Jeremys LED Lighting Page .pdf
Original filename: Jeremys LED Lighting Page.pdf
Author: Jeremy Welch
This PDF 1.4 document has been generated by Writer / OpenOffice.org 3.0, and has been sent on pdf-archive.com on 20/03/2011 at 15:21, from IP address 109.112.x.x.
The current document download page has been viewed 1506 times.
File size: 15.6 MB (13 pages).
Privacy: public file
Download original PDF file
Jeremy's LED Lights Page
What's in this document?
This document contains detailed electronic information and will hopefully be of use to both novices
and more experienced hobbyists and electronic DIYers. It assumes you are familiar with basic
electronic concepts such as Ohm's Law, and the operation of bipolar transistors, MOSFETS and of
course LEDs. If you are not, I suggest looking them up on wikipedia.
I have tried to gather together and interpret many threads of information that it has taken a long time
for me to comprehend, there being many pieces of information scattered about in the world with not
much help on how to use the principals described. I have included a level of detail on the use of
inductance for buck regulating circuits later on in this document that I have not found anywhere but
have had to work out for myself over the last few years.
What do I Need?
Construction of these ideas requires the use of soldering iron and other tools, such as pillar-drill for
accurate drilling of metal. Although I have put all information that I think is necessary I do not
guarantee you will be successful since only YOU can determine that.
Introduction to Power LED's
I have made LED lighting for various spaces to run from solar power. It is the highest efficiency
small scale lighting solution available at this time, thanks to the Z-LED P4, an LED illuminant
which boasts 100 lumens (A measure of total light output) for 1 watt of power. To give some some
scale to this, compare to a good 8W fluorescent strip at 460 lumens (57.5 lumen/watt) or a similarly
good 13W strip giving 1000 lumens (77 lumen/watt). Compact fluorescents are about
60lumen/watt. These LEDs are not cheap at €8 each, but lifetimes are in excess of 50,000 hours, so
replacement is not necessary on any usual time scale. Using them, at this stage, means a little bit of
work since I have not seen any commercial product with them in yet – they don't quite fit easily into
any of the standard fittings for halogen bulbs etc. because of their cooling requirements, though I
have seen some attempts, at nearly twice the price each, and probably using an inferior LED
judging by their ratings.
At full power their efficiency reduces but they still produce 240 lumen at 3.5W input power – 68
lumen/watt, the same as a normal (not special high-output) 13W strip, but the light is considerably
better than a strip, being dimmable, turnable, colourful, UV-free and with appropriate optics
focusable too. My impression is that they are very close in colour to sunlight, confirmed by playing
with my cameras white-balance settings, so do not be scared by the 'cool-white' label; and whereas a
single candle will suddenly “illuminate” a room full of strip lights, when lit by LED's the space is
already lit so the candle doesn't have the same effect. I tend to use them at about 70% since they are
still high efficiency and producing lots of light at that level. Turning them up further doesn't seem to
do very much either. At that level, 3LED's were drawing 0.51A from my caravans 12V lead-acid
battery and the entire caravan was lit well enough to read comfortably in the bed area or chop
vegetables in the kitchen, a task that four 8W strip lights drawing 2.7A had been unable to achieve.
We were using head torches to read and cook (the tubes are all fully functional though older 400
lumen rated, and the somewhat yellowed plastic diffusers removed) though this was partly due to
poor placement of the strip lights
Cooking by the light of a single 2W LED
Gentler lighting from 4W of up lighters
White LED's use between 3 and 3.8 volts each depending on how hard they're being pushed, and
this voltage will reduce as they get hot. For this reason all manufacturers of LED illuminants advise
the use of a constant current source. Other manufacturers have marketed various little devices to fill
the need, but they are all expensive and require over 2V overhead to work, for reasons unknown.
This means that they are unable to run 3 LED's at full power (3.7+3.7+3.7+2 = 13.1) from any leadacid battery I've ever used except while it's being charged – in other words during the day on a solar
12V system. Not very helpful.
