Celebrating the independent kiwi spirit of invention.

Caving Headlamp – Drivers

By Ian Mander 24-27 January 2010, updated 26-30 October, 7-8 November 2010, 24 June 2011, 6 April 2012.

Driver Options

The three drivers considered are based on the following buck regulator chips:

AX2002 (US$6.97 for 4, 0.93A or US$1.99 each, 1.25A at DX). I have several.
EQB8L (since sold out at DX). I have a couple.
PT4105 (since sold out at KD). I have a dodgy one I hope just needs resoldering and have been able to get another.
PT4115 (US$2.20 at DX). I have a couple which are ear-tagged for a 12V project.

The AX2002 has a low overhead and a moderately high maximum output, but is not efficient at low currents. The 1.25A version has a Schottky diode rated at 3A, which means that version is suitable for increasing the output current. The 0.93A version only has a 1A Schottky diode. DX's AX2002 boards have a 47µH inductor, which according to one claim (by Tobias B) makes it more efficient at lower currents than the recommended 22µH inductor.

The EQB8L is more efficient at low drive currents but will present serious problems for switching to turbo mode because its set resistor is on the positive side of the LED.

The PT4105 is no longer produced, and is sadly missed, since it made great low-current regulators. It has a negative side set resistor.

The PT4115 is the replacement for the PT4105 (which is sadly missed). With a minimum input voltage of 8V, it's out of contention straight away. The headlamp has to be able to work with 4xAA NiMH. It does offer dimming, though, so would be a great option if the headlamp used 3*Li-ion cells... I might think about that.

Testing

I've tested the AX2002 and EQB8L drivers at the various headlamp mode current levels, with fresh 4xAA and 3xAA to simulate a flat battery. The PT4105 figures came from a graph in the PT4105 datasheet. Tobias' testing also indicates a discontinued PT4105 driver's efficiency is around those figures.

Mode	Nominal Current	AX2002		EQB8L		PT4105
Mode	Nominal Current	Average Efficiency	Approx. Runtime (both beams)	Average Efficiency	Approx. Runtime (both beams)	Average Efficiency	Approx. Runtime (both beams)
Idle	20 mA	60%*	57 h (2.4 days)	73%	71 h (almost 3 days)	93%	90 h (3.75 days)
Low	60 mA	76%	23 h	86%	27 h	93%	29 h
Medium	180 mA	85%	8 h 10 min	89%	8 h 45 min	91%	9 h
High	500 mA	88%	2 h 45 min	83%	2 h 40 min	88%	2 h 50 min
Turbo	1500 mA	89%	20 min	-	-	-	-

*A linear driver would do just as well in this situation, because the driver input current is the same as the LED current. Perhaps a plain old bomb-proof resistor could be used instead of a driver, since it would be always on. It would be bad for running NiMH batteries flat, but good for not turning off when battery voltage got too low for the driver.

Where runtime most matters is the higher modes, because that's what a caver is most likely to be using, and that's where a caver is most likely to run out of battery power while using them. Runtime differences in the lower modes are not very significant.

For medium and high modes, there's not much in it, but the efficiency of the PT4105 makes it very attractive if not wanting a turbo mode – consider using for a 4xAAA headlamp.

Even idle mode having 13 hours extra using the EQB8L driver is no big deal – I don't plan on being in a cave that long, and that mode would not be used for active caving. With that in mind, and considering how much easier the switching will be, means the AX2002 driver is the better choice. (I'm not saying I won't keep an eye out for more efficient drivers across the board, though, or perhaps a way to use PWM.)

Another problem with switching drivers is that if the driver is left powered, with just the LED disconnected to connect it to the alternate driver, the driver's capacitor in parallel with the LED will be charged up to battery voltage. When the LED is reconnected it could spike the LED.

One possible solution is to use a separate battery pack to drive the turbo mode. This would mean that drivers could be run in parallel and each still output the desired constant current. The AX2002 driver would supply 1 A in addition to whatever driver was used to output the 500 mA of high mode.

AX2002

More AX2002 notes here.

A total of four Cree XP-G LEDs will be driven by either two or four AX2002 drivers depending on the mode.

The AX2002 chip should do quite nicely, but not all driver boards that use this chip are the same, because entry level versions use a Schottky SS14 diode with a maximum current rating of only 1 A. A driver with a 3 A rated SS34 is ideal (US$1.99 at DX). With an off the shelf set resistor of 0.2 Ω, it will be easy to get 1.5 A simply by adding a 1 Ω resistor in parallel, and with the smaller inductor value of 22µH it should be more efficient at higher currents. (However, a 12 Ω resistor is hardwired.)

Standard modes

Each AX2002 driver has a 12 to 12.5 ohm resistor hardwired, to give a minimum 20 mA as long as power is supplied.

Contact resistance in the 3-mode pushbutton could change the output current significantly, to drop the high mode as low as 450 mA. The value is below the sensitivity of my equipment – a good sign – so finding the actual value of the resistors might take some experimenting.

The output current for the AX2002 driver is determined by I_out = 0.25 V / R_set.

Nominal Output Current	Nominal Set Resistance	Hardwired Resistor	Parallel Extra Resistance Made From ...	Contact Resistance	Actual Output Current
20 mA	12.00 Ω	12 Ω			20 mA
60 mA	4.14 Ω	12 Ω			60 mA
180 mA	1.389 Ω	12 Ω			180 mA
500 mA	0.500 Ω	12 Ω			500 mA

Turbo

In turbo mode each LED is driven separately by its own AX2002 driver at 1.5 A.

