30V from 24V

On Wed, 05 Nov 2014 08:32:59 -0500, JW Gave us:

I hear you thought once too. :-)

Must have been some time ago...

Or maybe it was just that spot you were standing in.

You thought... but it was only right there, in that "here" spot.

Can ya heeer me now? --Baba Ram Das (Dr. Richard Alpert) :-)

So, you'd monitor the current through the LED and shunt excess to keep LED drive at the target value. That wastes some energy, but at least there's no 700 mA sharp edge transient presented to the power supply. When the pulse is over, turn the shunt 'way up to dump the field in the coil, of course.

Over time and temperature, is voltage-drive or current-drive the best for this kind of thing? Or, does one want to sample the light output and try to regulate that?

Of course, if you did a dual-beam measurement, it would remove a lot of the constraints on your LED drive. Also, the extra SNR might be handy when the inevitable real-world problems arise in the field. For example when someone changes the hens feed and thus the optical properties of the eggs.

ChesterW

On Wed, 05 Nov 2014 14:30:49 -0600, ChesterW Gave us:

The bulbs sound like an LED array on a standard industry type base. Doesn't have to be Edison's but this is the serviceability aspect here.

I am SURE that base has a circuit in it which takes the 24V and makes the individual currents for each actual LED in the 'bulb' array.

It would seem cheaper to open that bulb up, and likely only require a single part adjustment to that 24V-in-LED-currents-out circuit, to make it accept 28V gracefully. (ooops 30V)

Slap a new, in-house P/N on it, and voila'! You even get to mark 'em up a bit more, and may even be able to get a contract Mfgr to perform entire runs of the changeovers rather cheaply.

The LED drive is a great deal simpler than the rest of the instrument, so I'm not making any compromises for its benefit. There's no extra SNR in a dual beam measurement unless you're dominated by the source noise, which I'm not planning to be.

Cheers

Phil Hobbs

It's a 20W Philips SMD mounted device, the LXS8-PW40. I need the relatively high spatial coherence of a small source in order to be able to get the light onto the egg from a distance.

These ones are just wired in series.

Thanks, if that were what I was doing, it wouldn't be a bad idea, except of course that I wouldn't be likely to get clean 1-2 kHz modulation out of a scheme like that.

Cheers

Phil Hobbs

On Wed, 05 Nov 2014 17:27:08 -0500, Phil Hobbs Gave us:

Costco sells these flashlights. 3 per pack, and they are extremely bright.

A very good value at like $15 for a 3 pack.

Anyway, the driver for them is in the base cap, and cycles its connection to the case tube of the thing to pulse it, and it is pulsed, as I get moire patterns at night on my bike from some street lamp types, and it changes when I change the setting...

Back to anyway... you could get one of those, and see how they fire their CREE single LED "photon source", and maybe adjust your design from what you learn.

They are bright enough to light the street from street lamp height.

They also have a brighter, C cell x 3 fired job that has a movable lens. You could buy a couple of those just for bench test illumination, etc. These are single element LED with a tiny 16 device array on one chip. Probably like yours.

The smaller flashlights are a single LED.

Well worth buying one of each, just for utility purposes. This is not spam.

Cool! (thanks) Can I ask if it's a pure (585nm?) led or a blue into a phosphor?

George H.

Just a basic poor mans blocking oscillator in buck boost form. Turn ratio to give you the needed 4-5 volts on top of the already 24 you have now. Actually, you could drive a buck boost transformer from a controlled current source. This way your 4-5 volt boost is current limited and frequency controlled. Two birds with one stone.

I am current picturing the perfect circuit in my head as I type, looks like a cake walk. Jamie

Den torsdag den 6. november 2014 00.13.13 UTC+1 skrev Maynard A. Philbrook Jr.:

all in one:

-Lasse

...

Phil,

are you sure that 30V is actually enough? These high power series LEDs can have differences in Vf piece-to-piece and over temperature.

Bin Code Minimum Forward Voltage (V) Maximum Forward Voltage (V) G

The variation over temperature is considerable, plus Tc is negative, so if your LED is pulse-driven and T happens to be low, you can probably add another volt of Vf.

Depending on the type (bin) of LED and temperature, that might just hit

Plus, the LED packs a lot of power into a tiny package. That's fine in itself, and is basically just what you intended, but it also likely means that the chips could have thermal time constants in the single ms region. The whole package has obviously much more inertia, but the series chips are tiny, essentially cut to the size of their optically active areas, and, being connected in series, they are likely sitting on some sort of an insulating layer or film. That together could give them fast thermal time constants. Possibly fast enough to be partially inside your intended frequency of modulation.

High power LED light output does not scale linearly with current if you take temperature changes into account. As T rises, the LED efficiency falls. At the power levels of the LED you intend to use the effect of self-heating on efficiency is definitely there. It might even show up at your modulation frequency, producing little thermal "tails".