A Single Candle
2x8W Strip lights
2 LEDs at 0.5W total
2 LEDs 2W total
2 LEDs 4W total
2 LED's 6W total
These 6 photos were taken with the camera set on manual exposure for
comparison (f2.8, ISO1600, 1/30th second)
Resistive Current Regulators
A one amp constant current source can be constructed for less than a euro that only requires 0.6
volts overhead, and occupies less space than a decent eraser. It contains just 4 components and can
be made dimmable by the addition of 3 more. The circuit is as follows for the strait 1A version:
LED3 Z-LED P4
LED2 Z-LED P4
LED1 Z-LED P4
T1 IRFU8729 or similar
Battery, 11.7 to 14.5 volt
R1 (0.6Ω) sets the current when matched to the 0.6 volt drop across the base of the BC546 (my
standard general purpose npn bipolar – cost 4p). R2 turns on the MOSFET (T1) until the current
produces 0.6 volts on the 0.6Ω resistor and T2 starts to turn on and stop T1 turning on any more.
Different currents can be achieved by changing R1, eg 1.2Ω gives ½ A output.
The dimmable version is as Follows:
LED3 Z-LED P4
LED2 Z-LED P4
LED1 Z-LED P4
T1 IRFU8729 or similar
Battery, 11.7 to 14.5 volt
In this version, R3 and T3 create a reference voltage 0.6V above the current sense voltage and Pot1
sweeps from that reference - turning current-limiting T2 on with no current flowing - to the current
sense voltage, meaning that the full 1A must flow before current limiting activates.
A lower overhead version would be:
LED3 Z-LED P4
LED2 Z-LED P4
LED1 Z-LED P4
T1 IRFU8729 or similar
Battery, 11.7 to 14.5 volt
Notice that R1 is now 0.3Ω, which would allow 2A to flow before turning on T2 but Pot1 now
cannot reach the actual sense voltage, R4 prevents T2base getting closer than half way to it, so it is
effectively dimmed to half, keeping it at 1A. The lowest full-power operating voltage is now just
11.4V. This could be taken further but it must be borne in mind that the junction voltage of a bipolar
transistor changes with temperature and that if, for example, a 0.1Ω R1 is used, dropping just
100mV at 1A, a 50mV change in the voltage drop of T2 would change the output by 50%.
The IRFU8729 N-Channel Enhancement HEXFET MOSFET is vastly over-specified for this job,
having an on-resistance of 0.006Ω a switching time of 25ns and a current handling capacity of 50A
but costs only 41p and does the job. Its in an IPAK package, a bit like a miniature TO220 but with
no fixing tab, and in this circuit needs heatsinking – up to 5W dissipation if the battery is being
overcharged at the time. I have been known to fit it between the dimmer pot and an aluminium wall
plate with a little CoolTape double sided sticky pad to stop it sliding out. 100cm² of cooling surface
should be enough in any case, but it may get hot.
Mounting Power LEDs
The Z-LED P4 similarly needs cooling, as it has to absorb 3.5 watts of power and only turns about
15% of this into light. Therefore a suitable fitting must be constructed. 200cm² each is about right as
these LED's need to stay cooler than the MOSFET at about 70°C maximum, and lifetime is cooling
dependent so it's best to stay cooler. CoolTape is not sticky enough to keep an LED stuck on its
own, so m3 screws or pop-rivets are needed to hold it on. Be creative with your design, as this is the
visible bit – but sometimes simplicity is easier to keep neat. I find a hand plane makes a nice edge
to an aluminium plate, and keeping the LED's shielded from sight is essential since this is a lot of
light to be coming from a 2mm square. Simply bending a 100cm² bit of aluminium into a square Ushape with the LED in the bottom, then screwing one side to the wall with a spacer makes a nice
invertible uplighter/downlighter for a small space.
Proper fixing of the LED's is essential to their longevity – here's how I do it.
First cut an appropriate piece of aluminium to a good size – I use a 10cm by 10cm (4x4 inch)
square for these. Bend it in a vice to the desired shape, then paint it if necessary, remembering to
mask off the area where the led is to be fitted so the heat doesn't all have to get through that small
area of paint. Use as paint shiny as possible inside, to avoid absorption by the shade, or bright
white. On the outside I have used black – dark colours absorb but also radiate better than light ones.
The LED is pre-mounted on a small aluminium
'star' that makes this surface mount device
handleable by humans.
As It Comes
NOTE: the transparent dome over the chip is
SOFT and finger pressure can break the gold
wires that are embedded into it, breaking the
led. Handle it by the star.
Place the star on the bare metal of your heatsink/shade and mark the
centres of where the two screws should go. The solder pads of the star
are very close so unless you have a supply of insulating washers I
suggest positioning them as far out as possible. I put the centres in line
with the very tips of the star.