The output current for the AX2002 driver is determined by I_out = 0.25 V / R_set.

Nominal Output Current	Nominal Set Resistance	Parallel Resistance Made From ...	Actual Output Current
1.25 A	0.200 Ω	0.2 Ω	1.25 A
1.50 A	0.167 Ω	0.2 Ω + 1 Ω	1.50 A

My own tests of an AX2002 board found that with 4xAA, it has an efficiency of 87% at 930 mA and 85% with an XR-E (higher Vf) at 1.1 A. The AX2002 datasheet indicates similar efficiency with increasing output current up to 2 A. Because of the low Vf of the XP-G and the low voltage drop in the chosen cable, the driver's voltage headroom should not be a problem for almost the entire capacity of the battery (perhaps affecting runtime by only a single minute).

The AX2002 has built-in thermal shutdown at 140°C. I don't feel inclined to rely on it instead of having turbo momentary.

EQB8L

When I last had trouble with my LED car boot light I swapped the resistor-driven arrangement to a constant current driver setup. The driver conveniently has a rectifier built in, so I don't need to worry about which way to insert the LED light into the bulb socket, which means it's now a drop-in replacement – nice.

Now, the new versions of these drivers use the PT4115 chip, which turns off under 8 V. That's OK for automotive use, but no good for my headlamp. The EQB8L version which I used for my car, however, keeps right on working at less than 4 V; I've finally got around to doing some testing [January 2010]. The 13.5 V test used a boost power supply (I don't remember even having one [October 2010]), while the other tests, except where noted, used a multi-setting buck power supply. Both power supplies were run from a 12 V SLA battery. The results seem to be consistent.

Input Voltage	Input Current	Input Power	Output Voltage (with cool LED)	Output Current (assumed based on previous measurements)	Output Power	Efficiency
13.50 V	100 mA	1.350 W	3.08 V	350 mA	1.078 W	79.6%
12.82 V (without rectifier)	100 mA	1.282 W	3.08 V	350 mA	1.078 W	84.1%
10.95 V	120 mA	1.314 W	3.08 V	350 mA	1.078 W	82.0%
9.15 V	145 mA	1.327 W	3.08 V	350 mA	1.078 W	81.2%
7.53 V	175 mA	1.318 W	3.08 V	350 mA	1.078 W	81.8%
6.15 V	215 mA	1.322 W	3.08 V	350 mA	1.078 W	81.5%
5.32 V (4xAA Eneloop)	250 mA	1.330 W	3.08 V	350 mA	1.078 W	81.1%
4.64 V (without rectifier)	250 mA	1.160 W	3.08 V	350 mA	1.078 W	92.9%
4.50 V	300 mA	1.350 W	3.08 V	350 mA	1.078 W	79.9%
3.82 V (without rectifier)	300 mA	1.146 W	3.08 V	350 mA	1.078 W	94.1%
3.97 V (3xAA Eneloop)	150 mA	0.596 W	2.94 V	Out of regulation, didn't measure output current.
3.10 V	0 mA	0 W	No output

Interestingly, the net driver efficiency is very consistent across the voltage range tested. However, because the rectifier drops the same volts whatever the input voltage, while there's about 4-5% improvement in efficiency to be gained by shorting out the rectifier over 12 V, around 14% better efficiency can be achieved at lower input voltages. (Apparently the decrease in the driver's efficiency as the input voltage increases almost perfectly equals the effective increase in efficiency from the proportional rectifier losses, leading to a very consistent net driver efficiency.)

The driver is still in regulation at 4.5 V with the rectifier doing its thing. I've added a line for the (theoretical) performance based on the measured 0.34 V drop across each of the two Schottky diodes in service at 13.5 V and 4.5 V input. With the rectifier shorted the EQB8L driver can be about 94% efficient at the voltages I want – very nice.

Set Resistors

Because this driver has a set (or sense) resistor – presently 0.3 ohms – the driver can be easily configured to output 180 mA for my desired 75 lumens using an XP-G. This driver will very likely find service in a more portable device than a car in the not too distance future. (I'll use one of the more recent PT4115 versions in its place for the boot light.)

I believe the output current for the EQB8L driver is determined by I_out = 0.1 V / R_set.

Nominal Output Current	Nominal Set Resistance	Resistors Used in Parallel	Parallel Resistance Made From ...	Actual Output Current
20 mA	5 Ω	5 Ω	10 Ω + 10 Ω = 5 Ω	20 mA
60 mA	1.67 Ω	5 Ω + 2.5 Ω	3.9 Ω + 6.8 Ω = 2.48 Ω	60 mA
180 mA	0.56 Ω	5 Ω + 0.63 Ω	1.0 Ω + 1.8 Ω = 0.63 Ω	176 mA
500 mA	0.2 Ω	5 Ω + 0.21 Ω	0.2 Ω	520 mA

Contact resistance in the 3-mode pushbutton could change the output current significantly, to drop it as low as 450 mA. The value is below the sensitivity of my equipment, so finding the actual value of the resistors might take some experimenting.

Celebrating the independent kiwi spirit of invention.

* This would have been an ad.	When you buy stuff from Asian sellers:
Please don't buy stuff from a country in the middle of intimidating its neighbours.