Diode tests for ECDL operation

Return to home page

Return to ECDL page

See also:

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

Free running diode tests

 

Overview

Below you find some representative findings for a variety of common diodes; you will see that they behave quite differently and right know I wouldn't know any criterion to decide beforehand whether a laser diode performs well in an ECDL configuration -- single mode operation is by no means automatic! Generically it seems that a diode that is labelled single mode performs better also in an ECDL. But even for the same type of diode there are large fluctuations in the maximally avaliable stable single mode power. It is important that the back-coupling of the grating is optimized to the diode. It depends on the grating itself, but also on the collimator, the adjustment and the resonator length. See my findings below.

Here is a representative summary for some diodes so far:

1) Rohm RLD65PZB5 80mW/658nm laser diode:   unsuitable

2) Sony SLD1239JL-54 100mW/658nm laser diode:   unsuitable

3) "long open can" Mitsubushi ML101U29 150mW/660nm diode   some specimes did well to 100mW and more

4) Mitsubishi ML101J27 130mW/660nm laser diode  stable operation seems generically possible to up 100mW and more; good value

5) Opnext HL6385DG 150mW/642nm laser diode  stable operation possible up to 90-100mW, very good

6) Sharp GH04P21A2GE/PHR-803T 100mW/406nm "blu-ray"laser diode: stable to 15mW only

7) Mitsubishi ML101F27 150mW/660nm laser diode: unsuitable

8) Nichia NDB7412 1W/445nm laser diode, tests of feedback with glass plate: stable operation to 60-80mW, with compromises on stabilty and spectral purity more than 200mW are possible. Bad beam quality, not TEM00

9) Opnext HL63133DG 170mW/638nm laser diode  stable operation to more than 100mW, very good!  Highest quality and priced diode.

10) Osram PL450 80mW/450nm laser diode   ok only for ECDL, 40mW, stability tricky

11) Opnext HL45023TG 80mW/445nm laser diode  TBA

12) Opnext HL63603TG 120mW/638nm laser diode ECDL 70mW, very good value

13) Mitsubishi ML520G54 110mW/638nm laser diode ECDL 60mW, very good value

14) Osram PL520 50mW/520nm laser diode ok only for ECDL, 40mW+, good

15 ) Mitsubishi LPC-836 300mW/655nm laser diode: unsuitable

 

(Mitsubishi ML520G71 300mW/638nm, Opnext HL6388MG  250mW/637nm laser diodes: unsuitable due to multi transverse mode)

Note however that this needs to be qualified - while single mode operation seems often possible beyond 100mW, most diodes become unstable and the single mode zones become very narrow. Such operation regions may not be suitable for practical holography work. As a crude rule, I found so far that generically, ECDLs become unstable beyond roughly 60-100mW, quite independently of diode, grating and operating details.  Another crude rule is that best are diodes for red with rated CW power 100-200mW but not more.

The question is thus whether the considerably higher effort to build an ECDL is worthwhile. The answer is that only by a careful matching of diode, collimator, grating, resonator length, adjustment, and operating parameters, higher powers can be reliably obtained (and to find this out, takes most of the effort). Minor modifications, like changing the collimation or the resonator length, can change the picture a lot! The issue is not to just obtain high power single mode operation (which is easy), but also to have a robust and stable operation. One would also like to have a simple means to verify single mode operation; for some toughts, see here.

Below are the data for the gratings I used, which I measured to a few percent accuracy (I know that not all entries are consistent, but this is the best what I could do). All are 1800 lines/mm gratings blazed for UV, Grating 1 I got from ebay, Grating 2 is gold plated and taken from the commercial 780nm ECDL laser described here.  Gratings 3 (model  R11-075) and 4 (model 3-462) I ordered from Optispac and Optometrics, respectively, each was around $50 at 12.7x12.7mm^2 size. I had the opportunity to have some of them gold coated, which does not only enhance the reflectivity at red wavelengths but also protects the grating from long term deterioation. I denote the gold coated ones with an extra suffix "g". I didn't find a significant difference when scattering along or against the blaze direction. "refl" and "out" refer to the reflected first order beam back to the diode and the zeroeth order output beam, respectively. Moreover "vert" and "horiz" refer to p- and s-type polarization, resp.

