Overdriving LEDs

I have been toying with some flash ideas to go with PICAXE flash controller I built a while back.
Air gap flash does keep comping up, but I came across something called a vela one.
https://www.vela.io/vela-one-high-speed-flash

now while relatively cheap (£787) and a fair amount less dangerous than the 20+kv of an air gap flash, I wondered how they got what they needed. they apparently use COB LEDs but they give no details on how its driven.

a picaxe @ 64mhz can get close with a 0.625us pulsout( length of flash could be varied too), which should be fast enough for just about everything but the fastest bullets. I have been thinking about using some CREE LEDs as they are extremely bright. I have a torch with 7 in and is rated for just 9000 lumen, something short of velas' claimed 1,000,000 lumens.
I have not looked into driving CREE LEDs as yet, but most have a very high ESD rating (often more than 2000v), so the current thinking is get hold of some and drop double rated voltage through them which should get pretty close.

Any thoughts? If the LEDs can just take power strait of the rail (negating drive issues with them), what will be needed to drive a high current load of a dozen or more LEDs for such a short period of time with the shortest possible rise and fall times - I know a large capacitor will be need on the rail to keep things going - maybe charge a large capacitor and drop the current through a power mosfet of some description. I am open to ideas, not just for creating the desired output but for features to try and include, figuring a charge/ready indicator would be a good start


from vela

After months of experimenting with different circuits and LEDs, we have built a circuit that drives nine LEDs up to 20 times brighter than rated, without damaging them or overheating

Click to expand...

It's not too hard to switch a few tens of amps within 50ns using power MOSFETs, it just needs a decent driver. I've built a few brushless motor controllers and found that without a decent gate driver the switching losses are high - you typically need to drive an amp or two into the gate of the FET to charge the gate capacitance up fast enough, and take the same sort of peak current out to turn it off quickly, too.

COB LEDs aren't that expensive, so it's probably worth playing around to see how hard you can over drive them. The limit will almost certainly be the chip bond wires, but my guess is that for a 500ns pulse you can exceed their fuse current rating by an order of magnitude or more, as very little heat will build up. The LEDs themselves will most probably take a heck of a lot more current than their peak rating, as this is a single, very short duration, pulse.

The only thing I'd look out for is counterfeit COB LEDs, as there are an awful lot of fakes and LEDs that have bad pixels around on the market. It looks as if the COB LEDs in that Vela flash unit are very similar to the ones used in outdoor floodlights. These seem readily available, it would be a matter of choosing ones with a forward voltage that would allow you to over run them a lot from your available supply voltage. My guess is that the Vf of the LED will be a heck of a lot higher when run at, say, ten times its max peak current, so some allowance for that would be needed. If you want to run this from batteries than you may well need to step the voltage up to the capacitor bank, LEDs and FET switches, to be able to get the peak current up to a high enough level.

Half of the "one trick pony" circuit is to turn the light off quickly, too.

If you don't, you'll get a trailing "smear" behind your bullet.

In another application, we used to do this with SCRs and a capacitor bank. We found that you had to add an SCR to the output (crowbar configuration) to kill the light quickly.

thanks tex, getting the thing to switch of quickly (fall time) was one of my concerns. I was thinking about a zener by its self. never heard of a crowbar circuit before now. A quick look round and found this

https://en.wikibooks.org/wiki/Practical_Electronics/Crowbar_circuit

if the ground from LEDs goes in (where the diagram shows 8v) as soon as the system is fired and raises above the zeners value SCR would be triggered.

Jeremy, I have a document called "Design and Application Guide For High Speed MOSFET Gate Drive Circuits" which I have only perused so far, however it show something called bipolar totem-pole driver. A lot of the circuits seem to be for PWM controlled power supply, speed enhancement is more of a concern on the off side of things, but with the off being taken care of by the SCR I am not sure that is going to be so much of an issue. Power the device as not been decided as yet. I have a high end gel motorcycle battery able to deliver well over 120 cranking amps (current it can supply to crank an engine over) and rated for around 36 amp hour @ 12v, although the battery's ability, or even a standard power supply's ability to keep up with sudden drains on power is poor by comparison to that of a capacitor. before o can even consider how it going to be power the LEDs will have to chosen and obtained so that where I will start my research I think

Crowbar circuit.

