100mW+ single mode external cavity diode lasers (ECDL)


Return to home page


1) General remarks

2) Design notes

3) Simple test setup

4) Sample Designs

5) Operation and stability considerations

Other relevant pages:

Detailed individual diode tests

Converting a 780nm to a 658nm 100mW+ single mode extended cavity laser

Free running diode tests


1) General remarks

 An ECDL, which is stabilized by an extra grating as feedback, holds the promise of producing just a single longitudinal mode, and a very narrow line width of a few hundred khz or so (which is by itself not important for holography). Another potential advantage is that such a laser is supposed to be less sensitive from inadvertent external light feedback; apparently, plain laser diodes can be destabilized by back reflected light even if it is weaker than 10^(-6), ie, less then a tenth of a microwatt!).

Much can be said about this kind of lasers and I may do so later.  A classic web site is this one, which has links to all sorts of literature too. See also here for a concise overview, as well as these articles: No1, No2, No3, and No4. See also the interesting notes here and in particular this one.  Note that for applications in amateur holography, the technical requirements are much less stringent as for most professional applications of ECDL (spectroscopy, laser atom traps etc), as we neither need tunability nor linewidths in the sub-Mhz region nor a great spectral purity nor extreme long-term stability. If we aim for a coherence length of a meter or so, which should be sufficient for all practical purposes, then line widths and drifts of tens of megahertz are perfectly fine.  So we don't really need special antireflection coated laser diodes and other special parts; with some luck a home-brew system with surplus parts but nevertheless sufficient spectral properties can be build. 

Whether this leads to a laser that can practically be used in day-to day holography, without the need of frequent retunings, is a different question. As will become evident below, single mode operation is not at all automatic, and depends on careful design (and lucky circumstances as far as the diode is concerned). Single mode zones may be quite narrow in the diode current/temparature plane, and changing the diode current by a fraction of a mA, or changing the temperature by one hundredth of a degree, can induce a jump to a different single mode. The design challenge is to safely remain in a stable zone and not drift though a mode jump or even into a chaotic zone over time, for long enough such as to reliably shoot holograms. This requires very stable diode and TEC drivers and strict thermal shielding. We find that by careful design it is possible to stay in single mode operation without jumps for an hour or more.

To give some idea how differently an ECDL behaves in comparison to a free running diode, see the following two brief movies. The left one shows how modes jump in an ECDL when the current is continuosuly increased by a few mA. To the right the analogous movie for a free-running laser diode:

We see how in the ECDL the spectrum is confined to a small region, and small jumps occur within a region from one single mode to another one (one division corresponds to approx. 0.5nm). When the border of the zone is reached, there is a larger jump back to the other border. Sometimes in a transition we see briefly several modes showing up. For the free running laser to the right, the wavelength is not confined but keeps on shifting with the current. The spectrum is mostly multimode.

All-in-all, here is the summary of what I found so far. As compared to a free-running diode laser, the ECDL configuration can yield single mode operation up to high powers (order of 100mW) and does not require cooling to low temperatures (which alleviates problems with condensation). On the other hand, the single mode zones are typically narrow, unless one happens to have a good diode and an optimal setup. Stringent temperature and current stabilization is critical to avoid mode hops. The main problem is to either prevent, or at least reliably detect multimode operation, which can an ECDL can drift into over time.

2) Design notes

The simplest way is to use a "Littrow"-configuration as shown below, where the first order reflection goes straight back into the laser, and the zeroth order (naive reflection) yields the output. One definitely needs a 1800 grooves/mm or more grating and not less, because otherwise more than one diffraction order appears, and this would reduce power.

For a given number of lines per millimeter, the diffraction efficiency depends on the wavelength, "blaze" wavelength, and also on the polarization, as is exemplified below:

(Source: Optometrics). Shown here are typical diffraction efficiencies for two kinds of holographic gratings, where the red lines refer to the polarization ("p-type") being parallel to the grooves of the grating, and the blue ones to the polarization being parallel to the plane formed by incident and reflecting beams ("s-type") . For the Littrow configuration as shown above, "p-type" thus refers to vertical and "s-type" to horizontal polarization of the laser.


a) Gratings:   So far I used 1800l/mm gratings blazed for UV. See the diode test page for details.

b) As for the LD driver, at first I used the Thorlabs Model LD1255 constant current laser diode driver. Later I switched to my own driver board described here, which has a performance at least as good as the LD1255.