Grating 1 refl vert out vert refl horiz out horiz
642nm 5% 85% 18% 67%
658nm 3.8% 84% 17% 63%
445nm 12% 76% 28% 30%

Grating 2g refl vert out vert refl horiz out horiz
642nm 8% 86% 31% 59%
658nm 6.7% 87% 31% 60%

Grating 3 refl vert out vert refl horiz out horiz
658nm 7.5% 81%    
445nm 20% 70% 28% 30%

Grating 3g refl vert out vert refl horiz out horiz
642nm 9% 84% 25% 61%
658nm 7.8% 87% 25% 61%

Grating 4 refl vert out vert refl horiz out horiz
642nm 3.6% 78% 11% 50%
658nm 3.4% 80% 12% 50%
445nm 7% 65% 2.8% 26%

Grating 4g refl vert out vert refl horiz out horiz
642nm 8% 80% 26% 46%
658nm 6.6% 82% 24% 50%

The entries marked in red I found suitable for high power ECDL operation. Note that the return coupling for blue light is generally quite high, which is not too fortunate, but this is a consequence of the UV blazing and it would be much worse for blazing at visible wavelengths.

Note added (1/2012): Apparently Optispac is not doing business any more. I got a couple of new gratings from China Star Optics Technology, these are holographic gratings (so have somewhat lower diffraction efficiency), blazed at UV. Grating 5 has 1800 l/mm as the gratings before, and Grating 6 has 2400 l/mm, which I got for experimentation with 405nm and 455nm lasers. Here are the data I measured:

Grating 5 refl vert out vert refl horiz out horiz
638nm 8.8% 77%    
658nm 8.3% 78%    
450nm 17.8% 57%    

Grating 6 refl vert out vert refl horiz out horiz
638nm 6% 78%    
658nm 6% 80%    
450nm 15.5% 61%    

 

 


1) Rohm RLD65PZB5 80mW/658nm diode

Below the result of some automated scanning over the diode current/temperature plane, done similar as explained here:

On the left: shown is the linewidth of the ECDL laser as per my optical spectrum analyzer; green/blue correspond to single longitudinal mode operation.
On the right: AC noise in optical output, which is a measure of mode hops.

The granular structure shows that single - and multimode operating points lie almost dense next to each eather, and this means that hitting and maintaining a single-mode point would require quite a fine tuning; in practice this laser would not be useable. It also exemplifies that an ECDL configuration does not automatically guarantee single mode operation.


2) Sony SLD1239JL-54 100mW/658nm laser diode

Terrible mode chaos, just like for the free-running diode. Perhaps deservedly, the diode died from overheating induced by a faulty thermistor connection.


3)Mitsubushi ML101U29 "long open can" 150mW/660nm diode

I achieved 100mW at approx 220mA without problem. Almost no sign of multimode operation at currents < 170mA. I used Grating1 and vertical polarization. Here a scan over the current/temperature plane, with noise superimposed as black:

Thus, we see differences as compared to a blank, non-grating stabilized diode laser. The general pattern is much more uniform than for a non-EDCL diode laser, in this case it is single mode operation up to approx 170mA. The slight variation of color is due to the change of output power (which simulates a slight change of line width). Mode hops, signified by "black" noise, occur on regular bands. Thus changing the current by just very few mA typically induces a mode jump, and the same applies to changing the temperature by a few hundredths of a degree. 

Here are zooms at around the same temperature, for lower and higher powers:

We clearly see that at lower powers, there is almost everythere single mode operation, and mode hops between different modes occur at regular bands. However at higher powers, seen on the right (around 100mW), there are multimode zones interspersed with single mode zones, and adjusting to, and staying in, a single mode zone becomes more of a challenge. So staying below 170-180mA gives less power, but apart from occasional mode hops there wouldn't be much to fear.