Usually used to keep a "runaway" power supply from putting too high a voltage on the output.

It can be used to quickly "short out" an LED to kill its light output. Proper design is important. Perhaps you'll find you don't need something this exotic.

I am having a read of the MOSFET driving document, you maybe right, there are claims of getting them to switch on/off in 50-200ps which is jolly fast.
There is a lot more than first meets the eye to driving MOSFETs correctly but the outcome should be worth it.

skimming through eBay, these aren't may not be the final choice but with a supply voltage of 12v it should be fairly easy to supply 2x voltage to them
http://www.ebay.co.uk/itm/Saving-Co2-MR16-LED-6W-COB-Chip-On-Board-Bulb-by-Future-Lighting-Ltd-60W-/181452088246?var=&hash=item2a3f6323b6:m:mEDdaBUoGifCYOigeVLLoNw

Matt Kane who developed the Velo One product seems to be quite forthcoming about the product, trials and tribulations on its kickstarter page and blog so it might be worth taking a look on there, as well as perusing Google to find other snippets of how he may have done it ...

https://www.kickstarter.com/projects/vela/vela-one-the-worlds-first-high-speed-led-flash

From what I can tell, though I may be wrong, the main thing is driving the LEDs well over their official current ratings. I imagine there was a good degree of experimentation and building on past experiences involved.

Turning the LEDs off is easy if you use a decent FET driver. It's all about getting the charge out of the gate in a few ns, and that means using a totem pole driver that can sink a few amps to discharge the 1000pF or so of each FET gate. I'd suggest using a dedicated gate driver chip, as these are easy to use, accept logic level inputs and can easily switch FETs on and off fast enough for this task.

Personally, I think the hard part is going to be current limiting. At the sort of turn on/off speeds you're looking at, resistors and wires are going to be inductive enough to slow the current rise time. The capacitors will need to be very close to the LEDs, perhaps with low value non-inductive resistors for current limiting.

I built a capacitive discharge tab welder a few years ago, and that operated at peak currents of a few hundred amps (close to 1000A at the start of the pulse). I learned a fair bit about the inductance of even short lengths of very heavy gauge wire doing that. The short (around 300mm long) lengths of arc welder cable, that I used to run from the capacitors to the electrodes, would jump about violently from the intense magnetic field created by the current discharge.

I'm certain that a fair bit of experimentation will be needed, and I'd be prepared to accumulate a fair pile of blown components before you get it right........................

found a little more info, the current is definitely where the main gain is (many time more was the term used), as per hippys post - although there is some over voltage too

highest lumen rated LEDs on mouser run at 54V and 2.6A. If we assume at least 4 time current with the same voltage that's 10.4 amp x number of LEDs, assume 9 LEDs and you got yourself 90amps.
however I think a good starting point maybe something more along the lines of a 1000-1500 lumens per LED.
keeping the voltage too low, the output drops off very quickly.
http://www.mouser.co.uk/ProductDetail/Cree-Inc/CXA1512-0000-000F0HN20E2/?qs=sGAEpiMZZMuCm2JlHBGefhlKgvu0rJfJswDAIIDnDmGLpLnXPXcmfw==
700mA and 18v. I have a 19v 7.1A laptop supply here which could be used to charge a large capacitor and the charge dumped into the LEDs. FET would need to be rated for at least 20 amps.
I do fear though that the current will have to be "driven" through LEDs as I don't think they will draw that much on there own accord - and with a 19v supply would only need a 1.5ohm resistor to get them into tolerance.

much thinking to do

I'd not worry about the LED current rating at all, as the max pulse current rating will probably be for a 20ms half sine, and you're looking at pulses that are around 1/40,000th of this duration, so the data sheet rating will be way off. The same goes for FETs, the current rating is not going to be the limiting factor at all, the Rdson is, so you may well find you end up using massively over-rated FETs, most probably in parallel, just to stop the FET Rdson limiting the LED current.