It's possible to cheat by using a constant power drive instead of the usual constant current drive. The rise of current that will happen (in constant-power drive mode) when the LED heats up and Vf drops due to the rising temperature will to a degree compensate for the drop of the LED efficiency due to that same change in temperature. I don't know in how far these effects are physically related, but in some high power LEDs, constant power drive happens to be a passable approximation when constant light output is needed and no active control of light output is available. Unfortunately it needs voltage sense, current sense and something that approximates a multiplier in the control circuit. Of course, sensing and controlling the actual light output in real time would be the "exact" method, but you did not say how "clean" you need the modulation to be, so an approximation might also do...

In any case, the control system (whatever its working principle) will likely need another couple of volts to have a good linearity, so one might consider something on the order of 35V on the supply to ensure that there's enough left for the LED (over bins and temperature), the current sense, the driver (to maintain control and linearity), plus a little for power supply variations and stray losses in the wiring.

Regards Dimitrij

It's a phosphor type. I picked the one that had the smallest slope in the 600-nm region, and it works fine. As a consultant, I don't have a lot of opportunities to iterate stuff, because customers quite reasonably expect my stuff to work the first time. That does mean that I have to play pretty safe on stuff like power supplies.

Cheers

Phil Hobbs

Hi, Dimitrij,

Thanks, that's good wisdom. (It's good to have more than one set of eyes on a problem, which is one of the reasons I like SED.) The current iteration (LT3580 @ 900 kHz, 15 uH inductor, 1000 uF output cap) makes about 32 volts, because of that 29V maximum you mention, ripple, and various tolerances, e.g. 2% on the internal reference and 1% on the resistors.

Misbehaviour of the LED goes away to leading order because of the spectral differencing measurement. The funny power law correction degrades that to some degree, of course, but it isn't too awful. On reflection I may put a photodiode in the TX module, to allow correcting for that problem.

The control system is pretty standard--just an NPN transistor with a resistor in its emitter, controlled by an op amp to keep the current constant. With a beta of >100, that's more stable than most of the other parts of the system.

The competition's gizmos use flashlamps and PMTs, as I say, which take a lot more TLC than the white LEDs and PIN photodiodes that I'm using. (I've used flashlamps, and I'm a _huge_ fan of PMTs, but both are badly out of place in an egg-grading machine!)

Thanks again

Phil Hobbs

Thanks. A single-chip solution would be pretty attractive, I agree. The problem is how to get a really well-behaved 1-2 kHz modulation, which is fairly far from obvious with that part. I'd probably have to phase lock the oscillator, to begin with, and then figure out how to make it turn on and off cleanly at that rate. (If I just put a MOSFET in series with the LED, say, it would discombobulate the current control loop pretty badly, it looks like.)

Cheers

Phil Hobbs

LT3517 maybe? has the boost and a driver for a series mosfet for PWM

-Lasse

Actually, you can correct that.

An active zener circuit can clamp the current output to maintain the last set point or with in reason when load is removed from the LED driver. This of course means you can't be modulating your current levels as part of your process, otherwise it'll inhabit it.

Example. Comparator, (-) input monitors the output side of the current driver to the LED, the (+) input is and integrator. When your pulse removes the load from the LED driver, you'll get a slight up swing, not much, before the active clamp comes in. The comparator of course needs to drive a transistor rated to clamp the output to maintain near constant current.

You will need to tailor the inputs so the comparator output will allow the switcher output to raise it's voltage just a little, otherwise at startup, you may see a condition where I'll simply clamp always.

Then again, I guess you could simply have another LED of the same type as a synchronous switch to maintain constant current load. You turn on the LED that is being used in the test and turn off the ballast LED, and vise-versa, etc.

Jamie

Philbrook Jr.:

for

are

The

in

are

care

headache,

benign--the

have

controlled

looks

And running the regulator chip start-stop at 1 kHz with a switching frequency of 250 kHz gives ratty start and stop edges at best.

What is your minimum pulse width? 50 us, less? if it can go as low as 1 us a current diverter gives a reasonably flat current pulse. That would be faster than the thermal tc of the LED lamp.

?-)

Thanks, I was sitting here wondering if a phosphor would have a more stable spectrum vs temperature than an "pure" LED.

(Hey, has anyone else noticed that google fixed the extra line problem.)

George H.

The phosphor is inorganic, so the shape of the spectrum should be pretty stable. All the optics have to be able to be hosed down and cleaned with Lysol along with the rest of the system.

Nearly all the eggs are blood-free, so we can do a continuous online calibration of the spectral ratio--just chuck out the outermost 10% of the distribution, and you're pretty well guaranteed to get the right numbers. That takes account of smears on the glass, the occasional drop of yolk from a cracked egg, that sort of thing. (Those events will be rare, but could othewise cause the rejection of hundreds of eggs if the software is stupid.)

The industry is used to flashlamp-based systems, so the aging of a white LED is going to look like pretty good news, I expect. Today they have to run calibration once per shift to get the best performance.

Cheers

Phil Hobbs