Hotwire LDO driver! 13mv drop with 1185

[NewBie]

Another method for doing this that has proven itself with the cavers is to go PWM.

The well known and infamous LVR regulators for bulbs are designed by Willie Hunt, and you can contact him at [email protected]. It has multiple selectable brightness levels, and has been around for many years, well known, established, and reliable. The multiple levels are extremely handy, especially when you don't need peak lumens and would like additional runtime.
http://www.cs.indiana.edu/~willie/lvr.html

His regulators can handle up to 25 Amps and are 99% efficient.

It varies the PWM duty cycle as the battery voltage drops to hold the bulb brightness constant.

Bulbs are much more suitable than LEDs for PWM, as they respond much slower, responding more due to the heating cooling time constant of the filament, and an average state is reached. LEDs respond extremely rapidly, and an average state isn't reached, even pulsing at 1,000,000 times a second.
(don't get me wrong, there is nothing wrong with PWMing LEDs, when the maximum current is limited).

Though, I didn't see where he mentioned inrush current limiting, there are several ways of doing it, you could run linear up to the point where the bulb filament heats up, then switch to PWM, or run the PWM very fast in the beginning, then switch to a lower frequency once the filament was heated up.

Or you could go all the way and add the cap and inductor, ending up with a switching supply, which you can fully soft-start and all, and if you choose the right switcher chip, you can go to 100% on.

 

[andrewwynn]

i think i typed my reply before JMs post but didn't hit 'send' 'til after..

I will probably not need or use the current limiting ckt.... i am putting in a startup cap that is 4.7x the size of the design, it will probably take about 2-4msec to turn on, the 'turn on' is linear in volts.. i will do some measuring of peak current and peak voltage when i'm testing.. my fluke will measure down to 1msec peaks.

in the ckt with the optional R4 (or the magswitch hack USING the switch of the maglite AS the R4).. it will limit the current to the max of 3.33 to 3.5A. I really would like to have that built in to the circuit.. i do have some high-quality 0.1 ohm sense resistors in-fact that i could use in the R4 case.

In the circuits where the Vbat is close to Vbulb (1185 solution).. the FET will get to the point of being turned on completely, and the 14.4V solution with the 1154.. likewise.. however in the 1160 / 7.2V and the 1166/ 14.4V solution the FET will never be fully turned on.. the FET i picked costs about $1.50 which i don't know that i need to find a cheaper one and this way i only need one for any voltage.. the same exact ckt will work from like 1.5V to 20V. (well the low end might not go quite that low).

I am going to use the FET with high current through it and high voltage across it.. it won't matter what FET it needs a major heat sink.

In the current design it will be built into the KIU switch which has a solid chunk of aluminum about 26mm diameter and 9-10mm thick... it will be in direct contact with the body and will have AA thermal compound keeping it connected.. probably good for 100W continuous (of course you'd not be happy to hold the 150C flashlight), but the FET would be happy as a clam.

The FET in a couple of my design possibilities was 9W continuous with fresh batteries and one was about 20W... i do want to have serious power dissipation capability, but 10W is a plenty and i think that what i came up with will only heat up about 15-20C with 10W continuous.

evan:.. yeah i got the D2pack on purpose.. the 1302s is the D2PAK datasheet.. I wanted something smaller and don't mind having to use arctic alumina epoxy to attach the thing to the heatsink... once i get in under the hood i might want to re-do that so i can use the TO-220 and just screw or clip the thing on but time will tell.

newbie.:

I actually found that exact website when doing my google for LVR. I emailed the email you suggested (and listed) and it bounced.

The beauty of a switching regulator is of course that there is virtually no power loss on the pass transistor! I was hoping that i would find an IC like the gem that john found that was a PWM switcher but in my case i'm trying to drop so little voltage in most cases that though the efficiency might be as low as 95% early on, they get to 98-99% in most cases where the Vbulb is close to the nominal Vbat.

I didn't notice if or what he did about the startup current in his regulators.. he uses µC so it can be anything... my bet would be a linear rise in Vgate 'til voltage is achieved.. he does have a really neat feedback circuit that shuts the thing off automatically in the case of no bulb.. quite a nice regulator.. but you can look at the schematics.. quite a bit more complicated than what i'm working on.. i want 'dirt simple'.