Note that there are in principle two or three relevant length scales (and thus possible longitudinal mode spacings) in the laser: first, the optical length of the diode cavity (which gives a mode spacing of say, 100Ghz), then the distance to the diode output window (which is irrelevant for an open can diode), and then the distance between the diode and the grating. While the CCD spectum analyzer can well detect the modes of the diode itself, its resolution may not be enough to distinguish modes from the external cavity (in particular if the cavity is large and thus the mode spacing small).

Thus to be sure I checked the spectrum also with my scanning interferometer which has much higher resolution. The result shows a line width of less then 5Mhz:

(T=17C, P=90mW)


4)Mitsubishi ML101J27 130mW/660nm laser diode

This diode behaves very well in free-running but even more so in ECDL configuration. Without any optimizations, for vertical p-polarization I achieved 110mW single mode at 160mA current right out the box, at around 19C. However, the single mode zones were quite narrow, like 1mA.

Here a representative landscape scan; as always, colors encode linewidth (cyan corresponds to single mode), and superimposed grey/black noise indicates mode hops:

We see that there is predominantly single mode almost everywhere, even up to 180mA which yields about 100mW. However within single mode zones there are mode hops between different single modes, which is indicated by the black noise lines. I also made a comparison how things look for different gratings and collimators, for the same diode and operating region. The results are interesting:

On the upper row, the collimator was the Lens-27 from Roithner. The label "4g" or "3g" refers to the gratings described above. On the lower row, the G-650-1 collimator was used; it has just a few percent less efficiency, but makes a drastic change with regard to the zones of single mode operation. The approx power at 180mA was (clockwise from upper left): 100mW, 104mW, 96mW, 100mW, which is consistent with the fact that Lens-27 and grating 3g are a few percent more efficient than G-650-1 and grating 4g.

We conclude that optimal results can be achieved when all factors (diode, operating point, collimator,grating, adjustment) happen to conspire together in a fortunate way; minor modifications can complelely change the picture.

When repeating the test for p-polarization (horizontal), the power was considerably lower, namely 68mW at 160mA. The diode died at 180mA, which is well below the maximal current rating of 200mA. However what counts is the power density at the diode facet and this must have been too large due to the larger feedback from the grating.

I have put that diode also in the converted, commercial 780nm ECDL described here. It uses Grating 2 which has somewhat higher reflectivity, and an unspecified aspheric lens. Here the result:

We see that there is generic single mode operation only at relatively low power. Below are higher-resolution zoom-ins at around the same temperature but different currents, the right shows the region near 100mW:

We see that at higher power the single mode zones become more narrow, which is as expected because of the higher gain so that more modes tend to oscillate. The separation of mode hoping zones is in the order of 1mA which requires a very precise current control.

 


5)Opnext HL6385DG 150mW/642nm laser diode

Lots of people have expressed a desire of a 100mW or so single mode laser at 640nm. I found that free-running diodes can be reasonably stabilized only up to around 70mW, so why not trying an ECDL setup. So I crossed my fingers and put my last HL6385 diode in the simple ECDL setup described above. I was careful and started with vertical polarization which gives a feedback of like 3.8%. I found it to produce 105mW single mode at 239mA at around 17C without effort, but remembering what happened to my best Mitsubishi diode I didn't dare to turn it up higher, so I don't know what the limit is; I just content myself with 100mW for the time being.

Below are two scan plots, the left shows an overall picture and the right one is a higher-resolution zoom at around 100mW. As always, colors encode line width (cyan corresponds to single mode), and superimposed grey/black denotes noise which signals mode hops:

Here a check with the scanning interferometer at 230mA:

Top scan is about 200Mhz/Div, lower scan is ten-fold zoom at approx 20Mhz/Div. The displayed line width is limited by the resolution of the SFPI and at most like 10Mhz. This corresponds to at least some 30m coherence length. However this is so only in optimal circumstances, minute changes of the operating point can create complete mode chaos.

Getting a bit more adventurous, I switched the polarization to horizontal to check out stronger feedback (approx 18%). Surprisingly, not only the power (75mW at 225mA) but also the spectrum got worse, it is like for the Rohm diode above:

Note the fragmented structure of the bands, it reflects hysteresis effects because the current scans in alternating directions. In other words, a mode can jump at slightly different values depending whether you had increased or decreased the current.

