TECHNICAL FORUMS > Equipment

High efficiency multi mode analogue LED driver

(1/8) > >>

royfellows:
I would like to first explain exactly what is meant by efficiency and how it affects the average user of lighting equipment such as headtorches and caplamps, and also describe a new lamp project intended to replace 2 existing models.

First, a bit of explanation. If you connect an LED directly to a battery you will probably get your light disappear in a puff of smoke just like a pantomime genii. So it’s necessary to employ some means of controlling the current to the LED or LEDs. This can range from a simple resistor, through linear drivers which are somewhat better, up to quite sophisticated switch mode drivers. The latter in very simple speak does what it says on the tin, it switches on and off at a very high rate and then smoothes the output into a continuous current. An oversimplification, but necessary for the un-knowledgeable to understand it.

Most of these drivers encountered will be whats known as buck drivers. This type is used when the battery voltage exceeds the voltage required by the LEDs and controls the current. As such, they are quite efficient, however there is a necessity for lower brightness levels, usually referred to as ‘modes’. This is usually done by a process called ‘pulse width modulation’ or PWM, it does not matter about understanding this for the purpose of this thread, but take my word for it that in a lot off the off the shelf buck drivers I have tested, efficiency goes out of the window on the lower output settings.
The product description will often boast up to 95% efficiency, the bit they don’t tell you is that this is at the maximum output, go down to the lowest mode and it can be less than 70%, no kidding. Of course, the current draw at that level will still mean that burn times will be quite long, but it is something of a thorn in ones side to think that a third of the power being drawn from the battery is just going to waste.

There are other efficiency factors such as use of under rated components to save space, typically the series inductor which can be quite bulky. High value plus high current. Mine only needs 2.2 uH, I have seen them on buck drivers as high as 40 uH and obviously underrated for the current to get it into a torch.

An attendant issue is that of falling output as the battery voltage decreases. Obviously, there will be a point when the battery voltage falls to that below the requirement of the LEDs, but this is from the word go. Many people will be attracted to the idea of a ‘slow death’ rather than an abrupt loss of light, but the slow death really should not be from the very start?

Anyway, moving on, I have been for some time now considering the design of a high efficiency driver using analogue mode switching. Well necessity is the mother of invention it is said, and matters have been precipitated by the cessation of availability of the drivers I have been using in my popular Scorpion X12, and also the X8.

Now Stenlight uses an analogue system based around a DC-DC converter and series of reed switches actuated by a little revolving magnet. There is quite a lot of info on the Sten website which is quite factual rather than sales BS. Stenlights are not the most powerful of lamps, and this is for good reason. Regardless of the efficiency, the small amount of energy that is wasted is obviously converted to heat, and all inside a tiny chip of a few millimetres. So there are practical limitations as to output.
 
The DC-DC converter is very efficient mainly because there is no external fly back diode, a source of energy loss. But it also has its own internal switching MOSFETs, actually 3 according to the block diagram, hence the heat issue. I overcome this by mounting my converter sort of ‘piggy back’ on the other side of the PCB from the other components. This enables me to heat sink it directly into the aluminium case of the lamp. It has already been extensively tested at full power, 3 amps.

All for the Stenlight, but my lamp users have grown to accept and love my little waterproof toggle switches, so that is the line of experimentation I have followed.
After many hours on the web I sourced a converter suitable for my purpose and my first testing confirmed an efficiency of a remarkable 98%. This was 3 series Cree XP-L emitters powered from a 11.2V input voltage with the driver set to a 200 mA output. This represents an output of about 550 lumens from the 3 serial LEDs for a battery hit of 150 mA. The battery capacity of the X12 is 3500 mAh, so do your sums. Its worth mentioning that some other lamp manufacturers use prepacked cells of much lower capacity that those I use, Sanyo NCR 18650 GA.

The switching is about 850kHz, so my design revolves around a low inductance ground plane, the overall prerequisite being SAD – short and direct. Deviation from this resulted in quite a few failures, and somehow the inevitable trashed components.

The switching methodology uses 2 serial flip – flop circuits giving 4 logic states controlling Rsense, on this design Rsense is not serial with output, but output to ground.
 Thus -
State 1 - off state – fixed default output
State 2 - output one – state 1 plus output 2
State 3 – state 1 one plus output 2
State 3 – state 1 plus states 2 and 3
Then looping back to state 1.

At this time I have arbitrarily set the outputs as 140mA, 220 mA, 1000 mA, and 2950 mA with the outputs switchable between the 2 beams plus one flood (as the old X12) and three flood LEDs, (as the old X8 but 50% more power). I can easy customise these outputs for individual users.

It will be well into the New Year before this lamp becomes available due to the need to continued testing.

The efficiency does vary as I have had results as low as 92%, but the current output does not. Its steady as a rock on the lower settings down to about 9V, switch off is at 8.1V. The intention is to fit in a flashing red that will cut in at about 9.5V.

The current production of Scorpion X12 batteries is a very compact aluminium cased unit that weighs in at 280 grams, so no great weight to carry on your helmet. Capacity is 3500 mAh at nominal 11.1V. Expect long burn times, yes.

I welcome comment from the knowledgeable on here, also questions, although obviously I don’t want to give away all of my secrets, and don’t intend to. Don’t even ask for the schematic.
I also have to say that there could still be unforeseen issues that could turn this whole project belly up.

ChrisJC:
So is this magical IC specifically designed for driving LEDs in a highly efficient manner directly from a suitable battery pack?

Chris.

Cantclimbtom:
Thank you, I understand now why some torches cause flickering/strobing when recorded on some GoPro settings, but wondering why a small capacitor isn't in series with the emitters to avoid that

Maybe this is only a feature for simpletons like me, but after recently trying your 3 position toggle

* Walk Mode
* Low
* Turbo (photo or looking up/down stuff)

IMHO that's unbeatable so I'm puzzled why all these logic states etc it seems a step backwards, each to their own.

ChrisJC:

--- Quote from: Cantclimbtom on November 26, 2021, 04:11:59 pm ---Thank you, I understand now why some torches cause flickering/strobing when recorded on some GoPro settings, but wondering why a small capacitor isn't in series with the emitters to avoid that

--- End quote ---

The circuit I designed (for which I have published the circuit diagram!) specifically avoids that by having a true constant current supply, which is adjusted for the various brightness levels. No strobing.

An awful lot of circuits just flash faster than the eye can see (just). I notice that most car rear lights are actually flashing on and off because when they pass me I see a series of lights rather than a continuous smear.

Chris.

royfellows:

--- Quote from: ChrisJC on November 26, 2021, 03:41:36 pm ---So is this magical IC specifically designed for driving LEDs in a highly efficient manner directly from a suitable battery pack?

Chris.

--- End quote ---

I would say yes after the many hours spent studying data sheets etc, but probably not by the manufacturers intent.
Figures are figures and numbers are numbers.

Navigation

[0] Message Index

[#] Next page

Main Menu

Forum Home Help Search
Go to full version