1
\$\begingroup\$

I'm some way into a project which involves switching 0.5W laser diodes at 100kHz, with a pulse of around 1us in width. I'm having issues with the diodes (which I don't have a datasheet for) failing after about a billion switching cycles.

Although I see a pretty well behaved waveform with my 60MHz scope, I'm thinking that the issue might be transients, possibly too fast to be seen on it.

The circuit I use is as follows (a bit simplified):

schematic

simulate this circuit – Schematic created using CircuitLab

To switch the diode on, the FET is switched off, obviously. The 10R resistor keeps about 2V across the laser in the off state - that may be superfluous, haven't tried without it. The voltage across the laser in the on state is about 5.5V.The current source is an LM317 with a 6R2 resistor.

Now I am remiss that there is no transient protection, but I am wondering how to do it. A diode to protect against negative voltage is a no-brainer. But what about a clamping diode against overvoltage? Zeners are not fast enough I think. But what about a something like this?

schematic

simulate this circuit

D3 and D4 would be Schottky I suppose. The idea is that the clamping voltage is always present, so the speed of clamping is only dependent on the diode, which should be very fast. It's a little bit wasteful of power and parts, but that is not really an issue here.

The leads from the driver PCB to the diode are a bit long (about 15cm), that might also be improved on a driver PCB redesign.

Thoughts? if anyone has direct experience of these circuits and real-world experience to offer, that would be much appreciated.

\$\endgroup\$
  • 1
    \$\begingroup\$ You might consider a gate resistor to slow down your switching times, so your current source doesn't have to react so fast to a changing load. I know this is not what you're asking, but if this is what's causing your transients it may eliminate them. \$\endgroup\$ – John Birckhead May 22 at 20:26
  • \$\begingroup\$ yes, in fact there is something like that in the circuit, and I might look at it again and try to slow the switching down a bit more. It also occurs to me that this circuit doesn't do much about switch-on transients. There are several levels of DC-DC converter supplying the +12V, the last one being a Traco power one. That might also be an issue. \$\endgroup\$ – danmcb May 22 at 20:29
  • 1
    \$\begingroup\$ I would take a second look at the AC output impedance of the LM317 as a current source, I suspect it's quite poor. \$\endgroup\$ – sstobbe May 22 at 22:19
  • 2
    \$\begingroup\$ That's only a few hours of runtime which is pretty poor. Definitely avoid overshoot on current and avoid parasitic inductance of long traces and reverse voltage . Overvoltage? your laser is a faster than zener Just avoid overcurrent in >10 ns range \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 23 at 0:23
  • \$\begingroup\$ So, you are using laser diodes that don't have a data sheet eh? Good luck. \$\endgroup\$ – Andy aka May 23 at 7:31
1
\$\begingroup\$

Use a GROUND plane.

And make the enclosed area of M1, R1, D1 about 0.5 cm on a side. In other words, you should get the inductance, through which the switched-current flows, down down down to just a few nanoHenries.

Given Vspike = L * dI/dT,

with 100 nanoHenries inductance (4", 100mm, of wire NOT over a plane) and 0.2amps switching in 5 nanoSeconds, the

Vspike = 100nH * 0.2amp / 5nsec = (the 'nano' cancel) == 4 volt spike.

Notice the spike-generator is M1/R1/D1. You can place the gate-driver some distance away, but use twisted-pair or coax-cable to connect to the FET's gate. And explore using source-termination (approximately 100 ohms at input to the TwPair, or 50 ohm input to coax). If you choose to look at the gate signal, use a 1pF or 2pF active-probe, with the probe's GND connected only 1cm away to the GND plane.

\$\endgroup\$
  • \$\begingroup\$ yes, this is helpful. Indeed we probably need to get the driver electronics much closer to the diode. Also we can slow the rise/fall times quite a bit, which will also help. Thank you. \$\endgroup\$ – danmcb May 23 at 10:16
0
\$\begingroup\$

Thanks to those who posted suggestions, and particularly analogsystemsrf. The solution turned out to be a redesign of the driver board. Instead of a MOSFET switch, I went to a discrete bipolar design, basically because it turned out to be easier to achieve a controlled ramp up and ramp down of laser current. The current source was also improved by using a discrete PNP transistor.

We also did some physical redesign to reduce lead length to less than 100mm.

With the redesign, the voltage/current waveform is very well behaved, with rise/fall times of about 100ns, and no detectable ringing when viewed with the scope. As the pulse width has a minimum value of around 500ns, this is quite OK.

I also added some TVS diodes at all input ports to give some protection against ESD from normal handling of the unit.

On a first test, we've been able to scan for some hours with no issues with the laser. Hopefully the problem is now solved.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.