I then drilled a 3mm hole for the M3 set screws. I sometimes drill
2.5mm and tap them with a thread tapper instead, as this can be neater in
Try to get all drilling done before the LED goes on, so make the holes for the wires and for the
fixing to the wall at this time. Then the LED goes on. I use CoolTape because I have it – I got a roll
in the mistaken belief that it would render the screwing obsolete, which it doesn't. The glue lets go
when it gets hot – the worst possible time – and things get broken. A possibly better alternative is
high melting point grease. You also get some purposeful thermal compounds for this too, the idea is
to fill the air gaps in the scratches in the aluminium surfaces so the heat gets through better – the
same as the oil in a frying pan.
Here is the fixed on LED:
Note the careful positioning of the screws so that they don't short
out the + and – terminals via the heatsink.
The 3 LED's can be on their own fittings or all together on one
larger one, but the wiring is the same – the battery + going to the
+ of the first, - to + between them, and the – of the last going to
the drain pin of the controlling MOSFET.
If your DC lighting system is other than 12V, work out the
corresponding number of LED's for the circuit. The IRFU8729 is
only rated for use up to 30V so find a higher voltage one if you
need it. Remember to think of the amount of heat it will generate
if your system voltage goes as high as the battery does while
Inductive “Buck” LED Controller
More complex, but more versatile is a class of regulator known as “buck” which uses that magical
and mysterious component known as the Inductor. An appropriate circuit for our requirements is as
+ C2 470u
+ C3 470u
+ C4 470u
11.5V to 16V
C6 + 10u
C1 2200uF +
The MOSFET is now handled by a MOSFET driver, a kind of digital buffer IC with very fast
response and very high output switching current. This one, the SC1301A made by SEMTECH, has
a 2A output that can swing the gate of our IRFU8729 from U1
12 volts or back again in just 20ns,
and responds 60ns after a change in its input. It has an enable pin (not shown) which is tied to the
+V supply in this circuit. Its input draws 5μA and has built-in Hysteresis, which this circuit relies on
to operate. The hysteresis means that the output turns on hard when the input reaches 1.8V and stays
on till it turns off hard when the input falls to 1.0V. C5 and C6 are in compliance with the
manufacturers instruction to have low impedance capacitors as close as possible to its supply pins,
to enable the 20ns 2A pulse to be drawn. It also should be close to the MOSFET gate, obviously.
So how does the regulator operate? The secret is in the Inductor. When you put a voltage across the
coil in an inductor, current starts to flow. Almost instantly, though, the magnetic effect of this
current affects the ferromagnetic core, and it starts to get magnetised in proportion to to current.
This increasing The result is that the (in our case negative) applied voltage is maintained while
current slowly increases at exactly that rate which generates the voltage that is applied (voltage is
generated in a coil by change in the magnetic field, requiring a changing current). When the driving
current is cut off, however, the core ceases to become more magnetised and the (negative) voltage
across it drops. It continues to drop and then go the other way as the magnetic field starts to reduce,
until the current that must continue to be flowing for that magnetisation to exist at all finds
somewhere to go – in this case via the Schottky barrier diode (SD1). The current then slowly
decreases at the rate that produces this opposite voltage in the coil, and the current output is still
flowing. This means that although 1A may have been flowing through the drive transistor while it
was on, it is only on for part of the time – so the current input over time is less than the current
output. That 1A also flows through the diode when the transistor is off, but this time it is flowing
round the short loop via SD1, R1, LEDs 1-2-3 and back to the Inductor without involving the
battery. Effectively 2 things are achieved – voltage drop without heat generation (the voltage drop
from the battery to the LEDs is across the inductor, not the MOSFET and the energy is being stored
rather than dissipated), and reduced current input.
T2 watches all this happening (current slowly increasing and decreasing) via R1. It compares the
voltage at R1 to the reference held by T3, that at its maximum is 1.2V below the +ve rail. At 1A,
VR1 is 0.4V. VR2 is therefore 1.2 - 0.4 - 0.6(VBE of T2) = 0.2V. Most of R2's current is then dropped
past the transistors nose by the effectively constant current of R3(across T2's 0.6V) leaving a little
bit that is passed through the transistor in a cascode arrangement to R4. Because R4 as a lot bigger
than R2 the voltage across it varies much more than does that across R1, and consequently R2 – a
ratio of 82:1 – and in the opposite direction to R1. The input hysteresis of the MOSFET driver then
means that the MOSFET is turned on when the current is less than 988.5mA and turned off when
more than 1012.5mA, a change that takes about 66μs when 6V is across the coil in either direction,
in the 4mH inductor that came from a commercial 2.5A 240Vac dimmer that I dismantled some
years ago to make a non-inductive dimmer before I learned how to do the inductive kind.
