Anyone implemented a voltage step-up circuit?

[MikeB]

I saw that Marcus was going to evaluate a step-up circuit, and it looks like the new Brinkmann "Wal-Mart special" has one (inefficient?).

I've been thinking about trying one out, to squeeze as many electrons out of a Duracell as possible, and/or run white LEDs on 1- or 2-battery sources.

I've found a couple of Maxim chips which might be useful; they seem to be a lot like the part which Marcus described some time back...

I'm no Electrical Engineer, just a tinkerer. I have no real experience in this area, and I don't even know if there's milliamps to be saved this way (the step-up might just waste too much to be useful).

 

[DaveH]

I'm not an EE either, but I think I can clarify things a bit. There are several factors here, battery conveniece, circut efficiency, LED efficiency and battery life.

The advantage of a step up circut is that you can use one or maybe two standard voltage batteries when otherwise you would have to use a larger number of batteries without the circut to get the correct voltage.

The price you pay is efficiency, driving the circuit will take at least some of the battery power. So that's the trade off convenience vs. efficiency.

Some circuts use power modulation techniques, this gets brightness, and long battery life, but the circut still consumes at least part of the batteries power. ( Eternalight )

The most efficient circut in terms of not wasting any of the power, is if you directly connect the batteries to the LED, like a photon. Luckily, LED's are flexable enough, that is they can sink varying amounts of current. This is what Photon's do.

Even there though, you may be overdriving or underdriving the LED and not getting the maximum light to power ratio.

So, in other words, everything has a tradeoff, and you just need to decide which thing(s) are most important.

DaveH

 

[**DONOTDELETE**]

Some of the things that you have to consider when doing the LED thing are:

Voltage. LEDs have a defined voltage requirement. Not enough, and you either get very little light or none. Too much and the LED becomes an NLED or DED (Non-Light Emitting Device or Dark-Emitting Device).

Current. Related to the voltage thing. The reason the little lites work with the button cells is that they depend upon the fact that they can't provide enough current to achieve the second condition noted above. Two 3-volt cells have 6 volt potential. The LED (white) has a voltage bias from 3.6v to 4.0v. More and you stand a chance of it going DED.

The way around the above conditions involve limiting voltage or current, sometimes considered "conversion". This is kind of similar to an impedance mis-match in a transmission line. Maybe. You get your best efficiency when the source (voltage) is matched to the load. If you had a 3.6V source (steady, no voltage sag, etc.) then you'd have no conversion losses and you could calculate the light time by
Vbatt x Icap / VLED * Ibias
such as, 3 1.2V NiCds w/ 300 mAH capacity -- a poor 1.08 W, but the LED only uses 72 milliWatts (3.6v * 20 mA) which should be about 15 hours.

Limit the current through the LED and the voltage dropped by it will be limited also. When dealing with this, we get in to efficiency. A linear current limiting device (linear regulator) generally has a very poor efficiency, and this is further limited by the drop-out voltage.

Fer instance, the PAL that I love so much... uses a 9V battery, and is supposed to last 20 hours at full brightness. 9V will definitely drive the LED into a non-working state. I don't have the formulas on hand but I seem to recall that for the linear regulator, the efficiency is something like:
Target Voltage / Source Voltage

So the PAL starts out at about 44% efficiency and rises slightly as the 9V battery wears down... In theory, it would work down to 3.6V, but the linear reg. has a drop-out voltage -- where the device drops x volts and you need to have the source x volts higher than the target voltage. If the linear reg. has a 3V drop-out, a 9V batt will work down to about 7V for our LED circuit (4.0V LED + 3 V drop-out voltage). The good news is, at that point, the efficiency has risen. Slightly. It should be about 57% efficiency if my memory is correct. Of the 5.3 W available in the battery, 2.6 have been burned in regulating the current to keep it from torching the LED. If you have a PAL, remove the Sanoprene top w/ a new battery and turn it on HIGH and feel the chips. Hot.

A switching regulator has a better efficiency, perhaps 82% - 90% (?). The time available from a 9v battery now is
9V * 595 mAH = 5.355 WH / .072 W = 74 hours. CAN that be right? Not yet... conversion efficiency = 82%, so max time would be 74 H * .82 or about 61 H

Maxim has some very good circuits, but if you've looked at the components, I think you'll see it's not a project for the home hobbyist. Linear Technology and National Semiconductor also have some good circuits (not as many and not as good).