The light output of an LED with change in current is fairly non-linear, and you will be way outside the spec sheet figures, so will have no easy way to calculate the luminosity at the sort of peak current that you'll be pushing through the LEDs. I'm pretty sure this is going to need a great deal of experimentation, with no easy way to actually measure the light output, it will probably mean doing a lot of camera tests to see what you can get.

Should be fun trying, though!

Here's another link which might be useful, someone else embarking on the same voyage of discovery ...

https://petermobbs.wordpress.com/2015/02/06/experiments-with-led-based-flash-gun-for-high-speed-photography

I don't think you will need a higher voltage, large capacity battery for generating what the LEDs need, and probably safer and better off not having that. The Vela One uses 4 x AA batteries. Don't forget it's "oomph" you are after and a low current pumped into a capacitor bank over time gives a huge current sucked out in a short time; the total "oomph" is still the same; V x I x T ( or something like that ).

I am not clear on how you mean that the current will have to be driven through the LEDs, won't draw enough current on their own. AIUI, LEDs will suck all they can, would be heading towards the eternal disco in the sky if you put them across any significant high voltage, high capacity supply and kept supplying them current. The trick is to give them far more than they can take by limiting how long that situation is allowed for.

Hippy's right, the LEDs are just forward conducting diodes, so Vf, the forward voltage drop, is set by the chemistry (for white LEDs the VF is typically around 3V at normal forward currents) plus the resistivity of the LED itself. The resistivity means that Vf will increase with increasing current, so there will be a maximum current the LED can draw for a given voltage.

Because of the resistance component of the LED, the Vf at the very high pulse current that you're looking for may well be way higher than the spec sheet. Although the Vela unit runs from AA cells, and the mean current draw may well be pretty low, my guess (and it's only a guess) is that there may well be a boost converter to charge the discharge capacitor bank. The capacitor bank (and it will probably need to be a bank in parallel, not to get the capacity needed but, as for the FETs, to lower the ESR, the capacitor Equivalent Series Resistance). The capacitors will need to be low ESR types, with as high a ripple current rating as possible, as the ripple current rating is a reasonably good indicator of the high current discharge capability.

It might be an idea to start by characterising all the components in the discharge path (including the wiring) in order to get an idea of the total circuit resistance before you select any current limiting resistor, as you may well find that most of the peak current limiting is coming from the series combination of the capacitor bank ESR, the wiring, the LED resistivity and the FET Rdson. The initial peak LED current will be limited by the series inductance, and most of that will be in the capacitor bank, so it may well be an idea to look at using a combination of low ESR electrolytics, with a couple of smaller value low inductance capacitors across them, with a low inductance, to deliver the initial fast peak current needed.

IGBT semiconductor as a switch

Not to belabor the point, but speedlights have evolved for a reason.

Here's some things to add to your research. Be sure and read to the second heading: "And now the point".

LINK: http://www.scantips.com/speed2.html

oracacle said:

I have been toying with some flash ideas to go with PICAXE flash controller I built a while back.
Air gap flash does keep comping up, but I came across something called a vela one.
https://www.vela.io/vela-one-high-speed-flash

now while relatively cheap (£787) and a fair amount less dangerous than the 20+kv of an air gap flash, I wondered how they got what they needed. they apparently use COB LEDs but they give no details on how its driven.

a picaxe @ 64mhz can get close with a 0.625us pulsout( length of flash could be varied too), which should be fast enough for just about everything but the fastest bullets. I have been thinking about using some CREE LEDs as they are extremely bright. I have a torch with 7 in and is rated for just 9000 lumen, something short of velas' claimed 1,000,000 lumens.
I have not looked into driving CREE LEDs as yet, but most have a very high ESD rating (often more than 2000v), so the current thinking is get hold of some and drop double rated voltage through them which should get pretty close.

