New 2-Transistor Current-Regulated Boost Circuit

[Ty]

This is my first post here - I've learned a lot from reading this forum, and hope I can return the favor. I've been tinkering with the 1-transistor boost circuit some folks have been using to drive LEDs from a single cell. With some simple modifications (adding another transistor and resistor), I've made it supply a constant current to the LED over a wide range of battery voltage, essentially the whole life of the battery. Basically, it maintains the same brightness until the very end of battery life.

The circuit and notes about it are posted on my web site (I had been looking for an excuse to build one). So far, it's
just in prototype stage - it's on a breadboard instead of in a flashlight, but I think it's a promising idea.

Have a look, if interested at:

http://tyapo.tripod.com/2t_regulated_led.html

To my knowledge, this is an original addition to the 1-T circuit discussed here and elsewhere /ubbthreads/images/graemlins/smile.gif. It is pretty simple, though, so I wouldn't be surprised if others had already done something similar.

 

[OddOne]

Think that may be adaptable to driving a LED array or a 1-watt Luxeon?

oO

 

[Ty]

[ QUOTE ]
OddOne said:
Think that may be adaptable to driving a LED array or a 1-watt Luxeon?

oO


[/ QUOTE ]
You can drive more LEDs with the circuit as shown; just parallel LEDs with the same resistor value as that used for current sensing. The number you can get away with will depend on the exact transistors and coil you use.

You might also put a few LEDs in series in place of the single one in the circuit - especially with reds and their low voltage drop. Whites may not work well.

As for Luxeons, I've simulated this circuit using 3V supplies (and a few other tweaks), and you can manage to get 100 or 200 mA out of it without much trouble (except you need to find some small sub-10-ohm resistors for the current sensing). I'd bet you could make it functional for a 2-cell possibly weak luxeon driver. Then again, by making it somewhat more complex, you could probably do much better.

This circuit isn't going to replace IC switching supplies like those from Zetex, Maxim, etc; it's mainly of interest because the parts are commonly available and cheap...and of course it's interesting if you're just a circuit freak like me /ubbthreads/images/graemlins/wink.gif.

 

[Albany Tom]

Nice!

I have to ask a stupid question. How did you determine the value of inductance of the transformer?

 

[FreeBSDboy]

Could you replace the transistors with something more "meaty" to get a higher current out of it? Or would that also involve modifying the inductor?

And for my last question, how clean is the power out of this thing?

 

[OddOne]

Okay, if this can be made workable for a 1-watt Luxeon Star it would OWN. Sure it won't be a Zetex/Maxim/etc. circuit but hey, it's MUCH easier to get a couple 2222s/TIP31s/etc. from Rat Shack than having to order ten-packs of ZXSC300s from DigiKey.

If you pursue a 400mA capable version of this please let me know - I'd post a writeup on it on OWLI.

oO

 

[Doug Owen]

Prolly not a good idea for high current LEDs......

I seriously doubt this is a good way to drive high current LEDs. Consider the currents involved. Several Amps peak, you'll need 'a lot more iron' in the core and do some serious rework on the windings. Driving high current transistors is also problematical as gains tend to be low.

Consider also efficiency. These converters tend to be in the 50% range, consider the Arc AAA and CMG Ultra. They both draw *average* currents of about 180 mA at 1.5 volts and seem to deliver performance similar to 30 mA at 3.4 Volts at the LED. This means low battery life and heat problems. There are, IMO, sound reasons that such circuits don't often find their way into traditional electronic products.

Mr. Al has given us a couple of simple transistor drivers (using Radio Shack type parts) for those inclined. Say for instance <http://hometown.aol.com/xaxo/page1.html>, My favorite idea for driving Luxeons is, however, his LM334 circuit. Less than $2 worth of parts and we can drive LEDs from 3 NiMH cells at close to 90%. The last few 30 mA drivers I've built have given 50 hours of *continuous, full power* run time from 1850 mAh cells. I strongly suspect that the circuit under discussion cannot come close to running 30 mA for say 15 hours on a single such cell? While I haven't flattened any batteries with any 350 mA versions yet, from shorter runs I expect 4 hours or so.

Or so it seems to me.

Doug Owen

 

[star882]

Re: Prolly not a good idea for high current LEDs......

"Driving high current transistors is also problematical as gains tend to be low."
You can cascade transistors(use a smaller transistor to drive a bigger one).

 

[OddOne]

Re: Prolly not a good idea for high current LEDs......

Of course it's not the most efficient circuit. The Zetex driver MrAl came up with (his CCR-2) will probably hit in the mid to upper 80s.

However, it's a viable option for those that can't deal with surface-mount parts, aren't big on mail-order, have budgets that allow for spending three bucks on a pack of 2N2222s as compared to mail-ordering $35+ in parts at a time, etc. etc. etc. The only part that'd be tricky to come up with will be the toroid core, and enterprising Hams probably have at least one that size or larger in their junk drawer.

