Current Controller for Laser Diode

Any other constraints? Does the diode need to have one terminal grounded or can it float? Any frequency response issues?

If frequency response isn't critical and if the diode can have both terminals floating here's an easy circuit:

VCC Vref + + | | |\ | '----|+\ '---------- Vcntl ___ | >-----. to laser

------|___|--o---|-/ | .---------- R3 | |/ .-. | | R1 | | | | | | | | C1 '-' ||-+ | || | ||

Move C1 from the FET's gate to the opamp's output.

Oh dangit. Yes, do.

Like Rene says, at 10 mA just use the opamp.

Vctl-------- + out-------+ - | | laser | | +------------+ | 1k maybe | | gnd

But the laser sounds unusual. I've never heard of a laser whose wavelength was proportional to current, or one that would lase below roughly 10% of max rated current.

John

Many of the ones I've seen incorporate an internal photodiode to monitor their power output. Its usually better to regulate the laser output using this as feedback rather than just laser current.

A precision current source can be built from an operational amplifier, by measuring the current from a shunt resistor to the inverting input. On the non-inverting input, you have the voltage preset from a DAC. A standard circuit really. You're lucky that your speed is close to DC.

Rene

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

The OP slightly misstated that feature. What semiconductor lasers will do is lase at a wavelength that varies approximately linearly with current, provided you remove an offset term. The effect arises mainly from thermal modulation of cavity dimension. It can be used for interesting purposes when the limited wavelength range spans particular spectral lines one might wish to examine.

Also:

Depending on the op-amp, put an equal RC on the non-inverting input. This removes the common mode bias current and noise current.

Make sure Vcc is well regulated/decoupled at high frequencies. The circuits's impedance falls at high frequencies.

In article , John Larkin wrote: [...]

The wavelength is more like Lamda = A + B*I

This is true for most/all semiconductor lasers. The "N" of a semiconductor increases as the carrier density increases. Since the cavity is formed by mirrors on the ends of a chunk of semiconductor the wavelength tends to decrease as the current increases.

It is handy that the wavelength also increases as the temperature increases so the increase in temperature that goes with the increase in current has an effect in the same direction. If this wasn't true "line locking" a laser to an absobtion line would be very troublesome.

In article , Larry Brasfield wrote: [....]

This is true in the long term. For the very short term, the carrier density in the junction area dominates.

Take a look at NISTs "chip scale atomic clock" project for a good example.

The NIST program is using a vertical cavity laser in exactly this way. It works quite well.

This is better in terms of performance but way too much money for many projects.

This is true but rather unreliable since the diodelaser cavity is too short. There are just too many modes in it. A better approach would be to antireflex the laser diode and have a controlled external cavity with a grating or fabry perot.

Rene

A grating and a lens doesn't cost much and are quickly mounted. Movement either by a piezo, or an RC servo. Yes, the antireflex coat is not for free but doesn't cost that much either. At least you can achieve single line emission under the gain profile of the few nm.

Rene

In article , Rene Tschaggelar wrote: [...]

You've got my interest. How much for the needed hardware?

NIST is trying for a atomic clock that competes with the OCXOs. They've got quite a ways to go yet on the development.

BTW: in the NIST system, they have to modulate the light at the 9GHz frequency because they are using first order coherent population trapping. Making a 9GHz cavity that small would be trouble some. Doubly so because you really want a wide peak on it so you don't get pulling.

The hardware is fairly cheap. You need the laser diode of yours, but on the backside, where usually the monitor is, you need an antireflex coat. A lambda quarter of falcium floride or such. Then you need some optics to expand the beam, An achromat or a microscope lens. Having the beam widened up, it goes as moreless parallel beam to a grating. 30$ or so at Edmund Scientific. The grating retroreflects the wanted wavelengths back. The selection of the wavelength is the angle of the grating. This job is mainly mechanical, setting up the lot on a sturdy plate, adjusting the angles, remove hysteresis ... Note that laser gain equation have now the lenses and the grating in it as losses. This means the lens system should accomodate for the large NA. A longer laser cavity has less longitudinal modes and the grating is selective amongst them.

I read some articles about that. Considering that I get an OCXO in less than half a cubic inch, running between 0 and 50degC, at less than a watt, for 500$, that is quite a task.

A perhaps better newsgroup than s.e.d would be alt.lasers or sci.optics

Rene

In article , Rene Tschaggelar wrote: [....]

Define "fairly cheap". Looking at what you've listed below, it looks to me like the laser system will cost over $1000US. I was hoping you had come up with something that the group in the land down under had missed. It looks like you've suggest the same basic kit as them. I guess that is reasonable as it is most likely the best way to do it if you don't have the new vertical cavity technology. The one thing they stressed was the need for very good mechanical stability in all of the parts.

They seem to be inching towards actually doing it. The laser production yeld is still a bit of an issue. The vertical cavity laser has to run at the 894 line. Only a small cross sectioned ring of them on the wafer end up wanting to run there. The rest are either too high or too low. This is a silicon growing issue that they have to get a handle on.

Running the light through the body of the cell twice by using a mirror on the far side solves one of the big mechanical issues by forcing the light to average to parallel. Switching to CPT turns the 9GHz from RF cavity to modulator issues.

One of the folks who is working on this told me that he can see a production price of something like $100 in the future. There is nothing in the system that is by its nature expensive other than perhaps the laser its self.

Well, an aspherical lens, a few squaremillimeters of AR coat, a squarecentimeter of grating is doable for a couple of dollars, it is the mounting and handling that is time consuming. With some automation, in numbers, the price should come down. Yes, the mechanical stability is important. Possibly thermally stabilized. But a micro setup should be doable.

Ah, they are locking to maximum/minimum absorption with the 9GHz sidebands ? That is not that trivial. Nor immediately to be made small. With some custon microwave chips though...

Rene

In article , Rene Tschaggelar wrote: [...]

Its the fine machined parts that worry me the most about the costs. One nice thing about the NIST project is that the fine machining is done with normal silicon processing steps. The disadvantage is the same thing. Mere mortals can't round up enough cash to buy the machines to make millions of low cost atomic clocks.

They have overlooked a couple of things (on purpose perhaps) in their presentation. The cell has to be warmed to about 60C and there has to be a very good magnetic shield and "C coil" around it. In the proto-type the cell is suspended inside another glass chamber using Kapton tape. None of this is done with normal IC processing stuff.

It is the machine parts that worry me most with optics too. For those that are used to work with inch bolts, the Thorlabs mounts may appear workable, but when the expectations are somewhat higher, there is only custom parts. Yes, I can can spend a day or so behind the milling machine but the results are, well, on the border to be useable.

Well, these are laboratory solutions for one-of.

Rene