Any thoughts? If the LEDs can just take power strait of the rail (negating drive issues with them), what will be needed to drive a high current load of a dozen or more LEDs for such a short period of time with the shortest possible rise and fall times - I know a large capacitor will be need on the rail to keep things going - maybe charge a large capacitor and drop the current through a power mosfet of some description. I am open to ideas, not just for creating the desired output but for features to try and include, figuring a charge/ready indicator would be a good start


from vela

Click to expand...

LEDs and xenon flashtubes are about as far apart as galaxies and atoms. Once a xenon tube is triggered, the xenon ions will continue to conduct the discharge capacitor supply voltage, until it drops enough to extinguish the tube. A neon lamp acts in exactly the same fashion, requiring more voltage to turn it on, than to keep it on once "ignited".

To make an high speed xenon flash, you need to not only trigger it quickly, but extinguish it quickly.

An LED is another animal altogether. The wire bonds on the die will likely be the limiting factor. As a diode, they will take all of the current you can throw at them, until they don't. The limit is primarily thermal, although you'll see some severe output wavelength shift as the current goes up.

As a start, I'd get a bunch of LEDs, charge a very large capacitor (10,000 uF), to some voltage just a wee bit above the forward voltage drop of the LEDs, and then discharge the cap through the LED with a VERY short, VERY large wire (avoid inductance at all costs). You may lose LEDs, but sneak up on what they will tolerate as you gradually increase the cap charge voltage between each "shot". If you DON'T kill the LEDs, increase the capacitance until you do.

For a particular LED, you cannot get any brighter. It will have absorbed the maximum current that it is capable of. Later, you can get into things like snuffing the current at peak, rather than letting the LED do it.

@rq3 - I don't think that's going to work well as IF the LED is still intact the current will fall off slowly as the LED conducting voltage goes down - you need a fast turn on AND turn off... LEDs have a 'large' range of voltage conduction. No need to subject the LED to decaying current to get a possible peak current value... [BTW 70 years ago I was discharging a lot of capacitors through NE2s and an R4330 100ws xenon lamp... awesome!]

that link is gives a handy starting point, thanks hippy.
tex, a high end speed light is half the cost of the vela and less than a 1/4 the speed, we know faster is better - and why not for a project. I did look at IGBT, but I am unsure if they will have high enough switching speeds. I aiming to try and run picaxe @64mhz and pulsout 1 for 1/1,600,000 second light burst. its something that still needs a bit more research. maybe a bit of trial between each.

capacitors will most likely end up being standard flash capacitors, they will exceed the voltage requirements, low ESD ratings and are designed for pulse use. 10,000uF maybe a little excessive though

had a read of this thread, he seems to running 150v and 20-30amps through the 37v LEDs
http://www.glacialwanderer.com/forums2/viewtopic.php?f=8&t=2237

FWIW, I built a quenched xenon tube flash light around 30 years ago, to photograph milk droplets and splashes (I still have the photos around somewhere, they're B&W, shot on fine grain ASA400 film with my old East German Exacta 35mm). The trigger was a photo switch beam break and the flash gun was modified with a thyristor crowbar to quench the xenon tube. IIRC, I got down to flash durations of a µs or two, but had to push the film to something like 2000 ASA during development, even with the widest aperture I could use and still retain enough depth of field. Speed lights were unaffordable for me at the time, especially as I was only really playing to see what could be done without spending a lot of money.

Getting back on topic, I think your best approach is to start small. Buy a small COB LED, with a low Vf, plus some low ESR capacitors (those for switch mode supplies are good). Do not use a capacitor with a voltage rating much higher than you need, as the chances are that it will have a higher inductance, and inductance is something you'll want to minimise.