The Variable Resistor has the effect of reducing the reference voltage down to 800mV, at which
point the unit effectively shuts down because the voltage across R2 is insufficient to get current
down to R4 even when no current is flowing in R1. The switches, on one side, and the capacitors,
on the other side of the LED's allow individual switching of the three LED's and are not in the
controller but should be near the light fittings. The other two capacitors (C5 and C6) are to ensure
that the report of the current flowing in the inductor at R1 is “up-to-date” so the oscillator works
properly! The other diode (D1) shuts the circuit down when all three switches are closed (this
arrangement turns off each light by short-circuiting the LED) so that the controller doesn't waste
power maintaining a 1A flow through nothing but the current sensing resistor. R8 allows the circuit
to turn on again when the short-circuit is removed – but may be unnecessary.
The individual switching is optional (as is dimming) and switching globally, as for the resistive
current regulators above is acceptable. The individual switches require individual smoothing
capacitors for each switched element (i.e. if only 2 switches with 2 LEDs on one, only 2 capacitors
are needed) so that when one LED is short-circuited the excess voltage that would be stored in a
global capacitor does not have to be absorbed by the remaining LEDs.
I noticed in operation that this circuit does not dim all the way, but gets to a low setting and then
cuts out. This is because of the difference between the turn-on current and the turn-off current. The
turn-on current cannot be below zero, or nothing happens, so the turn-off current is always a set
value above this, and the actual output will end up in between at this minimum operating level. To
get it to dim all the way would need an independent oscillator and comparator circuit that more than
doubles the complexity.
Some of you will be interested in costs – this total circuit (minus LED's and recycled components)
cost me about £2.50. The parts that were paid for were the MOSFET(44p) and Controller(56p), the
1K variable resistor (92p), the 2 transistors(4p ea), resistors(1p ea), and the 2200uF capacitor(22p).
If you had to buy an inductor or a toroid core (might even be classed as a ferrite bead it's so small)
then add another 50p or so, and make sure it can operate at above 1A for this circuit – see
information below on Inductors.
The MOSFET and the IC are both static sensitive and you should be earthed at all times until the
circuit is complete. The finished article is not static sensitive. I use a piece of wire stuck into the
ground at one end and tucked into my underpants (at the side) at the other.
If you should try to construct this circuit be advised: the IC is very small. Its in a SOT-23/5 surface
mount package that is just 3mm long and 1mm wide and has 3 legs down one side and 2 on the
other! I have, however succeeded in soldering several of these to standard strip-board in the
Cut 2 tracks at a hole, with a knife, to just wider than the hole. Solder a piece of tinned copper wire
to the track above passing from this side one hole to one side (soldered) to the other side coming
back through immediately above the broken tracks and then passing back through again through the
hole of the upper cut track. Place the chip with tweasers so that all 5 legs are touching the
appropriate track/wire, then apply heat and solder to the track end of the bit of wire connecting the
upper central leg. The heat and a little solder should flow down the wire and the leg will be
attached. It won't look like much, and don't test it yet since you will probably break the connection.
Solder the other 4 legs by applying heat and solder a little distance away from the chip (next hole
along) – if your board is clean it will flow to the chip and solder it. Before you say “my soldering
iron isn't up to that” take a look at the state of mine! Its stuck into a cheap 40W non-thermostatic
soldering iron that came with something that looks like a spade sticking out of it.
