new design of switching laser diode driver, need your advices

[shrad]

hello there!

this is my first real contribution to this forum, so this is a milestone on my road here

in my laser diode driver designs, I have slowly been attracted to swithed mode current sources

I draw several designs on paper to have clearer ideas of what could be done, but everytime the fact that spikes and EMI were unavoidable came in my mind and I have never tried to achieve a decent prototype

but lately I have found about new low-noise IC designs from linear technology

I quickly draw up a schematic under ltspice and had these results:

schematic


plot of current, voltage and coltrol loop voltage


so it comes that that design has in fact no noise at all and seems very stable in theory

but here's the problem: 30mA overshoot in the first 30 milliseconds

the curves indicate the coltrol loop is some kind of PID, and they result in handpicked values for R1, C4 and C5 for approximate best result

so do anyone, especially electronic gurus, have any clue on how to reduce overshoot but maintain good final value?

is this overshoot really nasty for laser diodes?

if this proves to be a valid design, the next step is to design analog modulation by replacing R3/R4 with an active equivalent (with a foot trimmer to set standby current), and then provide a prototype to a friend who has 5Ghz spectrum analyser and all the stuff needed to spot unvisible bad spikes

thank you all for your interest

[mixedgas]

thank you all for your interest [/QUOTE]

what is what color on the graph? I cant read the blue label.

The luxeon has a huge die compared to the LD, its frequency response is 1/200 that of the diode, typically 200 nanoseconds. Its slow. It has a huge mass and can take a hit.

You need a better model for the LD. A typical diode can do 1 ghz small signal. There are derating curves on the laser diode data sheets for pulsed power and average power. Integrate the area under the curve by hand, just divide it into a bunch of rectangles and figure out their area, and look at it vs time. See if you exceed the pulse specification. The diode is a small spec of semiconductor, much smaller then the LED.

If you reverse bias the die more then say 2.25 volts, they go poof , so watch the ringing. you'll need a shorting relay across the diode for your first couple of tries, to let it get to constant current first.

How about a dedicated laser diode driver chip?
Steve

[shrad]

hey mixedgas thanks for your quick answer!

the blue trace is the Vc signal (with R1/C4/C5) which represents the PID action (I think)

I was thinking about shorting the diode for the firts 100ms to let the current settle correctly even for slow components, but I wonder if the PID won't have to settle again and throw an additional overshoot

for dedicated laser diode driver chips, I had a look, but they usually come in awful bga or even smaller packages, making them really difficult to hand-solder, and this one is SO-16... plus, they are usually designed for low power telecom diodes

I made this choice because it is adaptable to most of the switching regulators from LT, and because this serie of low-noise regulators also come in higher current switching capabilities (this one is rated at 700mA max)

it must be possible to add multiple parallel mosfets for switching higer currents, though (30A easily achievable with a fast circuitry and 2 mosfets like IRF450)