Typical capacitance of a bigger laser diode?

[Joerg]

Folks,

Looked through numerous datasheets in the 100-200mW power range from
Sanyo, Sony, OSI, Ondax and so on ... nada. My experience when calling
is that there is no further info available.

So, does anyone know what the typical capacitance of an IR laser diode
in that power range is? A SPICE model would be even better. Reason I ask
is that we have to pulse-modulate the heck out of one, pretty much
redline it in terms of what it can give.

 

[Jamie]

Folks,

Looked through numerous datasheets in the 100-200mW power range from
Sanyo, Sony, OSI, Ondax and so on ... nada. My experience when calling
is that there is no further info available.

So, does anyone know what the typical capacitance of an IR laser diode
in that power range is? A SPICE model would be even better. Reason I ask
is that we have to pulse-modulate the heck out of one, pretty much
redline it in terms of what it can give.

If you know the "I" spec's, why can't you drive it with a current
source? You modulate that source instead.

I mean, most lasers have to be current regulated to some degree
anyway, I would think if you have the current source at its max rating
of the LD, capacitance wouldn't matter because you wouldn't have much
control over that unless you plan on doing some sort of Peak and hold in
which case, the peak condition will most likely vary with the diode.

Unless we're talking about something else?

Jamie

 

[Tim Williams]

I mean, most lasers have to be current regulated to some degree anyway,
I would think if you have the current source at its max rating
of the LD, capacitance wouldn't matter because you wouldn't have much
control over that unless you plan on doing some sort of Peak and hold in
which case, the peak condition will most likely vary with the diode.

There's no such thing as a current (or voltage) source at RF. The junction
has some resistance, even if it's just the (small, but certainly not
negligible) ideal diode resistance. Reality adds parasitic junction
resistance, wire bond and lead inductance, junction capacitance, etc.

At DC (or approximations thereof, i.e. up to frequencies where reactives can
be ignored for the circuit), one can build a current source, and simply not
care what voltage the load generates.

At RF, everything drives transmission lines and the speed of light applies,
so it's impossible to create a true current source. A "current source"
driving a TL generates a known voltage, dependent on the line's impedance.
When the energy reaches the load, the same voltage (in the line) divides
between the line and load impedances: to a first approximation, the line
becomes the source driving the load, not the source proper. It takes
several cycles of reflections, back and forth along the line, before source
and load are in agreement about what voltage and current stabilize at.
Therefore, it's much easier to simply match source and load, so that
although you need to know the RLC characteristics of each in order to do so
(matching networks and whatnot), you can simply transmit power into the
thing, and it gets modulated accordingly (in this case).

Tim

 

[Joerg]

'Scratch, scratch', I have no idea. I've got some slightly used ~100
mW Sanyo's in the lab, I could try and measure something. (or put one
in the mail...?) These are out of production so... any data may not be
that useful.

Only if it's easy to measure. Otherwise I'll contact some university
folks who deal with laser stuff a lot.

The capacitance a zero volts? Or when it's running at some current?

Essentially from close to zero current (which is usually at several
hundred mA) to full rated continuous current where they typically sit
around 2V forward.

You want to modulate it at some high frequency?

Yup, AM. Got to figure how much muscle the driver will need.

These have a relaxation oscillation up near 6 GHz, so capcacitance may
not tell you all you need to know.

That would not be cool but capacitance should not have anything to do
with oscillatory behavior.

 

[Joerg]

We're working with some butterfly-packaged parts in that sort of power
range, peak currents roughly half an amp. Capacitances are awful, 200
to 700 pF, so they are hard to drive fast. We're just now testing a
new gaasfet driver circuit, target being a 1 amp pulse around 200 ps
wide.

That's like LEDs, I was almost afraid it would be that high as well. I
took a look at abusing LM5113 chips but their pull-up is weak and kind
of sluggish. They are great when pulling down but I need both directions.

Some laser suppliers are Eagleyard, Innovative Photonics, Lumics.