Sometimes economy outweighs efficiency.


I bet if a larger toroid is used (or possibly the existing one if it is wrapped with #24), Q1 gets replaced with a TIP31 or similar, and a 2-ohm resistor is used for R2 the circuit could crank out 350 mA. I might have to try it. /ubbthreads/images/graemlins/grin.gif

oO

 

[Doug Owen]

Re: Prolly not a good idea for high current LEDs......

[ QUOTE ]
star882 said:
"Driving high current transistors is also problematical as gains tend to be low."
You can cascade transistors(use a smaller transistor to drive a bigger one).

[/ QUOTE ]

Yup, that fits my general definition of 'problematical'. Adding more parts gets away from the idea, doesn't it?

And you have to be very careful with this idea in this sort of circuit, Darlingtons, for instance (the traditional) won't work due to higher Vsat.

Others have considered raising the power of this modest circuit *over ten times*, not trivial. I strongly suspect that more core will be absolutely necessary, when this one saturates (not good for efficiency, BTW), there's just not much energy stored there. You'd need to raise the operating frequency too high IMO. I've some modest experience, I used to work for Mc Lean Electronics in lovely Fountain Valley California. We made 12 DC fluorescent lamps for RVs (like Fleetwoods). A bunch of lamps, like 1500 a day. The largest user (at the time) of Sylvania 15 Watt tubes. TI would send out an Engineer and a Sales type from time to time to negotiate the next *quarter million* transistors. We bought '3055 type NPN, selected for Betas over 35 at 4 Amps and 4 Volts, Veco over 95 Volts. Paid $.43 or so each 20 or so years ago. The transformer (it was a proper transformer, not some less stable flyback scheme like here) was bobbin wound double C cores of 20, 25 and 625 turns, and about the size of a golf ball. Smaller ones didn't handle the load (about 15 Watts total). The dual 8 Watt tube unit (closer to one Amp) could use a *slightly* smaller transformer, but volume called for the same unit in both models.

Inductors aren't magic any more than other parts. They follow the same rules.

If the idea really is low cost, high efficiency, easy to get parts, I suggest the solution is in a different direction.

Cheers.

Doug Owen

 

[MrAl]

Re: Prolly not a good idea for high current LEDs......

Hello there,

Looks like a very nice circuit for driving the smaller
LED's. Not sure if i would like to use something like
this for driving a LS though, as the relatively large
sense resistor will eat up a little too much efficiency.

I'll probably do a simulation of your nice circuit over
the next day or two and post the results.

Good luck with your LED circuits,
Al

 

[Ty]

[ QUOTE ]
Albany Tom said:
I have to ask a stupid question. How did you determine the value of inductance of the transformer?

[/ QUOTE ]

"uninspired guesswork"

I actually have tried making a bunch of inductors and transformers on these particular cores that I found cheap, so I knew about what inductance was practical with them. Basically, I figured I wanted enough turns to get good coupling, but that would still fit. This lead to about 12 turns, which measured out to just over 300uH on my inductance meter /ubbthreads/images/graemlins/smile.gif

SPICE simulations showed this would be OK, but SPICE's inductor model bears almost no resemblance to reality - cores never saturate, etc. So the real test was to try it out - and it seems to work OK. For an inductor like this, I could probably try between 1 and 20 turns (about what can fit max) in an afternoon to optimize things...

 

[Ty]

Re: Prolly not a good idea for high current LEDs......

[ QUOTE ]
MrAl said:
I'll probably do a simulation of your nice circuit over
the next day or two and post the results.
...
Al


[/ QUOTE ]

I'd like to see what you come up with. I first tried this in SPICE before making it, and results depended on which of the various transformer models I used. I finally ended up using the "coupled inductors" model, which appears to simulate well (but may not match reality!). Simulating oscillators in SPICE can be a real pain /ubbthreads/images/graemlins/icon23.gif

 

[Ty]

Output cleanliness is next to....

[ QUOTE ]
FreeBSDboy said:
Could you replace the transistors with something more "meaty" to get a higher current out of it? Or would that also involve modifying the inductor?

And for my last question, how clean is the power out of this thing?

[/ QUOTE ]

I think this circuit is currently "inductor-bound" - that's likely to be the least ideal of any of the components. Advancements to this particular circuit are probably going to come from better magnetics, if anywhere. Having an infinite beta transistor with zero Vce(sat) wouldn't hurt, though. Let me know if anyone finds any (cheap!) /ubbthreads/images/graemlins/wink.gif

As for the output cleanliness, with the shown values, you might be looking at ~1mA ripple on a 20mA output for 1.5V source (guess based on SPICE sims). That's not too bad. As the battery voltage drops, the ripple is reduced, since each pulse delivers less energy. If you reduce the filter capacitor, the ripple will obviously get worse...

When I get a chance, I'll throw it on the 'scope and measure the ripple for various loads and battery voltages. So far, it seems fairly clean, though.

 

[MrAl]

Re: Output cleanliness is next to....