Bummer.. no ICs in the mail today! i have this IC and another neat little IC that is a simple op amp, comparator, and Vref in a single 8-pin chip.. i can build a very simple LDO from that, it's like 6 parts vs 10, not to mention 8 vs 14 pins.. i think for the 'copycatter' that might be a lot nicer and i'll have the circuit online in this thread soon.

that design is 'ludicrously simple'.. but still uses a 'fancy' part so i have another design i'm working on to make a ludicrously simple ldo just with an FET, a zener diode and a couple pots.

you can see the thread in batteries included (typically right next to this one)

-awr

 

[NewBie]

Maxim makes a part similar to what you were talking about, but only good to 7V:
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1221
High-speed for 5.5V:
http://pdfserv.maxim-ic.com/en/ds/MAX9000-MAX9005.pdf

Microchip also makes one that works to 5.5V:
http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=1335&dDocName=en010431

Linear Tech makes one thats good for 12.6V:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1154,C1004,D4396

Another very interesting part for 32V with ref, current source, and two op-amps:
http://www.st.com/stonline/books/pdf/docs/4950.pdf:


Of course, if you don't need the internal reference, since there are many low cost small external references, you can also always use the very low cost LM392 op-amp/comparator combo.

 

[andrewwynn]

hey i really like some of those.. they are just like the one i found but some have higher voltage maxes.. the one i found has a max too low to work with my 14.4V solution... i need to put a Vdropping diode in series to make it work..

I finally took a closer look at how the KIU socket works.. and it looks like i can fit the circuit inside between the KIU base and the plastic magswitch case... the heatsinking to the case will be fantastic. I should be able to put the pot in in such a way i can access it from the top to be able to adjust the output.

STill didn't get the controller ICs yet..

Anybody know of a PWM version of the ICs in this thread? It would allow a wider range of control.. the problem is the circuits get far more complicated when you have to add in the inductor that most switchers use.. it should be possible to have strictly FET/PWM to operate incandescent lights just like the ones mentioned above that willie makes.

-awr

 

[NewBie]

Could you make it work with one of these, skipping the inductor and integrating or averaging the output waveform before feeding back in?


http://www.st.com/stonline/books/pdf/docs/7335.pdf

I've seen a number of circuits that are used to control peltiers/heaters to a temperature, via PWM on the web, which is essentially very similar to what you want to do, but do it vs. average voltage on the output.

 

[andrewwynn]

i think i saw that exact device .. you would think that putting a cap on the bottom leg of the feedback bridge you could get an average voltage.. the concern i had was w/o the inductor the load will get full voltage pulsed.. maybe just a big enough output cap to absorbe the pulses? I want a device that will go to 100% duty cycle when the battery is low enough and has almost zero dropout, so there can't be any series resistance (inductor, sense resistor).

This is the kind of device i was thinking about.. and with PWM i could use the tiny pch hexfet i found that is 8mohm resistance reading on it says that it has a max duty cycle of 95%, so there would need to be some voltage overhead for this to work, but for several of my designs that would be fine.. i like it, i think i'll have to try one out.

-awr

 

[NewBie]

i think i saw that exact device .. you would think that putting a cap on the bottom leg of the feedback bridge you could get an average voltage.. the concern i had was w/o the inductor the load will get full voltage pulsed.. maybe just a big enough output cap to absorbe the pulses? I want a device that will go to 100% duty cycle when the battery is low enough and has almost zero dropout, so there can't be any series resistance (inductor, sense resistor).

This is the kind of device i was thinking about.. and with PWM i could use the tiny pch hexfet i found that is 8mohm resistance reading on it says that it has a max duty cycle of 95%, so there would need to be some voltage overhead for this to work, but for several of my designs that would be fine.. i like it, i think i'll have to try one out.

-awr

Right, thats one of the drawbacks of PWM, be it LED or Incandescent/Halogen source.

Of course, as you know, a switcher which can hit 95-98% efficiency, with a little skill, and choosing a controller that can do 100% duty cycle for the switcher, and you get the low loss of the switcher, without the heat produced by a linear, the slow gentile ramp up of soft start, plus the power that had to be burnt up in a linear, can now go to runtime, and a tiny part of it to run the switcher.

Also, since the switcher acts as a method of transforming the power to what the bulb needs, when the batteries are fresh, the load on the batteries (especially current) is reduced, resulting in less loss in the batteries.

Food for thought.