I don't know of any Spice models. We make our own measurments of
capacitances, inductance, diode curves, and optical outputs.

Helmut Sennewald has contributed models for smaller versions on the
LTSpice Yahoo group. Just not for big ones.

 

[Joerg]

An ohm or two, resistive. The capacitance isn't usually an issue below
100 MHz or so.

The resistive part is fine but unfortunately it'll have to be more in
the gigeehoitz range :-(

 

[Joerg]

If you know the "I" spec's, why can't you drive it with a current
source? You modulate that source instead.

Current steering is the usual method but it won't be quite that easy in
this case. I still need to muscle capacitance around using some sort of
"gooser" circuit on top.

I mean, most lasers have to be current regulated to some degree
anyway, I would think if you have the current source at its max rating
of the LD, capacitance wouldn't matter because you wouldn't have much
control over that unless you plan on doing some sort of Peak and hold in
which case, the peak condition will most likely vary with the diode.

Unless we're talking about something else?

It's fast and furious AM I have to do, almost 100% swing.

 

[Boris Mohar]

Current steering is the usual method but it won't be quite that easy in
this case. I still need to muscle capacitance around using some sort of
"gooser" circuit on top.



It's fast and furious AM I have to do, almost 100% swing.

How about optical modulator?

 

[Joerg]

Then you'll have to worry about the laser rate equations as well as the
capacitance, and you can't readily fix those with a T-coil. The
relaxation frequency is often about 1.5 GHz.

What's the model number?

Don't know yet, just that it'll be in the >100mW class and IR. If we can
get a nanosecond of transition time that'll be ok for the beginning. I
could muscle the capacitance with a gooser circuit but that's always a
bit scary because laser diode can commit optical suicide in less than
the blink of an eye. All it takes is they come out with a less
capacitive model and ... poof.

 

[Joerg]

How about optical modulator?

Too many bucks

 

[Joerg]

Do people limit voltage swing when modulating these fellows, i.e. 0 <
Voff < Vf?

That would make driving a bit faster.

Oh yeah, otherwise the driver will cause global warming and could get
carbon-taxed

You don't even have to go to zero volts.

 

[Joerg]

Hi Joerg, Well I think this may be mostly meaningless for your
diode. But I ran a pulse into a Sanyo DL-7140-201S. (this is a 70mW
diode with typical I forward of 100mA) Signal terminated into 50 ohms
and then 1 kohm into diode. With no diode (and just x10 scope probe)
I had a 1/e RC time of about 20ns. (16pf probe so seems reasonable.)
With the diode the RC time depended on drive level.

Drive level 1/e RC time
(V) (ns)
2.0 150ns
3.0 80ns
5.0 30ns
6.0 25ns

You'll have to subtract the probe C to get a guesstimate... At low
voltage it didn't really look like an RC.. it had a linear region...
anyway. Not much use for your high power diodes.

That looks like surprisingly little capacitance. Many such diodes will
run at 2V at full bore though.

As far as relaxation oscillations. Well, just saying the words I'm
already starting to hand wave... But my limited understanding is that
this sets the limit in terms of modulation frequency. Lots of hits
searching for "diode laser" +"relaxation oscillation" I'd post a
link, but no idea as to the quality of any of them.

Yes, that is the 2nd and much tougher barrier that Phil mentioned. I
don't think we'll quite hit that region yet.

 

[Joerg]

1 ns shouldn't be hard with a 100 mW laser diode. You can do that with
dirt-cheap cmos drivers, or use a couple of ADN2871's.

I think that's just a current shunt driver. Could be tight here. But the
real issue with all things electronic around lasers is obsolescence.
Some major telco deal falls through or has run its course and the stuff
may become hard to purchase. A warning sign is often when distributor
stock drops to the low three digits everywhere.

CMOS drivers aren't very fast unless you use logic level ones and then
holding the diode close to the current limit is a white-knuckle ride.

Any idea why the LM511x series is so slow pulling up?

The real killer is series inductance, internal to the diode package.
We're shooting for 100 ps edges at 1 amp, which puts 10 volts across
every nH of stray indictance. With packaged diodes, a gaasfet driver
isn't going to do that.