So all-in-all, it seems that the stability gets worse with larger reflectivity of the grating. Thus, due to the larger feedback, the laser has an increased tendency to run multimode -beforehand, I had thought it would be the other way around!

 


6) Sharp GH04P21A2GE/PHR-803T 100mW/406nm "blu-ray" laser diode

 

Since the longitudinal mode spectrum of the free running blu ray diode is intolerable, an obious question is of whether an ECDL setup is of any help. A single mode laser at 406nm would be particularly interesting for DCG holographic emulsions since the sensitivity of those is dramatically better as compared to green DPSS lasers, say.

I placed the diode in an Aixiz 12mm mount tube (itself located in an alu brick), of which I had removed part of the 9x0.5mm thread in order to fit a braodband, 400-600nm coated Geltech C330TM-A lens.The grating was the same as above, which is UV blazed at 1800 lines/mm. The angle needed to be changed from 54 to 69 degrees, see the diagram above. Also I used vertical polarization.

I achieved 14mW at 50mA, 38mW at 85mA and 51mW at 100mA, which is in the order of 50% as compared to the bare diode. On the CCD spectrum analayzer the mode spectrum looks as follows:

(T=15C, I=50mA,P=14mW)

(T=15C, I=60mA,P=20mW)

(T=15C, I=100mA,P=51mW)

Thus, while this is much better as compared to the free running setup, the diode is not really useable for holography beyond 15-20mW or so beause it ceases to run single mode; this is consistent with what I saw for the the 405+/-nm laser diode specs at Toptica. I didn't yet perform a systematic scan over diode current and temperature, and thus cannot exclude domains which are better behaved. Moreover, perhaps also fine tuning the adjustment, in particular of the collimator lens, would improve the mode structure. I may try if I got around doing it.

Note also that the apparent line width is somewhat larger as compared to the one shown in the plots for the free running setup. As said above, the CCD can distinguish only the modes coming from the diode cavity itself, and these are the ones shown here. In principle, the laser could still run with several modes of the extended cavity, and the apparent increase in line width may just indicates this. On the other hand it may also be due to the fact that the diffraction order for 406nm is different and thus the resolution of the CCD spectrometer. In order to check this and find a possible fine structure with the scanning interferometer,  I'd need however suitable high reflective and curved dielectric mirrors for 406nm for the latter, and I have no idea where to get those from.


7)Mitsubishi ML101F27 150mW/660nm laser diode

This open can diode is similar to the ML101U29 diode: free running totally useless for holography, but decent diode for ECDL setups, I achieved 80mW single mode at 172mA without problem (grating #2, diode vertical polarized). The high resolution plot looks like this (the diode was wronlgy labeled on the plot, it should read ML101F27 #1):


8) Results for the Nichia NDB7412 1W/445nm laser diode

The first diode I checked works surprsingly well SLM to 50-60mW in free running mode, and next thing was to check its ECDL potential in a test setup. The first results were disappointing, where I used grating #1 as defined above, vertical polarization and a Lens-27 for collimation: the single mode zones didn't extend beyond the free running SLM zones, and the power was less while the stability was worse. However the next try was better, where I used the well-known Aixiz A-HGL-905-3H collimator (made for 405nm) and grating #3 which gives more feedback:

Here a comparison of the power levels (the dashed line refers to the free running diode):


We clearly see how the lasing threshold is significantly reduced by the feedback of the gratings. Moreover we see a plateau until approx 60mW for which single mode operation is frequent, and the rising slope beyond that seems to reflect that the diode then predominantly runs multimode, with higher efficiency.