Forget about IGBTs, you're way below the voltage where they make sense over using a FET. A decent FET will have bond wires that are good for 80 to 100 A or more and an Rdson of just a mohm or two, and no IGBT can come close to FET performance below voltages of 100V or so. When building BLDC motor controllers I still use FETs up to about 120V, and only over that do IGBTs start to have a very slight advantage. IGBTs are also harder to control in some ways, especially compared to something like an IRFB3077 FET (typically 2.8 mohms Rdson, 75V Vds, 120A bond wire limit, 280A silicon limit, see here: http://www.infineon.com/dgdl/irfb3077pbf.pdf?fileId=5546d462533600a4015356153f0d1def ). The gate capacitance is also reasonable for a low Rdson FET, at 9400pF (still needs a LOT of gate drive, though!).

I'd be inclined to start small, and just experiment to see how the LEDs behave when pulsed with very high current. Should be a fair bit of fun, I think, as I doubt many people have explored operating LEDs like this.

I built such a device to do a similar photographic experiment like Jeremy did...but in my case it was a bouncing marble.

I made the mistake of using white LEDs. Why a mistake? because white LEDs utilize phosphors, and phosphors have latency. Dozens of milliseconds worth of latency.

From my work with CRTs, I know that different phosphors have different latencies. But you will never get instantaneous light turnoff when you interrupt the current like a Xenon flash..
You may want to use RGB LEDs instead.

Given the latency of phosphors, it might be an idea to start this project off by making something to measure the light output. Fast photodiodes, with a "daylight" spectrum are around. One of these coupled to a storage 'scope might be all that's needed to get a feel for the LED turn on and off times, plus the actual flash light pulse duration when you get that far. You might be able to get a Picaxe to measure the light pulse width, if you used a fast comparator between the photodiode and the Picaxe, to square up the light pulse. You could get a resolution of 625ns at 64 MHz clock speed, which should be OK. Testing this with an existing flash gun might be a good starting point, and would make an interesting project on its own. It would also open up the possibility of running the final unit in closed-loop mode for set up, where the sensor is used to measure the actual light pulse width and adjust the turn on and off times for any given setting to allow for them.

Interestingly, the Vela unit seems to use white LEDs, not RGB ones, so I'm guessing that the COB ones they are using may use phosphors with a lot less latency than you experienced, fernando_g. I'm pretty sure I have some fast photodiodes around somewhere, and I may have go at measuring their response time at the weekend, if I get time.

seems like a good excuse to get a scope that's meatier than the PCB picaxe PCB scope.
pretty sure I posted something similar before but might take the plunge.
http://www.ebay.co.uk/itm/Hantek-6022BE-PC-Based-USB-Digital-Storage-Oscilloscope-2-Channel-20MHz-48M-Sa-s-/360781230592?hash=item54003cb200:g:NcMAAMXQL99Sc1fc
says it can go down to 1ns sample rate. It would be nice to be able to get more accurate time readings on the picaxe camera controller. I know that's fast, to the point where the PSB scope struggles to keep up.
its funny how you think that project should shouldn't be too hard and then suddenly you need to buy new tools and equipment before you can even start.

Jeremy Harris said:

Interestingly, the Vela unit seems to use white LEDs, not RGB ones, so I'm guessing that the COB ones they are using may use phosphors with a lot less latency than you experienced, fernando_g. I'm pretty sure I have some fast photodiodes around somewhere, and I may have go at measuring their response time at the weekend, if I get time.

Click to expand...

When I first wrote the post, I could not come up with the proper technical word....it is not latency but persistence. Sorry for the confusion.

Back then, I measured them up with a photodiode in photocurrent mode, coupled with a fast opamp as a transimpedance amplifier. The turnoff response looked fairly similar to an RC discharge curve. Thus I assumed the value which was 63% of peak to be one time constant. Then by using the rule of thumb that after 4 time constants one could consider the time to fully turnoff something like 12 milliseconds or about 1/83 of a second. Not fast at all!

But that was back in 2003 using Luxeon 3-watt stars. I am sure that faster phosphors are available now.

I still managed to get a decent image of a bouncing marble and water droplets, triggered from a videocamera's vertical sync pulse (1/29.97 for NTSC).