My Soldering Iron – the result of abuse and the flux in lead-free solder
Choosing an inductor:
The key component in this circuit is the inductor, and not any old coil will do. It must have
sufficient inductance, and it must have this inductance with the current that we need running
through it. Not every core is capable of both these things, and some detail of the functioning of an
inductor is needed in order to know that you have a component that will work. Core materials are as
varied as transistors but some basic general info is helpful. 2 types of material are used for cores,
those based on metallic iron and those based on ceramics known collectively as ferrite. The two
most interesting features of any core material are its μ, or permeability (how many times more field
than air it generates for a certain number of current-turns) and its saturation field strength. Iron has a
basic μ of 2200 unless it is gapped, and saturates at 1.8 Tesla Ferrites have μ anywhere from 10 to
50000 and saturate at about 0.3 to 0.4 Tesla Ferrites work best at high frequency because they are
basically non-conductive, whereas a solid iron core would act as a fat short-circuit winding to any
coil wrapped round it. At mains frequencies (50 – 60 Hz) Cutting the iron into a thin strip helps,
because it then sees only 1/nth of the field per slice and so sees only 1/nth of the volts per turn
applied to the coil. Pulverising the iron to dust and setting in plastic continues this further. An air
gap in a core reduces the μ, requiring more turns to get the same inductance but since the
inductance is proportional to turns squared and μ and the field generated is only proportional to the
number of turns and μ, this provides a way of increasing the saturation current of an inductor
without loosing inductance. Coincidentally, at also improves the losses of a core by reducing the
volts per turn of the coil. Iron Dust cores are made with Distributed Air Gap to get the low values of
μ needed for an inductor to work at medium frequency and high current. They are usually yellow on
the outside, or colour coded. Ferrites Are usually dark grey and metallic looking, unless they are
coated. The Inductor I used for this circuit had a yellow toroidal core (I don't even think about other
shapes, they all leak) of about 10mm depth, 39mm path length and the ring about 5mm from inside
to outside. I measured its inductance directly (the time taken to get to 1A current with 1V applied =
inductance in Henrys, in this case 4ms), counted the turns (100), and from all this deduced the μr to
be about 250 and the saturation current to be about 2.5A – just what the dimmer it came from was
rated at. Yahay! Then I chose the ratio of R2 to R4 for a frequency of about 30kHz (by setting the
current difference from off to on), and matched R3 to get the 0.2V across R2 with 1.4V on R4.
I used 3 equations to do the deducing:
and finally, to convert from Tesla (B) to Webers (Φ):
l = magnetic path length (meters)
a = magnetic path cross section area (meters2)
I = current (Amps)
N = number of turns
μ = Absolute permeability of the material (vacuum = 10-7)
V = EMF (Volts)
Φ = total magnetic flux in the core (Webers)
B = magnetic flux density in the core (Tesla)
t = Time (seconds)
Δ (delta) means 'Change In' whatever comes after it.
In working out the saturation current of the unknown inductor I used the second equation with the
third substituted into it and the results from the inductance test:
which translates as:
B reached =
V applied⋅t taken
N turns⋅a core
The tester I use applies a constant 1V via a current sensing (1.2Ω) resistor and stops at 1A,
measuring the time taken as voltage increase on a capacitor fed while the inductor is being run by a
constant current. It reads 1V/ms (1mV/μs) and is probably accurate to about 10%, which is close
enough. The coil has 100 turns and the core has a cross section of 50mm². It took 4ms to reach 1A
with 1V applied. This gave a result of 0.8 Tesla The core colour (yellow) indicated an iron based
core with therefore saturation at 1.8 Tesla, meaning this inductor is safe up to 2A with its current
The three become one to give an equation for μ:
V applied⋅t for⋅l magnetic
4 ⋅I reached⋅a magnetic⋅N turns2
This absolute permeability can be turned to a relative permeability by dividing by the permeability
of free space (μ0) which is 10-7. Watch out though with spec sheets, I was very confused by one core
manufacturer (Epcos) who use a μ0 which includes the 4π from the equation above! Poking about on
the internet I find that this is not uncommon, μ0 being 1.26x10-6 and the 4π being omitted from the
equations. But the effect is the same in the end.