(yeah, I know, you are trying to keep it braindead simple)

 

[JohnMuchow]

Andrew, this was posted by Bob Monsen over on the sci.electronics.basics Usenet group. Just tossing it into the mix.... :naughty:

Another way to control brightness is to use what is called a PWM circuit. Basically, you give your pass device a pulse every once in a while, and control the percentage of time it is on. Since you aren't just burning up energy with a resistor, it is often cooler and more efficient. A simple circuit for this consists of a cmos 555, an N-channel JFET, a cap, and a pot:

-----------------------o--------------o--Vcc
| |
| ||--'
.--------. | || J1=BF245B
.------+Vss Vcc+----' .->||--.
| | CN555 | | |
| | | | |
| | tr & th----------o------o
| | | R=470 |
| .---+OUT DISC+---o-\/\/\/\--------o
| | | | | ^ |
| | '--------' '----' |
| | |
| | |
| '-Gate of logic level NMOS ----- C=1uF
| -----
| |
-o------------------------------------o--Gnd

Call the current through J1 'I'.
The period is going to be the sum of the charging time
Tc = (1/3 * Vcc) * C / I

and the discharge time is
Td = R*C*ln((2Vcc - 3RI)/(Vcc - 3RI))

The duty cycle, which is what you are interested in, will obviously be Tc/(Tc+Td) since OUT is high during Tc.

The odd thing is that if you simplify this, the duty cycle doesn't depend on the size of the capacitor; it only depends on vcc, Id(J1), and R. Thus, you can pick a cap that is small enough so you don't see a flash, but not too so the pass transistor requires too much dynamic current. The duty cycle defined by this monster:
Vcc
D = ---------------------------------------------
Vcc + 3 R I ln((2 Vcc - 3 R I)/(Vcc - 3 R I))

So, when R is 0, D is 1, and when R is Vcc/3I, D is 0

That means that when R=Vcc/3I, the thing will simply stop, with output low. The reason is that the discharge pin won't be able to pull the trigger pin lower than Vcc/3. When R=0, it won't take any time to discharge the timing node (well, almost no time) so D <- 1.

The N-JFET will vary as to how much current it will source in this configuration. If it is sourcing too much, put a small resistor between the drain and the point where the gate attaches; this will lower the current. However, you need to make sure that your new resistor times the current isn't bigger than Vcc/3. If it is, the output will get stuck trying to pull the timing cap up to 2/3 Vcc, which means output will be high and your pass transistor will be on all the time. If you can't find a JFET, you can use a couple of PNP transistors as a reasonable current source, like this:

--------o---------o-------vcc
| |
| [Rxx]
| |
e |
b--------o
c |
| e
o--------b
| c
| |
[Ryy] '----- current out
|
GND

Rxx sets the current to near I = 0.615/Rxx. Ryy should provide 1/10 of the
current through Rxx, so
Ryy = 10*(Vcc - 1.3)/I
---
Regards,
Bob Monsen

 

[andrewwynn]

i think you meant to use the ubb

 tag to format those... 

it will monospace and keep the whitesapces.

the last posts we were talking about using PWM.. i can't tell from that circuit but it 'sounds' like it's just a PWM like the lioncub.. no regulation just a duty cycle percent of output.. can you 'modify' and put that into a 'code' tag so it's readable? 

-awr

 

[JohnMuchow]

Awwww, crap. I thought I had monospaced it. I'll get on it.
It's nothing special, just more stuff to add to the mix as possible options and inspiration to better designs.


[Edit] Arrgghh....CPF stripped out all of the spaces. I'll have to manually put them back in later or go get the post again. <deep sigh>

 

[andrewwynn]

i would copy/paste from the source again.. somehow the whole post is in a funny font though, only the 'code' should be.

there is art to ascii art for sure.

-awr

 

[JohnMuchow]

OK, I tried everything I could think of and it won't format properly.

...yes, I tried that.

...yup, tried that too.

If anyone is interested in the circuit, send me your e-mail address and I'll send a Notepad file with the post in it (where it's properly formatted). If you want to, check out Usenet's sci.electronics.basics group, 10/2/05, 2:10pm post. Look for Bob Monsen's response to the original post.

I give up.

 

[andrewwynn]

-----------------------o--------------o--Vcc 
                       |              | 
                       |          ||--' 
        .--------.     |          ||    J1=BF245B 
 .------+Vss   Vcc+----'       .->||--. 
 |      |  CN555 |             |      | 
 |      |        |             |      | 
 |      |     tr & th----------o------o 
 |      |        |     R=470          | 
 |  .---+OUT DISC+---o-\/\/\/\--------o 
 |  |   |        |   |    ^           | 
 |  |   '--------'   '----'           | 
 |  |                                 | 
 |  |                                 | 
 |  '-Gate of logic level NMOS      ----- C=1uF 
 |                                  ----- 
 |                                    | 
-o------------------------------------o--Gnd

you have to put

 before and after ONLY the 'code'.. the second needs / before the c in code.