I'm working on a step-up circuit that would let me use two or even one AA -- parts are murder, and the board is gonna be a b!tc4 for me to fabricate at home... my proj has stopped, but the _theoretical_ results would be 40 hours per AA. AA's cost what... 42 CENTS in bulk. The LED would be biased fully "on" 100% of the time, no voltage sag, no over-current. Efficiency is lower w/ one AA, but think of a fully-biased LED in a single AA package. I think that the two AA would be just fine, and if side by side, it wouldn't roll. ;-) Multi LEDs? Take the available power capacity, multiply by the cct efficiency, and devide by total power reqts.

IF it works, it would be on the order of...

(3V * 2.850AH * 82%) / (0.072 * n LEDs)

4 LEDs would yield 24 hrs from two AAs.

Ah, I dream...

Anyway, did my tired rambling explain anything?

Marc


------------------
If you know what's good for YOU, vote Libertarian.

 

[MikeB]

Thanks everyone for the feedback.

Yeah, Marcus, I was thinking along the same lines (long life & stable voltage/current as long as possible).

I've learned a lot about the linear DC-DC converters, and the switching DC-DC converters sound better, from an efficiency standpoint. But building the circuit with all the components (a coil, maybe a schottky diode, some other caps and resistors, in *TINY* surface mount sizes) seems like a major drag.

I'm definitely not ready to do my own board layout, especially on a budget.

I called Maxim for a price on the eval kit for the 1676 (the 1675 sounds about ideal), but they want $60

to part with one. Oh well.

Again, thanks for the info!

 

[**DONOTDELETE**]

Good review Marcus. I think the PAL's major fault is that they could have put two diodes in for the 9V supply and could have used a low drop out regulator (commonly available.)

The Maxim 1675 looks really good. I got some samples of the 1642 but have not done anything yet. The 1642 is a very similar product. PCB etching should not be too bad if you are careful. You need to watch the etching times carefully and could lay it out with decent "draw" software on a PC and print out the positive or negative with a good laser printer. You could also try the hack method and carefully solder things together in one big heap but the stray inductions of the circuit could be problematic.

If only I had more time rather than having to work on other circuits for a living...

 

[**DONOTDELETE**]

Yeah, switching is the way to go. Why waste power as heat to get a voltage that a diode can handle? Re: Maxim... yeah, I've been that route, and that's why I don't have an eval board. Plus, when you get it, it's configured for the "standard" output and you have to remove one of the microscopic resistors and find a replacement. They're about the size of deer ticks. Tiny. Definitly not the project for a home experimenter. I don't think it's something I'd want to try w/ the hand-wired techniques.

Re: the finished size. Given the proper equipment, I could make one that sits on top of an AA. With the LED.


Someone sent me an email asking for the details of the cct. (how does it work, etc.). I tried to email him back and got a "too many files open" from his mail server, so I'll post my attempt to explain it here.


Subject: Re: Step-Up Circuit....

Doug,

My best guess at the Infinity (guess, because they've started sealing the electronics, now) is that it a simple voltage doubler. The LED WILL bias "on" (er, sort of) w/ a lower voltage. Try it w/ a white LED two new Alkaline batteries... you will get SOME light output.

The circuitry I am trying to use is kind of like a charge pump, but w/ some additional "logic" and switches, rather than just a resonant circuit. Most basically, there is a clock/ square-wave generator, there is a voltage comparator, an inductor/ capacitor cct, an internal voltage reference, and a diode. When the cct turns on, the comparator checks the output voltage against the reference, and if Vout is lower, it switches current to the inductor. After the current rushes into the inductor, the cct switches the input to gnd and the inductor, trying to keep it's current state, discharges through the diode/capacitor part of the cct. The cap is across the output -- it's voltage starts to rise, and the cycle begins again. When Vout is where it's supposed to be, the comparator doesn't fire, the inductor doesn't fill, and doesn't discharge across the cap. This is done at 200 kHz or higher... the "stepping" is very small and the resulting ripple is also very small. You can get a sub-1% voltage regulation that way. Down side is, w/ the lower input voltage, the efficiency is somewhat lower, the output current can be limited, and you need to "bootstrap" the input to the output to bring Vin up. Whew. Anyway, no "doubling" or other multiples of the input voltage. Vout is independent of Vin (to a wide extent) and is dependent upon the internal Vref and a resistor network.

The proj is on hold because I don't have any of my equipment or supplies, and don't have the requisite parts to experiment further. The resistors are about the size of a deer tick, and the caps are about the size of the OLD fashioned (very tiny) chiclettes... I don't have the proper resistor values to make my only cct run at 3.6v -- it's biased at 5v.

--- Doug Ricci <[email protected]> wrote:
> So, how does this circuit you mention > below, differ from the one in
> the Infinity? Is the LED on the Infinity > not biased fully on?
> This would make since, since the Infinity > is rather dull in brightness.
>
> Thanks,
>
> Doug

------------------
If you know what's good for YOU, vote Libertarian.