Bodging Inductor Cores:
It is possible to bodge a pretty decent core for an inductive drop or boost circuit out of an old tapewound toroidal mains transformer core, or if need be, several can come from one. The metal used in
these is a long strip of about 0.33 mm thickness, slightly coated for insulation, of silicon alloyed
iron which has a lower conductivity than normal iron. It is not normally considered for high
frequency use because of eddy current losses in the metal, but with a bit of thought it can be got
around. These cores normally would have a primary of over 1000 turns at 240V AC, 50Hz (in
Europe) but for some unknown reason frequency is thought to be the problem with thick layered
cores like these. Actually what it comes down to is the reason people want to use high frequencies
for transformers – the reduction of the number of turns. What generates heat in a strip-wound core
is not high frequency, but high Volts per Turn. With a little thought its easy to see why. The reason
for using strips in the first place is to divide the flux into thin strips. Imagine the core as solid – the
entire flux is within the outer surface of the core, so that outer surface sees as many volts as every
other turn on that core, except that this one is a short circuit. If you divide the core into ten equal
layers, then each sees only 0.1 of a turn. The length of metal it flows through now is approximately
half that of the original but is ten times wider (in a spiral) and ten times thinner so its resistance is ½
that of the whole core, but the effect of the lower voltage is greater since Power =
We cannot make the metal thinner, but we can still reduce the voltage seen by the thin metal layers
by increasing the number of turns on the inductor. An iron core with 1600 turns will have a very
high inductance, enough to last the whole of the 10ms half-cycle with a whole 240V across it
without saturating, but it has almost no current capacity so is useless in its original state. Seeing that
it has about 6 turns per volt is a good indication of where we want to head. For a core working at
12V for some of the time, 72 turns would be good to head for. To get some current capacity with
that many turns we need a low μ core, but iron has a μ over 2000. We get this by making an air gap
in the core with a hacksaw, 80% through (so it doesn't fall apart). How do air gaps work? Lets take
another look at the first equation above. The left hand side is a simplification. It should read:
The fancy thing on the left is a Greek capital sigma and in mathematics means Sum. The next bit is
the part that's being summed and the ·dl after it means “with respect to changes in l”. Put them
together and it says “Take all the length sections of the magnetic circuit and add up B·l/μ for them
all and this altogether equals 4π·I·N”. B (flux density) is pretty much constant round the ring, so its
l/μ that makes the difference. For the iron part of a (lets say) 200mm length magnetic circuit μabs =
2200x10-7. For the 1mm that we cut with the saw (once the 20% that remains is saturated) μ = μ0 =
1x10-7. The magnetising force requirement of the 200mm of iron is divided by a number 2200 times
bigger than that of the 1mm of air meaning that the air takes 11 times more magnetising force to
reach a particular flux density than the whole rest of the core. The total energising force requirement
for the whole core is therefore 12 times what it was and the effective μ of the whole is now 183⅓.
But what about the 20% of iron that remains in the gap? Well, what we have just created is called a
“Step-Gap Swing Choke”, a kind of inductor who's inductance is different at different current
levels. Initially it will respond with a μ of 2200 until that 20% remaining iron by the gap is saturated
and the magnetic force must pass through the empty space. The current level at which this happens
is much lower than 20% of the full current possible with the core, it is 20% of the full current
possible with the ungapped core, which is 1/12th of that possible in the fully gapped core. The other
80% will then saturate at a further 80% of what the fully gapped core would be (theoretically)
capable of. We can then wind more turns, get more current capability and/or more inductance
(because inductance increases as the square of turns and magnetising force increases as turns) and
with more turns reduce the number of volts per turn and induce less eddy voltage in the strips of the
core, so reducing heat generation in the core, though at the sacrifice of DC resistance. And nowhere
is frequency directly part of it.
Example: a 150W toroidal transformer core is converted for use in 4 buck converters from 24V to
12V by cutting the tape into 4 equal length pieces. One is then wound and is discovered to saturate
at 5A with just 6 turns on it – 2V per turn, at lower currents generating about 12W of heat. A 1mm
slit is cut through the 15 layers and an aligning card inserted, the core is now rewound with 48 turns
of the same (2.5mm² stranded) wire and now is found to saturate at 18A and at 0.25V/turn
generates no discernible heat (in theory reduced by 64 times to 0.18W). The inductance also
increases by 5 times. It is now reworked to serve at 180V to 12V and is generating 2.5W heat. It is
rewound with 100 turns and another gap cut, maintaining the current capacity (the second gap only
halves the effective μ) and the heat loss is quartered back down to 0.6W, further lengthening of the
winding deemed too far due to DC resistance and the need to operate at up to 15A continuously (for
some of the time) when the windings would be generating 15W of its own. Remember that the
winding losses are current² related so only apply while the device is at maximum power whereas the
core losses are (V/turn)² related and so apply all the time if the input/output voltages are constant.
This is well beyond the scope of LED lighting and moving into the realms of DC power
distribution, solar energy and inverters that will be another topic for another page.
One final thing to note is that the most important factor of the capacitors at the input of the
inductive LED driver is not the capacitance, but its ripple current capability. In this case, with a 1A
output the maximum ripple it will experience is 0.5A, when the output voltage is half the input
voltage. The equations for calculating RMS (root-mean-square) ripple currents in capacitors will be
Since I bought them 6 months ago, the IRFU8729 seems to have gone out of production and been
replaced by the IRFU8721, which is faster, lower resistance, lower gate charge and cheaper. I'm not
I hope this is (more than!) enough info to get you started with your own LED lighting. If there is
more info needed let me know and I will add it in.