Project: 1.5V White LED Drive

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
Project: 1.5V White LED Driver
By: John S Rohrer

A result of many attempts over several years, this circuit alternately drives two white LEDs using a single alkaline cell (1.1 to 1.6 volts). A typical alkaline D cell will drive these two LEDs for over 200 hours while producing 10 candelas of light from the two LEDs combined. (Plenty to see in the dark.)

The complete utility light assembly adds about 25% to the volume of a D cell itself, and has survived repeated physical torture tests.

This design is for a small flashlight that is reliable, manufacturable on a small scale, and which provides more than 200 hours of light from a single alkaline D cell. Such a light would be suitable for emergency situations. The D cell is chosen because it is available worldwide. Alkaline D cell voltage starts at 1.6 volts (V), with most of the energy expended at 1.2 V, and virtually no energy left at 1.0 V. For maximum life, this design will function down to 1.0 V.

White LEDs are more efficient than incandescent bulbs, more vibration resistant, and last much longer. A downside is the loss of the pleasing visual spectrum of an incandescent bulb. LEDs are also chosen because the 5 millimeter (mM) or T-1 3/4 package is available with a narrow (15 degree) viewing angle, eliminating the need for focusing reflectors or lens. A 15 degree viewing angle illuminates a 1 foot diameter area 6 feet away from the LED. This may seem small, but a beam covering a 2 foot area would be only 1/4 as bright. So the choice is concentrated brightness.

Given that LEDs become less efficient at higher currents, the best way to power an LED would be continuous direct current at about 60% of the LED maximum current. But the conversion from 1.5 V to 3.5 V requires oscillation or pulsing. The pulsed drive should have as large a duty cycle as possible to reduce the peak currents versus the average current in the LED, and thus maintain LED efficiency.

There should be no perceptible flickering in the light output. So the pulse rate should be well above visual sensitivity, which is about 60 times per second.
Some transformers and ceramic capacitors are microphonic, converting electrical signals into audible ones. Since noise is not desirable, the pulsing frequency also should be above the human audible range, which is roughly 20,000 times per second (Hz).

An easily buildable flashlight implies common mechanical parts, with no machining.

REASONING

Altho’ white LEDs require about 3.5 V to turn on, they are easily destroyed if the voltage is forced higher. In the simplest circuits that can do the job, a single inductor stores energy in a magnetic field. The inductor energy is released as current to drive the LED, with the voltage determined by the LED forward drop. The time required to store the energy at the lower cell voltage is greater than the time to release it at the higher LED voltage; specifically, the storage time divided by discharge time is proportional to the LED voltage divided by input (cell) voltage. This 3:1 to 4:1 ratio can be reduced by placing the inductor step-up voltage on top of the cell voltage, dropping the ratio to 2:1 to 3:1, but even this is less than 50% duty cycle for the LED. The problem can be eased by a tapped inductor (transformer), which decreases the time required to store the magnetic energy. Such designs are common, yielding a simple circuit that supplies the needed current during each charge-discharge cycle.

The tapped inductor can also be employed as a transformer to provide voltage to drive an LED while magnetic energy is being stored in the transformer magnetic field. Thus LEDs can be driven during energy storage and during energy release. But the transformer output voltage is of opposite polarity for energy storage versus energy release. Using a single LED implies diodes or transistors to steer the alternating current to it, but these would waste power. Two LEDs paralleled in opposite directions across the transformer output is more efficient. One LED would be on while the other is off.

If the transformer turns ratio is 4 to 1, then less than 1 volt can produce enough voltage to drive an LED during energy storage. This means that the LED powered by transformer action during energy storage will produce light for cell voltage down below 1 V.

IMPLEMENTATION
The following schematic is the result of several "Not that way!” experiences. For instance, transformer T1 is relatively expensive; however, when the assembly time and effort to produce a replacement are considered, it is a bargain.
View attachment 149212

upload_2018-3-28_16-20-49.png

Circuit Operation
When switch SW1 is turned on, DC current flows thru T1 winding 2-1-4-6 and R3 to the parallel combination of DZ1 and Q1 base. In other designs, the current increases until limited by transformer saturation. This design has a specific current limit implemented by zener diode DZ1, transistor Q1, and the paralleled emitter resistance R1-R2. Forward-biased DZ1 has a higher voltage drop than the Vbe of Q1 because the junction doping for zener diodes raises the forward voltage drop over that of a comparable transistor base-emitter junction. This voltage difference can be increased if a small area DZ1 is used with a large area Q1. The difference can be 40 to 60 millivolts (mV). This difference causes most of the R3 current to flow into the base of Q1, turning it on.