The second diode I tested under the same conditions turned out to be generically worse, consistent with its free-running behavior. However, after very carefully adjusting the collimation, choosing longer resonator and also slightly vertically tilting the grating, single mode regions were found even up to more than 200mW:

Here is the plot of output power (the dashed line refers again to the free running diode):

Another test was made with grating #4, which has considerably less reflectivity. First a long resonator was chosen (see the narrow spacing of the mode jumps), and the systems was very unstable even at low currents, apparently because there was not sufficient feedback:

Then a as-short-as possible resonator was chosen where the feedback is stronger, and the system turnd out to be much more stable. It turned out that even in regions with noise, the system was mostly single mode plus some weak extra modes. The complete absence of noise quite reliably indicated single mode operation in this case:

This seems a quite useable configuration, since a simple noise detector is sufficient to detect single/multimode operation. We see again, ECDLs are tricky and require a very carefully matched configuration, for each diode, and minor changes in the adjustments can have a major effect on overall stability.

Even better turned out a test for diode #3 in a short-resonator configuration with grating #3, which achieved more than 200mW quite stable single mode in the first try:

Feedback with simple glass plate

There were interesting reports here and here that actually a simple glass plate may provide enough feedback in order to stabilize the 445nm diode to single mode operation. Needless to say would is very intriguing as one may get rid off the grating. So I have tried two versions, namely with a simple disk and with an wedge plate, both uncoated fused silica of 1mm thickness. The wedge plate separates the reflections of the front- and back sides, which may be a good or a bad thing ;-)  Thus it effectively reflects approx 3.4% back, and the ordinary plate 6.8% modulo interference effects. Correspondingly I found that the wedge plate reduces the threshold current by approx 10mA and the ordinary glass by approx. 20mA. I coulnd't reduce it by like 40-50mA as in the reports, where an uncollimated beam was used -- I use an Aixiz 405nm collimator. Also I found that my mounting wasn't precise enough for just having the glass flat-on the mount, so I needed to make the glass plate adjustable; here is a pic of the test setup: 

Here is the result of a current-temperature scan for the wedge plate:

and here for the ordinary glass plate:

We see that when there is feedback from both sides of the glass plate, single mode operation is predominant up to approx 250mA (80mW), and then increased noise and instabilities appear. It seems that the relatively weak feedback (6.8%) stabilizes the laser where essentially one supermode wants to lase, while at higher currents the appearence of several incoherent supermodes cannot be suppressed.

Below is a movie  (3.3MB) that shows the spectrum when ramping the current up and down again; the increased noise and instabilities are easy to discern. Note the acoustical feedback at certain points, it arises due to the laser acting a a microphone:

Finally I made an effort to try a very short resonator, with the hope that the broad mode spacing would facilitate single mode operation. Alas, in effect I found that at higher currents there is an increased tendency to run extra side modes. Using a very low fl/high NA, high efficiency lens from Swisslas, and a special mount, I managed to get down to a resonator length of about 1cm. The scan looks as follows:

So again we see a transition zone above which the laser tends to run run multimode, though pretty stable without great mode chaos, ie., without much noise.Which is bad if one likes to run the laser with only monitoring the noise. The power around 200mA is about 60mW.

Feedback with thin plate

Following a suggestion by Ahmet to make use of an etalon effect of a very thin plate, I also tried a microscope slide cover glass plate of perhaps 0.15mm thickness. However the problem was to achieve sufficient flatness over the large beam diameter, and the output was predominantly multimode. Perhaps with a proper mounting technique things can be improved (I just glued it over a 5mm hole).

Feedback with "wrongly coated" lens

From the above results I got the idea that even a collimator lens with no coating, or a coating for red, could provide enough feedback for stabilizing the laser. So I put in a collimator with the lens G-650-1, which is coated for red and even has a distrinctively strong blue hue when looked upon. And indeed we see single mode regions beyond 260mA (checks for higher currents are underway):

(the pronounced horizontal band is due to a disturbation).

We see that plenty of single mode regions persist, which are widely separated due to the short resonator (perhaps 5mm). Moreover we see that there is, unfortunately, little correlation between multi-mode operation and noise in the light output (that's why I have not placed the plots on top of each other). The phenomenon of strong noise beyond 240mA persists here as well, I need to find out whether it is detrimental for holography or not; so far I couldn't discern any particuliarity in the spectrum apart from weak side modes. There might be a low-intensity noise floor or something like that.

