Custom 17mm 5Amp PWM-less linear driver-preliminary

I did some tests yesterday.

It was battle between nanjg(3 Amp version) vs. this driver (PWM vs. PWM-less) at low currents.

I used xm-l u2 on heatsink with small tir and hs1010A for lux measuring(at some distance).

First,I measured current and lux for nanjg at low (5%) mode:



Then,I did the same with this driver(current is as close as possible):



That's ~33% increase in brightness compared to 3Amp nanjg at same current.

Similar,I wanted to see at which current this driver will have the same brightness as with 3A nanjg:



3A nanjg consumes ~31% more current to get the same brightness.Same result as test #2 if we consider measurement errors.

Answer to question why this happens is:

Efficiency of xm-l u2(from cree pct) at 3Amp,25C: 98lm/W

Vf of xm-l u2(from cree pct) at 3Amp,25C: 3.34V

Efficiency of xm-l u2(from cree pct) at 0,150Amp,25C: 165lm/W

Vf of xm-l u2(from cree pct) at 0,150Amp,25C: 2.68V

Nanjg runs led at 3Amp on every mode,this driver runs led with constant current(no pwm).

Overall efficiency gain can be calculated as: (165/98)*(2.68/3.34)=1.68*0.80=1.34 or 34% which is very close to measured values.You can see that led is 68% more efficient at 150mA vs 3Amp,but driver burns more voltage because lower Vf of led at 150mA.Overall,gain is still pretty impressive.

BUT,to be fair this test should be 5Amp stacked nanjg(or DD fet driver) vs. this driver,since this is 5Amp driver.

Similar to upper calculation (this time I used some numbers from match graphs,since cree don't show 5Amp numbers):

Efficiency of xm-l2 u2 at 5Amp,25C: 84lm/W

Vf of xm-l2 u2 at 5Amp,25C: 3.8V

Efficiency of xm-l2 u2(from cree pct) at 0,150Amp,25C: 188lm/W

Vf of xm-l u2(from cree pct) at 0,150Amp,25C: 2.8V

Overall efficiency gain can be calculated as: (188/84)*(2.8/3.8)=2.23*0.74=1.65% or 65% increased brightness for same current or reduced current to get the same brightness.

And 65% reduced current means at least 65% longer runtime (that's like you have panasonic 3400*1.65=5600mah).

This is example for just one low current (150mA),but conclusion is pretty obvious.

When you consider one more important fact,that 5Amp pwm driver losses regulation very quickly(because of high Vf),pwm-less driver can maintain constant current/brightness until battery is empty (again because of very low Vf at low currents),advantages are even more clearer.


Let's consider one more case: 5Amp buck driver(pwm regulated) vs. this driver.

You'll expect that buck is always superior compared to linear when we talk about efficiency.Let's assume 90% efficiency for buck.

At low mode (for ex. same 150mA) we have next data for buck driver which is pwm regulated:

Efficiency of xm-l2 u2 at 5Amp,25C: 84lm/W

Vf of xm-l2 u2 at 5Amp,25C: 3.8V

Driver efficiency:90%

Input voltage:4.2V

Led power=0.15A*3.8V=0.57W

Battery power=(1/0.9)*0.57=0.63W

Current draw from battery(voltage depended,using 4.2V)=


I used 4.2V because this way I got same batt. current as lin. driver,so it's easy to compare.

Led lumen generated=84lm/W*0.57W=48lm


This driver:

Efficiency of xm-l2 u2(from cree pct) at 0,150Amp,25C: 188lm/W

Vf of xm-l u2(from cree pct) at 0,150Amp,25C: 2.8V

Input voltage:4.2V

Led power=0.15A*2.8V=0.42W

Battery power=0.15A*4.2V=0.63W

Driver efficiency=2.8/4.2=67%

Current draw from battery=150mA

Led lumen generated=188lm/W*0.42W=79lm

So,at same battery current of 150mA 90% efficient pwm regulated 5Amp buck driver will generate 48lm,while pwm-less linear driver will generate 79lm.That's 79/48=1.65 or 65% more,same as above calculation for 5A nanjg or DD fet drivers.

Note that linear driver efficiency is just 67% vs. 90%,but 188lm/W vs. 84lm/W of led efficiency totally changes final efficiency result(and that only matters).

The point of this calculation is to show that there is no advantage of having high current 5Amp buck driver(at least not for single li-ion),unless it's also pwm-less controlled(as we can see from various driver threads,even much simpler pwm controlled buck is PITA,not to mention size and cost).Also,that hypothetical buck driver looses regulation very quickly,since it can't increase voltage.

Stop showing off and take my money! :D

Led4power, really a great driver, I'm sure we all would appreciate something like this!

Since there isn't something as "enough", I'm going to ask the questions:

1) Would this be able to run a MT-G2 with the Zener mod done? (in other words, is it zener mod compatible?)

2) Changing the sensing resistor, what is the maximum amount of amps the board will safely take? I'm thinking 3 x XML2 (in parallel) driven from 1 x 26650.

Thanx, really looking forward to the release ;)

looks like the driver does as expected, nice!

Personally, I'm not that picky on constant regulation, I sort of like direct driving a flashlight with the output dimming as the battery drains, like the good old incan days :-) . But I do like efficiency a lot, that would be my reason to use this driver.