As Q1 turns on, the voltage across winding 2-3 of T1 increases. Transformer action produces four times that voltage across winding 3-1-4-6 and thus the LEDs. This drives Q1 on even harder thru R3 (and C2, which speeds the transition), until LED1 is turned on. Now there is a constant voltage momentarily across winding 2-3 while Q1 is saturated, and the current thru 2-3 increases at a rate determined by T1's inductance. As the current increases, the voltage drop across R1-R2 increases until that voltage plus Q1 base-emitter voltage begins to equal the drop across DZ1. This diverts more current into DZ1, limiting Q2 current. Measurements with an FMMT617 transistor, an R1-R2 emitter resistance of 0.5 ohm, and a FLZ5V1A zener diode yield a Q2 current limit of about 90 mA. This implies that the peak voltage across DZ1 minus the voltage across Vbe1 is about 90 mV / 0.5 ohm = 45 mV.

With a 1 to 4 reduction in current due to transformer action, the LED1 peak current is 23 mA. Each LED is on for only part of a cycle, with the current ramping approximately linearly (constant voltage) from zero to peak and back to zero during the "on" time. So a 23 mA peak current with linear current ramping averaged over a full cycle (0-23-0 mA) corresponds to an average current of about 6 mA per LED. The typical operating current of a 5 mM white LED is 10 mA.

The DZ1 current limit stabilizes the light output versus cell voltage. However, the time to store this energy each cycle varies with the cell voltage, so there is some brightness variation versus voltage due to this. But there is less variation than with no current limit. The current limit also prevents transformer saturation, which would lower efficiency.

As the current limit is reached, the "rate of current increase" thru T1 becomes zero, so the voltage across pins 3-6 drops to zero, as expressed in the equation dI/dT = V/L. This shuts Q1 off, and the magnetic energy stored in the transformer produces a negative output voltage across pins 3-6, which powers LED2 until the energy stored in T1 inductance is depleted. Then R3 supplies base current to Q1 again, and the cycle starts over.

When the cell voltage drops below 3.6 V / 4 = 0.9 volts, there is insufficient voltage to drive LED1, but the transformer still delivers energy to LED2 on the second part of the cycle during the magnetic field discharge. This provides an end-of-battery-life indication as LED1 goes out but LED2 stays on.

If Q1 is oscillating, it will continue to do so until the cell voltage drops below 0.2 volt because transformer T1 steps that up to 0.8 volt, which is enough to turn Q1 on to start a new cycle. And current is delivered to LED1 [Edit was LED2] each cycle, altho' it dims considerably as the battery voltage drops. And because the cell will usually recover from 0.2 volt to above 0.7 volt with a little rest, it will produce some light from a cell that would be totally dead in any other device.

Note: A transformer's ability to hold a pulsed output voltage constant versus time is measured by its "volt-microsecond" capability. A transformer with a 50 V-uS rating can hold 4 volts for: 50 V-uS / 4 V = 12.5 uS, which is the half-period of the ~ 40 KHz operating frequency.
C1 provides a low impedance across the D cell at higher frequencies.

Prototypes have been built using a commonly available (Eagle Industries) plastic D cell battery holder with the printed circuit board mounted to the battery holder. The shape of this assembly allows for convenient positioning of the flashlight when it is set down, and it fits nicely in the hand. The assembly does not break despite multiple 3 foot drops onto concrete.

The circuit draws about 80 mA at 1.6 V and 35 mA at 1.0 V. A D cell typically stores 12 ampere-hours (AH), yielding about 200 hours of life. Eight hours use a day means 25 days of life.

The brightness at 1.6 V decreases to about one-half at 1.2 V.

Costs
In quantities of 100, the circuit board runs about $4.00, the switch $3.00, the transformer $2.70, the battery holder $1.00, with miscellaneous parts adding another $3.00, for a parts total of $13.70.

With this, all design goals have been met.

upload_2018-3-23_15-33-9.png

upload_2018-3-23_15-31-59.png

upload_2018-3-23_15-32-40.png

upload_2018-3-28_6-27-41.png
Above: Component side copper which has all of the circuit connections.