Note also that the power loss was like 50% with this collimator, so alltogether the G-650-1 is not a good option.

Feedback with AR coated plate

I also tried a 1mm thick plate, AR coated on both sides, of which I estimate the reflection at 445nm to be at around 2%. The resonator length was about 2cm and the Swisslas collimator was used. Surprisingly I could get the threshold down to 148mA at 17.5C, which indicates a much stronger feedback than expected; perhaps the optical quality (flatness?) is better than of the other plates. I also used a much more stable and better adjustable mount, namely a MFM-50 flexure mount of Newport, which can be attached right to the diode mount for maximum rigidity and short resonator length. I found the whole setup quite promising, a test build looks like this:

The scan looked in line with expectations, ie stable single mode operation until ca 250mA or around 70mW; and beyond an increased tendency for instabilities and multimode operartion with isolated but unstable single mode spots up to 200mW and more:

Feedback with etalon

I have a bunch of quartz etalons for argon lasers around, they are approx. 1cm thick and their faces are extremely parallel; they are coated with a bluish tint and I estimate a few percent reflectivity on each face. I placed one close to the diode, with perhaps 2cm distance beween the closest face and the diode. The feedback was very strong, actually I found to most extreme reduction of threshold current I ever have seen, namely from 198mA to 93mA! Indeed the etalon acts as a strong reflector at certain wavelengths, and clearly lasing will primarily occur at those. The net result was disappointing: stable single mode operation was possible to 15-20mW only, above that there were always severel modes, not chaotic but isolated ones. This seems to be typical for too strong feedback.

Feedback with extra slit

So far I found that little feedback stabilizes the diode well only at low currents, while at higher currents chaotic multimode operation occurs, similar to no feedback at all. Stronger feedback pushes the chaotic regime up, but then there nevertheless remains a tendency for having several isolated modes at elevated currents. When tuning through those modes, by changing the current and also simultaneously spectrally analyzing the spacial components of the beam, it appears that many of those extra modes have different spatial patterns, so correspond to different supermodes. So I got the idea to suppress those modes by having strong feedback through a thin slit, which would hopefully favor certain transverse modes and suppress other modes that would otherwise be supported.

And indeed, first tests were promising, I used a slit with approx size 5mm x 0.2mm near the collimator in a grating ECDL, and immediately found a very clean spectrum! The threshold current was increased from 121mA without slit to 128mA with slit. As expected, the power is reduced and I obtained around 95mW at 350mA, however quite stable so. I didn't tune up higher since a lot of power gets absorbed by the slit and I wanted to avoid damage. The spacial far-field pattern was much better than before, so the transverse modes were drastically reduced. Essentially only the middle horizontal lobe gets amplified and spreads out, so the familar stripe pattern gets suppressed; that is certainly more than welcome! However, several vertical stripes occur, it looks as if they'd come from the slit because they move with it. Perhaps they arise from irregularities of the slit, and could be ameliorated by using a precision slit.

Here a pic of the preliminary setup, The slit is glued onto an aluminum slab that is press-mounted against the diode mount:

When scanning through the current, the mode jumps are very small, unlike for the other setups tested so far. This means that now indeed the longitudinal modes of the extended cavity are predominant and not any more the transversal modes of the diode!

So this seems to be a promising route to go, obviously much more fiddling is needed in order to see how far this can be pushed, as there are now two more degrees of freedom to play with: slit thickness and axial location within the resonator. First of all I need to find a good way to mount the slit in a stable but adjustable manner. Stay tuned.


Here as summary some comparative data for diode #3 in various configurations (at  approx 17C, short-fl aspheric lens from Swisslas):
feedback refl approx % resonator length, cm threshold current, mA
none -- -- 189
fused silica wedge 3.4 2 174
fused silica plate 6.8 2 169
NFK-5 plate 8 1 153
grating #4 7 2 142
grating #3 21 2 121
grating #3 plus 0.2mm slit ? 4 128
microscope slide cover plate 8 1 152
AR/AR coated plate 2-3 2 148
Ar laser etalon few 2-3 93
G-650-1 ? few mm 185

For a brief collimator efficiency comparison, see here.