Nice very very nice…quite efficient!

1)zener mod doesn't work correctly with this driver because driver constantly adjust some parameters based on battery voltage to maintain costant current during discharge.

Something like this would work:

2)It's hard to tell,heat in mosfet and sense resistor are limitations,and they strongly depend on how good driver is mounted/cooled,voltage difference between battery-leds,and amount of current.For mosfet there is a point of max. dissipation,where (Ubatt-Uparasitic-Uled)*I=max.For example you'll expect that max. disipation is at max. current,but it isn't.

In my tests,driver driver dissipates more heat at 1.1amp than at 5amp,and that's because led needs more voltage,so less voltage is burned in mosfet.

On the other side dissipation in sense resistor is I^2*R.I use metal alloy 1206 sense resistor with 1W@25C rating(this is much more than common 1206 resistors-0.125W).

It depends on all that.


I will take at least 10 drivers.

-modes: 0.3–1.5–20–100,standard half press to increase mode
-“turbo” mode option (140%4.2V,drops linearly to 110%2.8V),turbo enable via solder moon 2

I’m not quite getting this turbo mode. The driver is rated for 5A. So I would assume that 100% = 5A. How can turbo be higher than 100%
Is the 100% level less than 5A?
Is turbo just a timer to drop back from 100%?

Wait a second, why did you take an efficiency of 84lm/W (corresponding to 5A) for the buck driver?
In a buck driver, if the average output current is 150mA, the current will vary around that value in a sawtooth shape. The current to the LED will be very close to 150mA all the time.
Even if the ripple is very bad (caused by a too low value inductor), let’s say that the max current is 300mA and the minimum current 0mA (averaging 150mA), the lm/W value to take will be very close to 188lm/W.

I agree that a buck driver uses PWM, but it’s only to control the FET. The inductor smooths things out and the output current is very stable. That’s the whole point of a buck driver.

Either you did a mistake, or did I missed something?

I took this picture out of HKJ review of a buck driver which clearly shows the sawtooth shape and minimal ripple. There is not axis annotations and values for the current unfortunately. :open_mouth:

The blue curve is lux, measured with a photo diode, not current. There is no annotations, because the values are completely uncalibrated.

Generally buck and boost drivers delivers a smooth current to the led, but there are exceptions.

And sometimes pwm is used for brightness regulations.

Thanks for your input HKJ.
Did you review a buck or boost driver that used a separate PWM signal to regulate brightness? Do you have a link? I’m curious to see what lead the designers to do that…

Yes, many. Check the table: Look in type, max. modes and pwm columns.

There is a problem in that link because of the space replaced by %20. But I got around. :slight_smile:
My jaw just dropped on the floor. Why on earth would they design these drivers (I found more than 5 on your website) with double PWM, one at high frequency to switch the inductor, and one at lower frequency to manage brightness. I just don’t understand. They already have a feedback loop to get a constant current, so it shouldn’t be hard to reconfigure the comparators to get a lower current. That would allow better efficiency, less acoustic noise and no flicker. Is there a reason that I don’t see to justify that design? Most drivers you tested work this way so there must be a reason…

The reason to do it this way, is because the microcontroller is separate from the buck/boost regulator, i.e. not part of the control loop.

It is easier to just switch the buck/boost regulator on/off, instead of adding a circuit to adjust the control loop. There is also the question about space for the components.

I hope one driver is already on its way to HKJ :wink:
His authoritative driver testing reveals much we need to know.

I didn’t think of it this way. Thanks HKJ.

HKJ answered most of the question,I will just add another reason why some switching drivers use "additional" (low frequency compared to switching frequency) pwm for dimming:driver components are optimized for one current,usually for best efficiency at that current.To get 50mA CC mode from 5A switching driver,duty cycle must be extremely small,order of magnitude of mosfet rise and fall times,and this reduces efficiency a lot.

The point is,no matter which type of driver,linear or switching constant current on all modes has better overall Lm/W,even if drivers have lower efficiency at those currents,compared to pwm regulated ones.

I’m sorry but I disagree. The duty cycle of buck drivers follows this rule (in steady state): D=Vout/Vin
And that’s true for any current.
Given that the voltage of the LED (Vout) at 5A is about 3.5V and at 50mA it’s about 2.5V then the difference in duty cycle is not significant.
Compared to a system with double PWM control:
-A smaller current will reduce losses caused by parasitic resistances (FET, inductor,…)
-But that will be partially compensated by the switching losses being greater (not much)
-and if the current ripple is greater than 50mA then the saw tooth shaped current will cross the 0A line (current in inductor reverses for a small portion of time). That causes a lot of losses. It can be prevented by complexe systems though.

I totally agree on that point. Plus there is no acoustic noise or flicker. :slight_smile:

EDIT: I’m just questioning the difference in efficiency between a buck driver an your linear driver. That doesn’t mean that I don’t like your driver, because I do! It is more compact than a buck driver and probably cheaper to build in the end. Plus it has every function that I could dream of. :slight_smile:

Sorry if I have missed it, but any idea when these will be available for purchase and is there a definite price for them yet?

So far most updated info on 1st page last paragraph:

and this:

So I guess…. not yet :smiley:

Maybe that I missed it, but is It possible to reduce the max current to say 3A? Some flashlights don’t have the thermal dissipation needed for a 5A driven LED.