-------------------------------------
I note an error in the schematic. The original design used two 1 ohm resistors in parallel (R1 and R2) because 1 ohm resistors were easier to obtain than 0.5 ohm resistors. But today 0.5 ohm resistors are commonly available, and only an R1 of 0.5 ohm is needed, not any R2. See attached schematic.

upload_2018-3-28_16-28-15.png
 

Attachments

Last edited by a moderator:

takao21203

Joined Apr 28, 2012
3,702
The pricing is pretty exorbitant for 100pcs

The PCB should not be more than a dollar rather much less.
The switch would cost me 30 cents if i use a rocker switch like shown, I paid about 20 for 1000pcs small 3 pin kind.
Transformer for $3 a piece?

13.70 as parts cost for 100pcs? Maybe 25 or 30 final price?

You should try to bring the total to a dollar or so.
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
The pricing is high. This project was about a light that could be built on a small scale, like in 100 piece lots at home. The price estimates were done more than a few years ago. I was simply reporting on materials costs then in small quantities from suppliers like Mouser and Digikey. Since I intended for the light to be reliable, I chose a more expensive switch type. The pc board estimate was from US and Canadian suppliers.
 

takao21203

Joined Apr 28, 2012
3,702
The shell of the IBM PC would survive several Millenia for sure, massive sheetmetal and the total weight is 14kg or so.
While nowadays, tablets last a year or so then the screen breaks or the USB connector and battery wear out.
Solar garden lights from discount chains at best last a season then they corrode cell and battery are too small.

They had a production line in Scotland but closed down since a long while, people didnt want to invest as much as for a car or simply didnt have that much. For its time it was a very large amount of money, so a PC was not affordable for most.

But cost optimization might not be your design goal.
 

PeteHL

Joined Dec 17, 2014
473
Being fascinated with electronics and having some skill at it, at one time I thought that I might be able to become gainfully self-employed selling some small electronic device. After developing the design and construction of a LED flasher, I attempted to estimate what a small scale cost of production might be, and what price I would have to charge for it to make a decent wage for myself.

Sadly, what I concluded was that even if I devised an excellent flasher in terms of design and construction, I simply could not compete price-wise with similar items that are manufactured, mostly because I could not purchase parts for the same price that the large scale manufacturer can. You should come to the same conclusion.

There is also the matter of the case holding the electronics. A small scale operation simply cannot touch the finished quality of casing that a manufacturer can produce. Technically your design looks excellent. But the problem is that your device does not have the appearance that would allow it to compete with similar lights that are manufactured.
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
All that you have said is true. However, family and friends find good use for the lights and I am satisfied that they are enjoying good quality, serviceable utility flashlights.
 

Audioguru

Joined Dec 20, 2007
11,248
A solar garden light that costs one dollar will do it. It comes with a solar panel and rechargeable battery cell that can be thrown away.
It has an IC that operates its voltage stepup with an inductor and the very cheap white LED can be replaced with two good ones in parallel.
 

PeteHL

Joined Dec 17, 2014
473
A solar garden light that costs one dollar will do it. It comes with a solar panel and rechargeable battery cell that can be thrown away.
It has an IC that operates its voltage stepup with an inductor and the very cheap white LED can be replaced with two good ones in parallel.
Certainly the rechargeable cell (Is it a single cell?) can be thrown away, but it should not be sent to a landfill but rather recycled. There are very nasty chemicals in batteries that will eventually come back to bite us if allowed to contaminate the ground and then also possibly ground water.

My understanding is that even primary batteries are recycled in Europe. At least the Europeans have some sense.
 

tranzz4md

Joined Apr 10, 2015
310
Even at $25 I prefer the single AAA cell light I get at REI. I know it's a bit much, but it's a fantastic light, one of the best this 60 year old electrician has ever owned, and FAR better than any MagLite ever was.
 

DickCappels

Joined Aug 21, 2008
10,153
This is a good example of what a project description should look like.

What is not apparent to me is: What limits the current through LED1?
 

PeteHL

Joined Dec 17, 2014
473
That is an interesting not-run-of-the-mill-design, I suppose, and a little bit difficult for me to follow how it operates, especially the transformer. In your circuit operation section, when you refer to Q2, you must have meant to write Q1.
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
The current limit in the LED utility light
A constant current source may be created using a ground-referenced voltage reference, one bipolar transistor (BJT), and a resistor placed from the emitter of the BJT to ground.
This is common practice, and the collector current Iout = (Vref - Vbe1) / R, where Vbe1 is the base-to-emitter voltage drop of the BJT. Vbe is approximately 0.65 volts. For cases where Vref is much greater than Vbe1, then a varying Vbe has little effect, and the current Iout is forced to be constant.