Hello again,

I did some preliminary simulations of the circuit using
a model of a Nichia type white LED on the output.

I havent been able to determine too much yet except the
following:

1. Efficiency is going to be around 70%
2. The output cap has to be large enough to ensure
proper filtering of the output or else the oscillations
go unstable and the circuit doesnt regulate any more.
Havent determined what values are best, but when
the input voltage increases it makes matters worse.
I guess this means you should do some basic regulation
tests to insure the thing is regulating, and if not
increase the value of the output cap maybe in increments
of 2x.
3. Adding a small RC 'speed up' network around the
bias resistor increases efficiency somewhat because
it forces the transistor out of sat faster.
The difference is probably about 3% or so.
The resistor and cap have to be selected carefully
to avoid oscillation. The added benefit is the circuit
regulates to a lower voltage then without it.

The efficiency picks up somewhat with the Zetex 617
transistor over a standard Vsat transistor by about
4 percent or so. The extra gain of the 617 requires a
larger input bias resistor.

I dont know if i would expect more then 75% from this
circuit however, mostly because of the large loss in the
sense resistor.

It is possible to modify the circuit quite easily to
include an error op amp and reference diode, which
would boost efficiency about 8% or more at the cost
of the added parts. An LM358 could be used as the
error amp so that would still keep the parts cost
pretty low.

Good luck with your LED circuits,
Al

 

[Ty]

Re: Output cleanliness is next to....

[ QUOTE ]
MrAl said:
I did some preliminary simulations of the circuit using
a model of a Nichia type white LED on the output.


[/ QUOTE ]

Cool! Do you have this model available, I'm stuck with fairly generic LED models in my simulations. A luxeon model would be nice, too /ubbthreads/images/graemlins/smile.gif

[ QUOTE ]

...
2. The output cap has to be large enough to ensure
proper filtering of the output or else the oscillations
go unstable and the circuit doesnt regulate any more.
Havent determined what values are best, but when
the input voltage increases it makes matters worse.
I guess this means you should do some basic regulation
tests to insure the thing is regulating, and if not
increase the value of the output cap maybe in increments
of 2x.


[/ QUOTE ]

I found the same thing. I posted a few charts of LED average current for various values of R2 and input voltages last night:

http://tyapo.tripod.com/2t_regulated_led.html

the graphs show the difference between 100uF and 470uF, with 470uF clearly giving better regulation at higher voltages.

[ QUOTE ]

...
3. Adding a small RC 'speed up' network around the
bias resistor increases efficiency somewhat because
it forces the transistor out of sat faster.
The difference is probably about 3% or so.
The resistor and cap have to be selected carefully
to avoid oscillation. The added benefit is the circuit
regulates to a lower voltage then without it.


[/ QUOTE ]

Great idea, I'll have to play with this a bit...

[ QUOTE ]

I dont know if i would expect more then 75% from this
circuit however, mostly because of the large loss in the
sense resistor.


[/ QUOTE ]

Yep, the efficiency is limited to Vled/(Vled + 0.7) since there's always a 0.7V drop across the sense resistor (when in regulation). By making Vled larger (white LED, multiple LEDs in series, etc) you could get greater efficiency - of couse, you can't drive too high a voltage output with a single cell.

I think I need to test the prototype with some real batteries over an extended period of time. Since this circuit will run down to low voltages, you might get more usable life out of a cell than with some higher-efficiency circuits, because you can milk more out of the battery before it shuts down. Maybe, anyhow.

Thanks for having a look at it!

-Ty

 

[MrAl]

Re: Output cleanliness is next to....

Hello again,

If i remember right, i posted a few models a while back
on CP. Doing a search might turn them all up. If not,
I'll have to look them up in my notes and repost.

The 1watt LS had all optimized values too.

Take care,
Al

 

[Ty]

Original source for blocking oscillator design

For those interested, I found the original web site with the blocking oscillator design, and wouldn't you know it, the guy had also built one with this current regulation scheme! Not surprising, really, since it's such a simple addition. Here are the links:

blocking oscillator design

current-regulated design

It took a week to track this stuff down, since a copy of the simpler circuit was put on someone else's page without link or reference. /ubbthreads/images/graemlins/mad.gif

 

[MrAl]

Re: Original source for blocking oscillator design

Hello again,

I see what you mean :-)

Yesterday i was looking at the circuit again and
found a way to decrease the 0.7v drop needed for
current sensing. The 33 ohm resistor drops to
2.5 ohms, meaning the theoretical efficiency should go
up quite a bit. The design works but i have one unanswered
question about the temperature stability of it.
The upgrade requires the addition of one low cost transistor
and one 100k resistor (could be SM types).
If anyone is interested in building and testing this
idea i'll post a schematic.

Good luck with your LED circuits,
Al

 

[Doug Owen]

Re: Original source for blocking oscillator design

Mr Al, please do (he says with baited breath...).

TIA

Doug Owen