c) Temperature and current stabilization. This is very crucial as one needs to keep the temperature of the laser didoe and cavity stable up to a few thousands of a degree. For a laser that would be useful in practice, one needs to add an enclosure to shield it from variations of the ambient temperature. I found that my MAX1968 based TEC controller (which is a left-over spare from my Coherent 315M DPPS power supply project), with the PID time constants optimized for my Newport 700 based diode laser, does work as-is also for this ECDL configuration, although the temperature swings a few times until it reaches equilibrium after 1/2 minute or so. My own optimized TEC driver works even better. In fact, the diode current needs to be well stabilized too and have low noise.

d) Collimator.  Its adjustment is very critical for stable single-mode operation. In fact the back-coupling efficiency is greatly reduced if the alignment is not optimal. This does not show so much in the output power, but rather in the stability and line width. This requires to rotate the collimator very accurately; which may prove difficult after the unit is assembled. It needs to be done only once, but then the collimator needs to be very well secured; not the slightest wobble is allowed. I collimate the diodes by requiring that the spot 5m away has minimal size. I found that sometimes like 1/100 turn in the 9x0.5mm fine thread can make a hugh difference!  I also found that very precise axis alignment is important, as well as using a high-efficiency lens with little loss. The back-coupling into the diode is highest if the focus spot size is as small as possible; this means that the Fl of the lens should be as short as possible and it's NA (numerical aperture) as large as possible. A good lens in this respect is the Lens-27 from Roithner, but by far the best I found the collimators of Marco Lauschmann.

e) Resonator length. A priori, the distance between the diode and the grating should be as small as possible; then the resonator mode spacing is as large as possible and this helps to prevent instabilities and mode hops. However I found situations that when increasing or decreasing the resonator length by a few mm, the laser runs very unstable or not single mode at all (all other parameters fixed); it seems that there is a window of optimal feedback for each diode, and sometimes this is quite narrow and requires a painstaking cut-and-try method to find the best length.

In my experiments, I found the following behavior as a crude rule: if the feedback is too weak, then the laser is unstable and does not easily lock into single mode, and if it is too strong, then there tend to be extra side modes besides the main mode. Feedback is reduced when making the resonator long (say, 5-10cm), so this can be used for gratings with a large feedback (say, 20-50%). The wavelength selectivity is then increased as well, on the other hand the resonator modes are more densely spaced so there is an increased tendency for mode jumps. For gratings with low feedback, say 5-10%, a shorter cavity (say, 2-5cm) should compensate for this, but while the mode spacing becomes larger so that mode jumps are less likely, the selectivity decreases and this counteracts the stability again. Sometimes single mode stability can be improved by reducing feedback, by somewhat tilting the grating vertically or slighly misaligning the collimator.

f) Adjustment of the grating. The horizontal adjustment sensitivity is not as sharp and one has a range of perhaps 1/4 turn of the set screw on the mount. The vertical adjustment is much sharper, and it adjusts like any other laser resonator, and ideally one would have micrometer screws for turning the grating. It is just borderline doable with an allen key. The tricky thing is to find the right spot in the first place; approximately measuring the angle of the grating and height of the beam beforehand helps to get into the right ball park to start with. The sweet spot is best visible when running the diode just below the lasing threshold, and when you cross it there is a flash where the power multiplies.

A guide to the correct location is to watch for the backreflection of the laser diode facet, which is visible when one is not too far away from the sweet spot and when running at high power. It helps to first not completely fasten the holder of the grating and by wobbling around one can easily spot that backreflection. Then one can fasten the grating holder and turn down the power near threshold for finding the sweet spot with the set screws. Final adjustment is done by using a photo diode together with an oscilloscope, which allows to very precisely find the best adjustment. One may also just keep on reducing the laser current and adjusting the grating in several runs, in order to minimize the threshold current.

g) Orientation of polarization. Whether horizontal or vertical polarization works better, depends on the grating and the diode in question. It is important not to provide too much feedback to the laser, and it seems one should keep it at roughly 5-20%.  From the figures above one infers that for red diodes an UV blazed grating is generically better for horizontal polarization and a grating optimized for visible wavelengths may be better for vertical polarization (recall that for holography, horizontal polarization is often preferred). However it seems that the gratings and the diodes vary very much and probably the only practical way to find out what works best ist to try out. Note that due to the larger feedback for horizontal polarization, the maximal diode current before destruction is lower... as I had to learn the hard way! 

For representative tests of individual diodes, see here.

3) Simple test setup

Here a pic of a basic test setup:

On the right there is the laser diode in a press-on mount, incl. two thermistors and collimator, while on the left you see a 1800 lines/mm holographic grating optimized for UV. It is attached to a tiny New Focus 9876-K kinematic mount, and the whole thing is mounted on a cold plate sitting on a 30x35mm peltier element for temperature stabilization.

