are pre-bent to fit as if surface mounting.... the 'bottom side' is metal.. i will likely be using arctic silver to attach the bottom to some copper that can be made the right shape to fit the inside of a mag body, so i can use very thin layer of AA to conduct heat to the body.

-awr

 

[StoneDog]

the 1160 is an awesome bulb to put in a magcharger.. you'll get more like 500 to 550 lumen once you factor in switch resistances.. but it's a real winner and a total 'gimme'

2C 2x 18650 would need regulation to get the voltage down or you'll melt the bulb.. but it's a total winner.. 1.6C and like 37 minutes runtime.. the bulb can be turned up to 'ludicrous speed' or even up to 'plaid'.. i would bet 750-850+ lumens is not out of reach with that 2C 2x18650 solution..

the normal modders haven't done that combo because it requires the regulator.. in-general DD is the only thing people do in hotwire.. very rare to find a regulated ckt, what i'm trying to change.

-awr

Well, if/when this ckt is available I'm definitely going to give it a shot. 'Might even spring for a Mag 2C in HA.

Jon

 

[evan9162]

Looks like you got the '1302 in a TO-263 (D2PAK) surface mount package, though the 1302 datasheet shows a TO-220 package. Were you intending to use this package, or would you rather have the TO-220?