Want to increase brightness of LED headlight

I have one of these magnifiers and they're sort of almost useable at 2.5x but the light isn't very bright at all. I attached my bench power supply and determined that at 3.6 V a current of 20 mA would flow which is my guess at what the LEDs are rated for (maybe it's not; all I know for sure is that they're cheap, white and 5 mm). At 3 V it's 7 mA and at 2.7 V (batteries only a bit discharged) it's 2 mA!! I want to use a little wotzit to increase the current to 20 mA but I have a question regarding which wotzit to use.

It needs to be tiny to fit into this battery compartment and the current contenders are:

QX5252F available in TO-94 form
YX8018 available in TO-94 form
ZXSC380 available in SOT23 form
Alf available in pog form

The ZXSC380 has the advantage of a teeny tiny package whereas I think the QX5252F and YX8018 have a leg I could somehow connect to prevent overdischarging NiMh batteries. Basically a different solution is needed for alkaline versus rechargeable batteries, I think.

Could anyone tell me alternative ICs or how to connect the TO-94 parts to not discharge past 1.6 V (0.8 V per cell)?

 

I'm using cheap alkalines but I want to use rechargeables because then I don't need to keep remembering to switch it off when I don't need it, Fancy Energizer batteries sound relatively expensive but I already have a few rechargeable AAAs that aren't in use.

I think I can use an SMD inductor. I have access to a microscope to be able to solder the parts together,

Are you sure NiMH aren't damaged by discharge below 0.8 V? Many sources in my first couple pages of Google results suggest they would be.

 

Thanks! I'll be sure to check the inductor specifications for saturation currents; there's one series I'm thinking of using which have currents in excess of an amp.

It occured to me that the two LEDs must be in parallel in which case they might be able to take 40 mA total and not just 20 mA so that's cool. I wish the current-voltage curves of the ZXSC380 were flatter nonetheless.

Graph: http://imgur.com/QIf18jQ.png

 

Oh I see, thanks. The frequencies seem really high so am I right in thinking a small ceramic capacitor would smooth out the peaks so the LEDs wouldn't see large currents? Also I think I would need to use two ZXSC380 chips in parallel to provide the current I want to both LEDs? Could they share a single inductor? Presumably they would get half the microHenrys each?

Does that boost voltage somehow? Or is it a low-battery cut-off?

The only datasheet I can find for the QX5252 is this: http://www.mikrocontroller.net/attachment/158139/QX5252.pdf

The LEDs need 4.1 V to get 40 mA of current to them. That's 164 mW. I'm providing 2.4 V (dropping to ~1.6 V) .

Inductor needed at start = 2 x 2.4 V x 10E-6 / 0.164 W = 29 uH. Is that calculation correct?


Components are pretty cheap on AliExpress so I think I could try both approaches (in about two months when the components arrive!).

 

I have the QX5252F IC and it's in a circuit where it increases the current to the LEDs up to 29 mA depending on the inductor used but there seems to be a lot of current bypassing the LEDs, going through the QX5252F. Is that normal? My inductors are cheap, low power, moulded axial ones. Are they appropriate or do I need lower resistance ones?

33 uH (0.9 Ω) = 27 mA through LEDs and 140 mA total consumed
68 uH (2.0 Ω) = 22 mA through LEDs and 160 mA total consumed
68 uH (1.1 Ω) = 29 mA through LEDs and 150 mA total consumed (higher power/lower resistance inductor)
100 uH (2.8 Ω) = 15 mA through LEDs and 160 mA total consumed

Image not showing for some reason: http://imgur.com/ml4OtBM.png

 

Thanks. I'll try to find larger inductors to test.

I forgot to include the capacitor in my diagram which I think helps stabilise the current measured (updated diagram attached). The current is measured immediately before the LED. Do I need to consider the frequency of the circuit to choose a suitable capacitor? What I have is some kind of ceramic capacitor that says:
104
0M0

It looks like this and I thought it was 104 uF but maybe I should find a bigger one.

 

Thanks! I didn't think about the high frequency draining the capacitor. Is there nothing else I can do to smooth the current in lieu of a capacitor?

I might be able to use an oscilloscope at my local Hacklab/Maker space to measure the current.

 

Is the current reading on the power supply affected by the frequency? I put a capacitor across the PSU and there was no change in the current displayed. I think I'm gonna use a 68 uH inductor since the LEDs are then plenty bright and the power supply reports a 40-50 mA consumption which is presumably fine. Thanks for your help. I'm pretty much all the way there now.