This circuit is a simple blinker... you set the frequency or pulse width with the potentimeter.. not sure which, not too familar with the most famous IC ever made ;) in any event.. it would be neat to explore PWM solution but i need something with regulation not just dimming.. the chips talked about earlier show a lot of promise. Also in the 'ludicously simple' thread we have pretty much nailed down a 'dirt simple' design with like 2 transistors and a couple pots.,. could be possible to make a $2 high-tolerance Vreg for hotwire that is made from big enough components to hand-wire on a proto board so people can DIY vs buy pre-fab, i love it when that happens. 

-awr

 

[JohnMuchow]

you have to put

 before and after ONLY the 'code'.. the second needs / before the c in code.[/QUOTE]Thanks Andrew, I'm sure I tried that, but I guess being sure doesn't mean too much sometimes. :)

 

[andrewwynn]

got all the parts, what i need is more time.

-awr

 

[andrewwynn]

awesome.. finally wedged in about 90 minutes between nano production and a movie to assemble a breadboard with the hotdriver.

IT KICKS ***.

I should end the review there, i suppose how do i improve on that?

Well it's a bit too big to fit in a light just yet but it's just plain awesome and incredible.. i haven't figured out the correct value to use for the pull up resistor for the enable pin to get the shutdown ckt working yet but so far it's been incredible.. The Vout change with Vin change is negligible.. and i've operated it with over 3V of overhead w/o a hitch (FET does get a warm).. but i have a heat-sink on it so it's not a problem at all. I need to get some measurements yet to show the Vdrop etc but it's just plain incredible just as expected and can't wait to make the SOIC version to put inside a KIU socket (there is room between the magswitch and aluminum cup on the socket!).. aiming to put the adj. pot lined up with a hole so i can change the output voltage.

pictures and numbers to follow.

Ok.. we ran a few bulbs on it... a 20W12V bulb was the first tester... but here are a couple interesting models:

1166 @ 12.6V and 2.07A = 26.3W.. 965 lumen. 36.7 lum/W!
1160 @ 6.64V and 3.95A = 25.8W.. 834 lumen. 32.3 lum/W!

note: i may have fixed the flag at some point in the future but didn't update this post.. the next paragraph might be outdated if that happens:

I haven't figured out the flag loop correctly yet. connected like the ckt it just stays off.. i think i need to use an FET to get it to work.. doesn't have enough drain to turn on a PNP it seems... but the N-ch FET i haven't figured out exactly how to make it take the EN low when it goes low.. though here's my latest thought:

right now the schematic shows a pull up resistor above the EN.. the idea is that the flag goes low and should pull the EN down.. the problem is apparently the EN input resistance is so high that it's always down.

So.. new plan.. voltage bridge from hot to ground.. two 10k resistors.. the center is attached to both the flag and the EN. when flag is high (normal operation.. it measures like 3Mohm), and i didn't measure the EN resistance but it's probably also very high.. this would put just about 1/2 Vdd on the EN which will definitely be enabled.. when the Flag goes low (when Vout < 92% setting).. it pulls the center of the bridge down to zero? putting all the vdrop on the top half of the vbridge and shutting down the IC.

Well i'm sure there'll be an update soon.

-awr

 

[andrewwynn]

oh.. btw.. running the 1166 i had like 30-40mV drop because the voltage is so high that the Vgs was not high enough to get the 4mohm resistance.. but when i ran lower voltasges like the 1160 at 5.5V.. i measured 13mV drop.. pretty cool. I found a trick i can probably use in the high-voltage designs to reduce the Vdrop, i will be trying them out.

-awr

 

[andrewwynn]

OK.. some neat updates:

first.. the updated schematic:

I finally figured out how to use the shutdown flag.. i had R3 in there but apparently the chip assumes 'off' so the flag pin being grounded on 'energize' pulled down the EN pin... i went through like 6 hoops with crazy voltage bridges even designed a PNP transistor option before i realized what i needed was a simple 0.1µF cap across R3 so that on 'power on'.. when flag wants to be low... C5 holds the voltage drop on R3 low just long enough to turn on the chip.. at that point as long as the voltage is high enough to keep things running 'flag' goes high and R3 lets nearly full Vbat through to the EN pin.

Oh.. i just realized i never put C4 into my prototype I also used only R1 and no R2, but i'm changing that.. i don't like the performance changes at various voltages.. R2 is supposed to be close to 10k.

The latest test: running a 1160 bulb at 6.6V.. holy COW that's bright!

see for yourself... click here.