We knew what you meant Fernando.
I have ordered some mosfets and drivers (IRF1407 and TC4420). not about capacitors as yet.

however in other news, my handtek 6022BE arrived, had a bit of play and seems as though it should be up for the task in hand. Tried some timing of the my high speed trigger, in its default mode (no delay, camera mirror up, fire flash on input) I am getting about 300us which conquers with the timer based on hippys 20x2 timer

output from software, 100us time divisions, 2v divisions, yellow line is input to controller, green line is output - I was surprised at the fall time of the output, it is after a opto isolator which maybe causing some capacitance I supose

fernando_g said:

But that was back in 2003 using Luxeon 3-watt stars. I am sure that faster phosphors are available now.

Click to expand...

humm, a possible obstacle with high-power LEDs designed for operation on mains is that the phosphor might deliberately be long persistence to reduce strobing in intended service.

Another thinks; the "capacity" of the phosphor is matched to the raw LED source nominal output. Overdriving the LED may produce lots of blue & UV light, but the phosphor may not necessarily give more balancing yellow light

Light colour isn't that much of an issue. White balance can be set manually in camera and tweaked afterwards if needs be.

Thought I would do some testing and timing today, and came across the first snag

test code:

there was nothing attached to the pin, but I got the same result if connected and LED and resistor for testing sake.
the region above 4 is actually very close to 500ns. 2v is about 1.5us which is still more than twice the claimed 0.625us for a pulsout 1 @ 64MHz.
I don't know if it the cheap scope that is making thing a little off or maybe some timing issues with the resonator. This test was done one a axe091 with a retro fitted 3 pin plug for the resonator (was an pin DIL cut down to 3 pins)

It looks very much as if something is linearly integrating the output pulse. At a guess, I'd say that there is a capacitance on the output being charged and discharged at a near-constant current from the Picaxe pin. I suspect that Picaxe pins are probably near-constant current source/sinks when operated at their maximum current.

I think that the 'scope probe may well be the problem. Better probes have a compensation trimmer capacitor built in, that you can trim to optimise the probe using the 'scope calibration pulse train output.

With nothing connected you should indeed get a straight line up/down on the scope trace, not a triangle.

Hi,

I suspect the slopes might be due just to the sampling period of the 'scope. The positive and negative gradients have exactly the same duration and slope, so I think the "display software" may be simply "joining up" sampling points using straight lines. The PICaxe output pins have significantly different pull-down and pull-up currents (by a factor of 2 or more), so gradients due purely to capacitive loading should be equally dissimilar.

With PICaxe interpreter delays, I think the PAUSE 4 should be completely unnecessary. Also, it's very important to appreciate that a "sampling" 'scope (or one using sampling principles) cannot work with "single shot" events. The "PCB Scope" (or the DPScope-SE in its "boxed" form) is particularly slow because it has almost no dedicated hardware support (using just standard PIC ADCs), but it is well and honestly documented. I'm not so sure about the Handtek. Also, there really is quite a lot of "technique" (i.e. experience) and understanding needed to make good, reliable measurements of the more "difficult" waveforms, using almost any 'scope.

Cheers, Alan.

The indicated sampling rate of 48 MHz in the screen shot (assuming that it's correct, and not indicating a false "over-sampling" rate) should be OK, as it meets the Nyquist criteria with room to spare. If the 48 MHz is really an over-sampling rate, then I'd agree, it's a sampling error caused by the 'scope not being able to handle a reasonably fast single event.

For those not familiar with sampling, then the Nyquist criteria specifies that the sampling rate must be at least twice the maximum frequency of any signal being measured, and ideally a lot faster than twice the maximum to get a reasonably good measurement.

A true 48 MHz sampling rate should be able to measure frequency components up to 24 MHz as a crude approximation, and in this case a single pulse with a width of 625ns should have 30 samples, and that should be more than enough to give a reasonable approximation of the output pulse. I strongly suspect that the true sampling rate is a heck of a lot lower than the 48 MHz shown on the screen shot, perhaps down around 5 MHz or less, which would account for this being a sampling artefact, rather than a true measurement.

That 48MHz does seem to contradict what is being shown. It looks to me, given what the cursor time measurement for the triangle says, like the actual sampling rate is 800KHz, a sample every 1.25uS.