The nightmare are those butterfly packages. Whoever "invented" those
should be ...

Can't you make that and the diode capacitance part of a Pi-filter, like
a "peaker"?

 

[Joerg]

Ahh just to clarify the voltage is the size of the step before the 1 k
ohm resistor. After the resistor the voltage was always 1.5 volts...
at these low current that's the forward voltage of the diode. And I
have too much probe C to make any meaniful measurements at higher
currents. I guess with a fast photodiode one could look at the light
pulse rise time... but that's starting to become 'real' work.

Hmm just thinking, I really should have put a DC offset in my voltage
step, that might have gotten rid of the low voltage linear region.
I've got our fancy signal generator at home (testing some other
stuff.) But I'll finsih up tonight and bring it in tomorrow.

Thanks, George, that is great. But only if it's quick and easy. Not sure
if this diode works though because (6V - 1.5V) / 1k is only 4.5mA and I
am looking at 100x that level.

 

[Joerg]

What's your minimum on/off time? What's the max on/off time? Current?

This may not be difficult or expensive.

I don't think it'll be expensive, already got a discrete solutuion for
it. Except that I don't know the capacitance which can put a crimp in
there. Times not cast in concrete yet, on will be low single digit nsec
and then a few usec off, so very low duty cycle. Current roughly up to 1A.

 

[Joerg]

Dunno, it doesn't seem scary to me. You could always add a schottky
diode and do current steering, to get precise diode current and some
speedup for free. That would be cheap. You like cheap!

Current steering alone often isn't enough. Some of the current will
first have to get the capacitance back up from wherever you had to drop
it down to in order to get the current to the required low limit (or off).

Gosh, that is pokey. But it's a 100 volt power supply thing.


stuck to a corkboard with a big pin?

One of the engineers at a client noticeably winced when I took wire
cutters to a laser diode like this. He had given permission but it still
must have looked like de-clawing to him.

http://dl.dropbox.com/u/53724080/Circuits/Butterfly.jpg



A little inductive peaking can help, unless there's too much
inductance already.

Buy me some more beer and I'll show you our tricks.

Still got that wooden nickel so I'll have to come back anyhow

 

[Joerg]

Yikes, that's almost exactly what we're doing. I hope it's not for the
same customer. Starts with letter "n"

Mine starts with the next letter in the alphabet

But there are myriad purposes for all this pulsing. In another totally
different case we were tracking etalon positions for the sake of
measuring mechanical properties of materials.

For some reason I have more respect for this stuff than for high
voltage. Probably because I know someone who was going blind from pulsed
radar exposure. Once you feel the effects it's usually too late.

 

[Jamie]

Don't know yet, just that it'll be in the >100mW class and IR. If we can
get a nanosecond of transition time that'll be ok for the beginning. I
could muscle the capacitance with a gooser circuit but that's always a
bit scary because laser diode can commit optical suicide in less than
the blink of an eye. All it takes is they come out with a less
capacitive model and ... poof.

Sounds like you need to make the diode become part of a resonant tank
with in the power circuit. Using the diodes capacitance to contribute
to the remaining C required.

I can see that being tricky due to the C changing dynamically. At
least I would think that to be the case.

Jamie

 

[Jamie]

Dunno, it doesn't seem scary to me. You could always add a schottky
diode and do current steering, to get precise diode current and some
speedup for free. That would be cheap. You like cheap!




Gosh, that is pokey. But it's a 100 volt power supply thing.




stuck to a corkboard with a big pin?


http://dl.dropbox.com/u/53724080/Circuits/Butterfly.jpg





A little inductive peaking can help, unless there's too much
inductance already.

Buy me some more beer and I'll show you our tricks.

Talking about beer, that reminds me, I need to get some more

P.S.
I don't think I want to know about your tricks while intoxicated!


Jamie

 

[Joerg]

It's just a matter of more tricks.

Yup. I've got one in there which should make it work