4) Sample Designs

Due to lack of thermal isolation of the simple test setup, and also because I like to develop a robust design for production, I spent some thoughts and put together a complete unit, including my proven driver board and noise detector. First tests showed that the setup is very stable and easy to work with. Here is how it looks:

There are some plugs for externally monitoring and controlling diode current and temperature, as well a pot for changing the current by a milliamp or so, in order for being able to move away from a mode hopping zone.

The laser unit is put seperately on a head spreader block, which is mounted with a clamp such one can shift and rotate it a little, in order to have a straight exit beam. This is because re-adjusting the grating will always deflect the beam, so re-alignment of the base plate can compensate for this:


The cold plate is 8mm thick and sits on top of a 30x35mm Peltier element, which by itself sits on a 10mm base plate. The whole thing sits on a heat sink, which currently is just a massive slab of aluminum. Thermal coupling to the driver circuit is minimized in this way; the large common sink has enough thermal inertia in order to suppress a cross-coupling instability.

Here is a 1:1 drawing of the laser unit (dimensions are in mm, and the angles refer to a 1800l/mm grating for use near 650nm):

It can be taken out as a whole, which makes changes and crude adjustments easy:

Here pics of some other ECDL prototypes I built (left: 100mW 642nm, right: 110mW 658nm):

This is a 445nm ECDL with glass plate feedback and beam correction by prisms:




5) Operation and stability considerations

While ECDLs can provide single longitudinal mode operation in excess of 100mW, it is nearly impossible to guarantee a "adjust once and forget" operation for extended periods of time. Due to inevitable thermal drift, diode ageing, weak feedback, etc., the operating conditions change with time, producing either mode jumps (which lead to banding of a hologram), or worse, run into chaotic mode behavior where any hologram would be ruined. With a good driver and a good mechanical design, after sufficent warmup time, a laser can stay single mode for one hour or more, but earlier or later, one needs to readjust the laser. The simplest way is to make the current adjustable for a few mA in order to get back into a stable regime.

To make the point, below is some representative long term stability measurement of an ECDL running the Mitsubishi ML101J27 at 160mA/110mW:

We see that after approx 2700 seconds, there was a jump to multimode operation, signified by a brief burst of noise. After some while the diode current was changed by one mA, and single mode operation resumed.

The main problem is, of course, to reliably identify and reassure SLM operation in the first place, during day-to-day holography sessions; and obviously one would prefer not having to run an OSA or scanning interferometer concurrently. One way is to monitor, and make audible, the noise in the light output, which signals instabilities. This works reasonably well for free-running diodes, but only for certain diodes and configurations in ECDLs (some investigations were summarized here). In general, there can be multimode operation in ECDLs which does not reveal itself be elevated noise (see the movie below).

An ordinary Michelson interferometer would most accurately reflect the relevant operating conditons, but it is clumsy as well and prone to destabilizing backreflections. On the other hand, an old trick for checking diodes is to simply use a glass plate in a diverging beam, and watch the interference pattern from the reflections of front- and backsides. Such a setup is called Fizeau interferometer. The point is that the diode resonator is shorter than the thickness of the plate, so the interferometer can capture mode jumps despite the very small difference of path lengths. Obviously such an interferometer doesn't work for other lasers with long resonator, and one might suspect that this applies to ECDLs as well, which have a resonator length of a few centimeters; in other words, one would expect that the mode spacing of the extended cavity would not be resolvable by such a short path length.

However, the surprising good news is that it seems a general feature of ECDLs that mode jumps are typically much larger than jumps just between the extended cavity modes; rather the jumps occur between the native diode modes! This must be the effect of the combined superposition of diode die and extended cavity resonance conditions. I have extensively investigated this using both the grating OSA and scanning interferometer, and it seems to be a general rule. Thus, all what is necessary is to send a small portion of the beam through a fl=50...200mm lens across the room, reflect it back by a microscope slide, watch the interference fringes, and restore the contrast if necessary by minimally changing the diode current.

Here is a movie that shows interference fringes generated in this way, side-by-side to the display of the grating spectrum analyzer, while the current was slowly changed. One can clearly identify single mode operation in this way:

Included as soundtrack is the laser light noise, which allows to hear mode jumps; on the other hand, upon closely watching, one can see that multimode operation is not necessarily accompagnied by extra noise, so monitorung noise is by itself not sufficient. The movie refers to an ECDL with a 642nm HL6385 diode at around 100mW (260mA).


Return to home page

Vers. 4.2-11/2011