I also think we are seeing a sample before and after the pulse when low, and one when it was high, those three points then simply joined to give the triangle. The lines look too straight to be anything else.

In the future I need to pay more attention to what I have thing set to - it was set to 2ms/div which gives 1MHz sampling, I then increased the sample rate afterwards to get a closer look. rooky mistake.
so I spent about an hour seeing what I could actually get. if I set the software to auto in only stores what is being displayed. I switch to single and set the trigger to the channel that is being sent the pulsout.
Then with a bit experimenting I found that fastest sample rate I could get any result from was 5us, 16mhz sample rate gives a square pulse about 1us long.
it sometimes take a few goes to get it to capture. I am going to check to see if there is a newer software version available, but it may have to wait until the sunday due to work.

Seems everything software wise is up to date.
I thought it maybe wise to look at other ways to get the reading I need, with an un ending loop for the pulsout I managed to get a 200ns trigger.
I had PWM running more for curiosity than anything else

even then the pulsout seems to be around the 1µs. It seems to be reporting the PWM as 8MHz which is correct.
I do find myself looking at the PWM signal and thinking if a single peak could be captured pulses shorter than 100ns could be used. This would be a hardware problem

Edit: Would a shorter pulse be possible by directly addressing the SFR

I've had a bit of play with the mosfet, driver and 4 standard red LEDs (2 series pairs in parallel).

the green trace is from the source side of the mosfet above 0.2ohm resistor for current testing. Yellow trace is the picaxe pin.
I put some standard electrolytic capacitors in for the time being, 3x330µf. current initially peaks around 3.2A and quickly drops to around 1.2 for the rest of the pulse if my maths serves me correctly (V/R=I = 0.64/0.2=3.2).
the no capacitor run was nearly as good with a peak current of 0.2 less.
both runs were done @ 5v and no resistors on the LEDs some code as before minus the PWM. I'm not sure if the 990µf of capacitors were not enough, or the fact that they were not low ESR which made there performance a little under whelming.

Does anyone have suggestions on next steps, capacitors and a photo diode for measuring the out put?

oracacle said:

I've had a bit of play with the mosfet, driver and 4 standard red LEDs (2 series pairs in parallel).

the green trace is from the source side of the mosfet above 0.2ohm resistor for current testing. Yellow trace is the picaxe pin.
I put some standard electrolytic capacitors in for the time being, 3x330µf. current initially peaks around 3.2A and quickly drops to around 1.2 for the rest of the pulse if my maths serves me correctly (V/R=I = 0.64/0.2=3.2).
the no capacitor run was nearly as good with a peak current of 0.2 less.
both runs were done @ 5v and no resistors on the LEDs some code as before minus the PWM. I'm not sure if the 990µf of capacitors were not enough, or the fact that they were not low ESR which made there performance a little under whelming.

Does anyone have suggestions on next steps, capacitors and a photo diode for measuring the out put?

Click to expand...

I'm not sure what I'm looking at here. Basically, you want to charge a large capacitor to some voltage well above the LEDs normal voltage rating, and then use the LED to short the terminals of the capacitor.

LEDs are diodes. They will conduct any current above their turn-on voltage, and will tolerate any current until they don't. The don't is when the LED semiconductor itself fuses, or the wirebonds melt. The issue is generally getting the heat out of the LED before the next "shot".

You can run a standard red LED at say 2.2 volts and 20 milliamps, or 2.2 volts and 100 amps. The 2.2 volts is INTRINSIC to the LED itself, the current depends ONLY on how far the supply is ABOVE the 2.2 volts the LED needs.

The supply can be at any voltage, IF it is applied for a short enough time. You can charge a 1000000000 farad capacitor to 2.2 volts, and your red LED will run happily for hours. Or you can charge the same capacitor to 100 volts, and instantly destroy your LED. The difference isn't the voltage across the LED (always 2.2 volts), but the CURRENT the EXCESS voltage can push through that 2.2 volt LED drop. Your job is to discharge the most CURRENT that the LED can tolerate without fusing its wirebonds or the LED die itself without exceeding thermal limits.