3551K color.. 863.3 lumen.. 19.7 estimated hours.. (well i've heard they have the life span goofed on the 1160 on the site so we'll see how long that it'll really last.. the beauty is.. i can just turn down the voltage if it's too high!).

some numbers from the latest test:
7.2Vbat
6.6Vbulb
.48V FET

6.66Vbat
6.56Vbulb
15mV on the FET.. that actually calculated to exactly 4.0mohm! how friggin cool is that! The reason that is a bit of a surprise... to get 0.004ohm out of that FET.. you need to push the gate with 10V! well that means if you have a bulb with 6.6V on it you need to supply 16.6V to the gate from your 7.2V battery.. the magic IC has a voltage tripler in it... it has a zener to protect against putting too much voltage, clamping it to 16V over the source.. and the FET i picked out can handle 20V gs.

The thing shut off at 5.2V.. but that works better now.. right about exactly 6V it'll shut down.. but the reality is... since LiON batts are pretty much depleted at 7V.. you'll only see less than 1minute of dimming when the thing dies... other wise absolutely max brightness start to finish!

at nominal Vbat.. it is putting 3.87A through the FET with .48V.. so that's 1.86W.. with 25.54W on the bulb.. that means.. 93% efficiency almost immediately... it only goes up from there.

I can't wait to get the prototype for the 'ludicrously simple' driver going but that'll probably be in a month or so when i'm done making nanos... they are sooo cute.. you have to go check out the circuits: http://nano.rouse.com

I can't wait to run some tests with the 1185 and the 1254 bulb!

I've run the 1166 at about 12.5V.. and it pretty much looks exactly like the 1160 at 6.6..

All in all.. thanks to anybody that's helped on this little project.. i will be miniaturizing it to fit into the aluminum cup of the KIU socket soon, probably be like next weekend after i'm done with a couple nano batches.

-aw

 

[andrewwynn]

Here's a pic of the breadboard:

notes:

the skinny resistor where the right red clip is... i'm just using as a test lead.. same for the one sticking up on the right side.. the black is attached there.

The tiny white wire over on the right.. that is soldered directly to the gate of the FET.. there is a red and blue one on the drain and source as well.. i was getting errors of 20+mV when i tried to measure from a nearby hole.

the 4.7k resistors are both gone.. they have been replaced with a jumper from pin 3 (near side left is pin1).. to pin 14.

the skiny red wires on the top over the bulb go up to a 3mm SMD 200k pot i was using before i realized i could get by without it completely... now there is just a 50k resistor from Vcc to the EN pin in parallel with the left top blue capacitor.

notice that i melted the CRAP out of the poor breadboard... I missed it because when i'm running the thing i have a welder's plate in front of the bulb.. oh.. that zener right by the bulb.. that's the welder's plate holder so it doesn't lean against the bulb. Its' really neat to check out the filament when it's lit! (but whoa that plate gets hot too). I think i might wire this ckt up on a perfboard from RS to show people an easy DIY vs the notsomuch DIY that i'll be building with the SOP14 chip.. i will be able to use the SMT caps i bought.. just directly between the pins!

It just keeps getting better and better. I'm going to get this into my 3bore 2D light that runs from 7.2V LiON pack.. and push the crap out of the 1160 and see how long it can take 6.6V and 863 lumen! 560 out the front (using the usual 65% formula).

863 lumen.. i'll maybe choke that back to 800 just to conserve the bulb life.. in any event.. a helova light from a 2D mag!

The spookier model will be the 1166 running about 900 lumen from a 1D light.. it will be 'overdrive' on the batteries so i won't recommend people copy me but it will be a force to be reckoned with to be sure.

The 'next level' is we are working on putting the osram bulb they use in the USL into a 12 cell pack in a 3D mag host that is bored inside to fit it.. but regulate the voltage so we get a flatter runtime and with an add'l cell probably even more runtime... i'll only have like 30-40mV drop on the FET running 100W.. i would put that into the 'acceptable' column.

-awr

 

[NewBie]

Here's a pic of the breadboard:

The 'next level' is we are working on putting the osram bulb they use in the USL into a 12 cell pack in a 3D mag host that is bored inside to fit it.. but regulate the voltage so we get a flatter runtime and with an add'l cell probably even more runtime... i'll only have like 30-40mV drop on the FET running 100W.. i would put that into the 'acceptable' column.

-awr

Well, with an extra NiMH, you'd be looking at actually having to drop that additional 1.1V or so, at 8A load current (I assume thats what the bulb draws). Power = Amps * Current. 1.1V * 8A = 8.8 Watts dissipated (the MOSFET has to burn the power it is dropping). Guess what? The power the linear regulator is burning, is what the additional cell contributes. The advantage, is that you get to stay in regulation longer, before the bulb starts dimming. There is some additional power burned due the internal resistance of the MOSFET, but it is minimal, at about 1/4 Watt. Without the extra cell, you'll still dim like the standard light. The advantage? Less insta-flash with the circuit.