The voltage reference V itself may be created using a zener diode DZ1 and a resistor R3 from the power supply. R3 would provide an ultimate limit on how much current is available to the base of Q1. Assuming that R3 is small enough to supply all the current needed for DZ1 and Q1, if VDZ1 has a temperature coefficient, that will contribute to the temperature drift of Iout as well as Vbe1. If VDZ1 has the same temperature drift as Vbe1 in mV/ºC, then subtracting them will give a temperature drift of zero, and Iout will be temperature stable. But, zener diodes have temperature drifts, depending on their breakdown voltage. To get a drift of -2.2 mV/ºC (to match Vbe1) requires a zener breakdown voltage of about 6.8 volts. This was a common way to create temperature stable voltage and current references in the 1960s, both with discrete components and in monolithic form. But the minimum Vsply was about 7.0 volts.

As a side note, a forward biased AlGaAs red LED also has a -2.2 mV/ºC temperature coefficient and a forward voltage drop of about 1.65 volts. So using it in place of a zener diode can produce a temperature stable current output by using the AlGaAs LED as the voltage source at the base of the transistor (driven thru a resistor R3 from the supply voltage). And Iout = (VLED - Vbe) / RE = (1.65 V - 0.65 V) / R1 = 1.0 V / R1. This produces a temp stable current source for voltages down to about 2 volts.

This still does not explain how to implement a temperature stable current source with supply voltage as low as 1.0 volt. CRUCIAL INFO: Zener diodes have a higher forward voltage drop than comparable area base-emitter junctions, along with similar temperature coefficients. So a forward biased zener diode may become the voltage reference at the base of the BJT. However, the voltage difference is measured in a few tens of millivolts. It may be increased by using a very large BJT, which has a lower forward Vbe because of its large area, or by using a smaller area zener diode (not so possible because zener diodes have to dissipate power). So a 1/2 watt zener diode is used along with a 3 ampere rated BJT, and the difference turns out to be about 45 mV.

With R1 = 0.5 ohm and a 45 mV difference, Iout = 45 mV / 0.5 ohm = 90 mA as the current limit of Q1. As far as base current is concerned, an R3 of 1 Kohm will supply a base current of (1.2 V - 0.65 V) / 1 Kohm = 0.55 mA at Vsply = 1.2 V. This means that the required minimum beta in Q1 would be 90 mA/0.55 mA = 163.
Fortunately the FMMT617 BJT is rated at a minimum beta of around 250 at 90 mA.

Thus, this little circuit limits peak current to about 90 mA, and transformer T1 steps the voltage up 4 times and the current down 4 times so that the current in LED2 has ramped up to about 23 mA when Q1 switches off, and the leakage inductance switches that 23 mA current to LED2.

The term “ramped up” is used because the LEDs represent a near constant voltage load at the output of T1. An inductor with a constant output voltage V and inductance L can only increase or decrease its current linearly (di/dt = V/L).
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
Correction: In "The current limit in the LED utility light", the second to last paragraph, the first "LED2" should read "LED1".
 

PeteHL

Joined Dec 17, 2014
473
Correction: In "The current limit in the LED utility light", the second to last paragraph, the first "LED2" should read "LED1".
In this forum, it is possible to edit your posts (at any time, I believe). Just click on "Edit" at the bottom left of the post.

Thanks for your interesting circuit,
Pete
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
In this forum, it is possible to edit your posts (at any time, I believe). Just click on "Edit" at the bottom left of the post.

Thanks for your interesting circuit,
Pete
There is no "Edit" button on my screen. This may be because I operate a Mac computer with OS 10.6.8 (older), or because the post is in the "Completed Projects" category.
 

DickCappels

Joined Aug 21, 2008
10,153
You will notice that I made the edit on your behalf.

I am pretty sure that you cannot edit your post because your post count is not high enough yet. I used Mac OSX 10.68 for years on this forum and the "EDIT" button was there all along.

By the way, finally being forced into upgrading to 10.13.2 was painful and expensive in terms of time lost, a printer/scanner that was not supported and having to get a later version of MS Office. You are wise to delay upgrading.
 

Audioguru

Joined Dec 20, 2007
11,248
You are using a Mac? HEE, HEE, haw, haw, ho, ho.
I saw a Mac about 25 years ago but it wasn't mine and I did not try using it.
 

Thread Starter

johnrohrer

Joined Feb 20, 2018
8
You are using a Mac? HEE, HEE, haw, haw, ho, ho.
I saw a Mac about 25 years ago but it wasn't mine and I did not try using it.
For the benefit of others reading these posts, I have used both Windows and Apple operating systems over a 30+ year span and still find the Apple operating system considerably easier and quicker to use, a good tool.
 
Top