Strange photodiode and VCSEL phenomena

Hi, Chris,

VCSELs are evil. What's likely happening is that the diode has multiple lasing modes, and the modes have different emission angles, and the slightest changes (temperature, drive level, back reflections from the world) make the modes jump around, with the highest-gain mode-of-the-moment being dominant.

It's hard to get flat pulses from a VCSEL. As you change temperature or drive or just wait a few 10s of ns, modes will change and pulse tops will squirm all over the place. Most of my experience is with fiber-coupled vcsels, but I'd imagine that you are spraying erratic modes out at various angles into free space. If a pd intercepts most of the angular output, things will average a lot better than if you only grab a small angle.

The UHS gates, driving a typical vcsel, can produce optical risetimes of 100 ps or better. A smaller resistance or a divider might help a little, depending on diode capacitance, but optical risetime is usually a lot faster than electrical drive. Can you measure laser diode capacitance, at zero bias? I'd be interested.

A good UHS triple-buffer can put 4.5 volts into 50 ohms in, as I measure it, about 570 ps.

I recently purchased a bunch of Lasermate parts that are superb. They make clean, flat fiber-coupled pulses with very little noise and nary a hint of moding; they measure just under 2 pF. I also have a bunch of brand "O" that are ghastly. In addition to noisy pulses and mode jumps, pulse risetime varies as a function of rep-rate, with risetime getting insanely slow at *low* pulse rates... hundreds of ns Tr at 1 KHz prf! You might try sweeping the rep-rate and seeing if anything strange happens.

I like the concept of a "movie" that would show spatial emission patterns vs time, but that would be difficult (understatement!) to instrument.

It's hard to measure the risetimes of the 850 nm lasers, as the silicon detectors are slow and the InGaAs detectors are nearly blind at this wavelength.

What does the Thorlabs diode cost?

John

Yeah, that's more apparent now after I took a look at the beam mode while varying the drive level over a fairly wide range.

Yes, next I looked at two things: what happens to the signal for a fixed beam region while changing drive level, which quickly elucidated the fact that it isn't the PD at all, but the VCSEL. Mode hopping certainly explains things. Then I set up a lens and put all the light onto my 10MHz PD, which then saw a much flatter pulse, until I pushed the drive level up to where I think the thermal thing kicks in.

What do you think of the hypothesis that the high drive level exponential decay after the rising edge is due to heat working its way through the die?

Interesting. I hadn't thought about that until now, that the I vs. V characteristic would of course like any diode, result in a faster current risetime than the voltage across the diode+R circuit.

I wish I had a faster scope :-(

There are some folks with 1GHz ones. Maybe I should borrow one. But I'm kind of off on a tangent here. The main goal was to create clean&flat square optical pulses so I could test the response of a custom photodiode amplifer.

That's really wierd. I will check out Lasermate. The Thor VCSELs are about $22 or so. I don't like to make a habit of buying photonics from them since I'd rather know the original manufacturer. But these were obtained on an impulse, just to have some to try.

The Firecomms 665nm VCSELs are even more idiosyncratic.

There are certainly amplified photoreceivers with GHz bandwidths. NewFocus comes to mind for instrument-type units, even some DC coupled.

Thanks for the input.

Good day!