Which LED Load simulator is best for precipitating led driver instability problems?

[treez]

Hello,
We need to test our 165W, switch mode LED driver (as well as our future led drivers) and do not wish to just use a resistive load of the same power and voltage level as the led load. The led driver uses a negative feedback loop to regulate the led current. (Max LED loading = 46.4V at 3.52A)

If we just used a resistive load, then we would not be able to truly investigate the propensity of the led driver to go unstable. Thus we need a dummy led load which has a low dynamic impedance and a fast frequency response, -in other words, it must mimick the led load’s dynamic properties.

Do you believe that a switch mode led driver, as described, would be more likely to go unstable when supplying LEDs (or a LED simulator) than if supplying a resistive load?

We believe that a led load (compared to a resistive load) has a higher frequency RC output pole, and thus there tends to be more danger of going unstable…Do you agree?

Attached is a diagram if the led driver driving a led load, and also two different types of led load simulators….which led load simulator do you believe would be best at mimicking the leds?
(LTspice simulation of led load simulators also attached)

 

[D.A.(Tony)Stewart]

Can you define all input/output parameters for the design?

...besides 46.4V at 3.52A

Then the test criteria for performance comparison will be obvious.

e.g. Recall my formula for ESR of a diode? temperature drift?
what about response time or junction capacitance?
ambient temp?
overshoot?

 

[treez]

yes thanks I understand your formula and we know we have about 0.6 ohm dynamic impedance in each led.
This isn't really design specific, we are seeking an imitation led load, for use with any of our switch mode led drivers, now and future ones.
But it is a buck converter, vin = 48v to 53v, leds = 8 by 12 long array, leds are 350ma nom but we run each at 440ma, as we have water cooling.
As long as we get something that's going to be better than a resistive load then we are happy...I am pretty sure that we can't mimick a led in every sense....we just want to mimick those things that will assure us of stability when we eventually attach the driver to a led load....so dynamic impedance and frequency response to mimick a led is important to us.

As discussed, a resistive dummy load is the worst as a led driver could well be stable into a resistive load, but unstable with the leds, due to the higher frequency load pole due to the small dynamic impedance of the leds.

Steve winder writes about this in his book, but in the free sample of his book, you cant see how much he goes into it.

 

[D.A.(Tony)Stewart]

Since you did not specify, Not sure why you need a simulator when you can use diodes and add Rs to make worst case.

 

[treez]

if we use diodes then it will need 165W worth of diodes....so that would be about 150 to220 diodes, each dissipating 1 watt, and a fan to blow over them.....even then, it would be difficult to set a particular voltage, and we'd spend ages selectively shorting some out to get the required voltage...the above circuit can be done so that adjusting the voltage is easy and quick.

 

[chuckey]

You may have to think about the forward characteristic,the reverse characteristic and the change of capacitance with voltage. The forward characteristic is the most important. Get a sample LED, carefully plot out its I/V curve. Relabelled the voltage axis times the number of LEDs in series. get hold of a perspex rule, and mark three best fit lines over the curve.
Work out the V/I of the first line, say its 200, then use a 200 ohm load. Where the second line bisects the first line, read of the voltage, work out the DV/DI of that line, and hence R, suppose its 100 Ohms, you need to put a zener in series with a 200 ohm resistor with a voltage equal to that you have noted. This technique can be used for as many straight line segments as you want. Finally put a forward conducting diode in series with this lot. This should be a reasonable load for positive voltages. Now connect a diode on this load to provide the reverse voltage characteristics with its own selection of zeners and resistors. A pulse extender should be used, so if you dodgy power supply produces violent negative ringing, the overloading of the reverse voltage can be noted, then measured.
The other parameters I can see that might need to be simulated are the LED capacitance both across the forward conduction and the reverse direction. I think this is esoteric. But if you want it you will have to measure it on an actual diode and figure out a way of simulating it.
Frank

 

[D.A.(Tony)Stewart]

if we use diodes then it will need 165W worth of diodes....so that would be about 150 to220 diodes, each dissipating 1 watt, and a fan to blow over them.....even.

Is this intended to be a universal lab design verification tool or a production tester with fixed parameters. What budget? And how many required?

one could use your product (LED board) as tester with adjustable Rs, to increase voltage drop for worst case.

If 1 LED is 600 mOhm worst case
the 8S12P array will be 12/8 * 600 mOhm=900mOhm. So if actual unit is less, make up difference with Rs or Nichrome Wire.

or buy single 150W LED @ 48V and water cool.

 

[dick_freebird]

I don't know that you're arbitrarily interested in inducing
instability. It could be, but I'd expect you're really more
interested in drawing out behaviors that relate to the
load as it's likely to be. That means roughly right I, V,
Z with maybe some deviations you can make (like, a
few diodes more or less, for Vout; added shunts for
excess current; more or less filter C; maybe a kicker
switched shunt for load-step.

You probably could get away with less heat sink if you set
up for repetitive low duty cycle pulsed mode testing.

 

[treez]

it will be a universal lab design verification tool.

 

[chuckey]

This is not possible unless you can simulate all possible LED types and their possible configurations and running current as well as some that will not be released to the market until next year.
Get hold of some typical LED strings, learn about them, try out your PSU, see what matters and concentrate on that.
Frank

 

[D.A.(Tony)Stewart]

It is very possible to program a non-linear dummy load to represent the VI characteristics for any LED or LED array in any voltage over a limited range and ESR but not the way proposed.

The degree of tolerance matching of the characteristic and the accuracy of the set-points must be defined explicitly. But it is possible to make one for 1 to 200 Watts or perhaps 10 to 1000 Watts up to any string voltage.

 

[D.A.(Tony)Stewart]

Furthermore, it is possible to chose inexpensive nonlinear components that don't require heatsinking, and use a precise regulation to get a quasi-linear loadline above threshold with low ESR that is programmable.