Low Current=forever=average light output
Excessive Current=Dead=no light output
Somewhere In Between Current=somewhere in between lifespan=much greater light output

It's far from true to say that Vf stays at the particular value in the data sheet over a range of forward current, as it simply doesn't. There are two components to Vf. The first is the voltage needed to make the LED conduct, which is set by the particular chemistry of the junction (typically around 3V or so for a white LED). In addition to this, there is the bulk resistance of the semiconductor material, together with the resistance of the bond wires etc. This has a very significant impact on Vf at high current, and is the reason a very much high voltage is needed to push tens of amps through an LED that has probably been rated for a few tens, or perhaps hundreds, of mA.

A look at the data sheet for any LED will show this bulk resistance element clearly, as Vf increasing with increasing If. Vf clearly continues to increase with increasing forward current beyond the data sheet limit, as if the LED were a low value resistance. Most of the time we never worry about this, as we run LEDs at low current and assume that the datasheet Vf is valid, which it is under those conditions. It's only when looking to drive very high current through an LED that the bulk resistance component of Vf needs to be taken into account.

The traces look very much as if there is capacitance available for both runs, most probably from the power supply, I think. The use of a resistor in the ground-source path will slow the FET down, as it will increase the FET gate voltage required (with respect to 0V) as the FET starts to conduct. Ultimately it could well turn the FET into a constant current driver.

I don't think the capacitors are fully charging.
I added a pause 2000 into the loop before going back to the top of the programme to try and give them a chance to charge, I also added a 1nf between the picaxe pin and ground to try and reduce the ringing, seems to have helped a little
the first shot is with the probe set to 1x and software set to 200mV, the second probe set to 10x and software 200mV.
Not too worried about the ringing on the output side.

were would you suggest I take a measurement from Jeremy?

If you can reference the gate drive directly to the source of the FET, then measure the current across a shunt either below that, or in the supply, then that should be better. Any resistance that's included in the source to 0V connection is going to have the effect of reducing the gate drive voltage at switch on, if the gate driver is referenced to the 0V side of that shunt.

My experience of driving power MOS FETs in brushless motor drives is that it can take a lot of very careful tweaking to get good turn on and off times (and turning FETs off cleanly is often harder than turning them on). My last BLDC drive board used 1uF, low inductance, polyester capacitors about 5mm from the FET supply side and source connections, together with a few hundred uF of low ESR electrolytics as close as I could get to the FETs. I still had some ringing, and ended up using a small SMD ceramic capacitor on the underside of the board and fitting a ferrite bead around the gate leg of the FET. Even a few mm of track or component lead can have enough reactance to cause some odd effects. IIRC, I used the TI gate driver ICs, that can source or sink a couple of amps or so of gate drive, as the FETs I was using had a couple of thousand pF of gate capacitance.

I used current sensing in the 0V to source line, but ran the gate drivers from small DC-DC converters to allow the gate drive to float. This allowed the gate drive to be directly connected to the FET source, together with all the local decoupling, and the current sense shunt was in the supply negative lead, so allowing the control circuitry to all be properly referenced to 0V. IIRC I used six paralleled low inductance shunts, around 50 mohms each I think - I can check later. I do remember having initial problems with the shunt resistors having too much inductance in one early version. It's a few years ago now, but I remember going through a lot of iterations before I got the FET gate drive nice and clean, with the biggest single improvement coming from floating the gate drivers by using the DC-DC converters (with a fairly large reservoir capacitor, to dump current into the gate quickly).

Currently I am testing on breadboard - not good for getting the thing tuned in I suppose.
I haven't chosen the capacitors as yet, would 20 m,ohms be considered low enough ESR. I think I need to get this sorted as I can think about a PCB that will have less inductance/resistance/capacitance than a breadboard
I will replace the current sense resistor for something smaller, currently have 3W 0.2ohm carbon film. I chose carbon film for lack of inductance.