Summary of the NDB7412

For the blue Nichia 1W diode a pretty universal pattern seems to emerge. Relatively independenly of the feeback (within reason), the max. stable output is in the order of 60-80mW; what changes with increased feedback is that the threshold current is lowered, but the onset of instabilites/multimode operation does not improve, rather it tends to occur at lower currents as well. There seems a universal border at 230-260mA above which multimode operation becomes predominant.

What is surprising is that the diodes can be stabilized with very little feedback, of just a few percent, which already leads to a drastic reduction of threshold current. This suggests that the internal reflectivity of the output facet is quite small.  It is also surprising that gratings do not perform significantly better; I believe that the wavefront of the diode is so "dirty" that the grating cannot not really play out its advantages. In particular, multiple transverse supermodes are not well suppressed. It seems that this can be improved by inserting a narrow slit into the resonator which suppresses higher transverse modes.

With sacrifices on stability and mode purity, more than 200mW are possible in isolated spots, also relatively independent of the feedback.  However, there is then a tendency for several supermodes to lase simultaneusly, which means that there are several supermodes with different frequencies; in other words, different spatial regions of the beam are not coherent with respect to each other, which is diffcult to see if one samples the wavelength at a single spot only (such as in the plots above).

This can be seen by the naked eye in the following movies. They show on the top the display of the CCD spectrum analyzer, and on the lower part an interference pattern from an interferometer (Fizeau-type, just reflecting the slightly divergent beam off a glass slide). There is also a sound track from the noise monitor. The interferometer fringes are the vertical stripes, and the horizontal bands correspond to the transverse emission pattern of the diode (which arise from coherent superposition of supermodes in the broad ridge diode). Both structures jump when changing the diode current.

The left movie shows the spectrum and interference pattern in the region of 200-230mA. One can clearly see the effect of mode jumps and multimode operation on the interference pattern, which then blurs. The important thing to note is that all regions behave in a synchronized manner, ie., all fringes move in the same way. This means that in this region, there is only one supermode lasing, and all regions of the beam are coherent with respect to each other.

The right movie shows the spectrum and interference pattern at around 300mA, which is beyond the "stability border". We see that the situation is much less stable and often multimode operation occurs. The important thing to note is the various different bands tend not to be synchronized any more, and we see that the fringes in the three major transverse mode bands tend to jump independently. This means that several supermodes lase simultaneously, which are not mutually coherent with respect to each other.

Here a protope which also includes beam shaping with anamorphic prisms:


9) Results for the Opnext HL63133DG  170mW/638nm laser diode

This is a narrow-stripe single transverse mode diode, which is similar to but somewhat better then the proven HL6385 (and definitely more expensive). It runs moderately well in free-running mode, but runs great in ECDL mode; I used grating 3g and a aspheric collimator. Here a current/temperature plot with superimposed noise in grey:

We see that apart from the mode hopping zones, the diode runs single longitudinal mode practically everywhere! At 180mA an output of 100mW was achieved, and certainly more can be done but I didn't want to risk the expensive diode. As is looks so far, this is the best diode overall! Here a picture of nice clean, and stable single mode operation at 100mW:

Finally, here a plot of output power (blue= free running diode, red=ECDL with grating 3g):


10) Results for the Osram PL450 80mW/450nm laser diode (3.8mm case)

In free-running mode this diode is completely useless for SLM operation, and at first it seems to be not too great in ECDL mode either, but there is some hope that with careful optimization a decent SLM laser can be made.

Only in ECDL mode the spectrum is narrow enough such as to give a reasonable readout of the Burleigh WA2000 Wavemeter:

Bottom scope shows the single mode signal of the scanning Michelson interferometer inside the WA2000.

Here a comparison of output power at 448nm in free-running (blue) and ECDL mode (red):

Here a plot for my first configuration using an Axiz 445nm lens and grating Nr.3, and we see that SLM (cyan) arises only at very low powers:

However, this was for the first setup, and next I improved the back coupling of the grating by using a high quality collimator lens (from Marco Lauschmann). Also, a short resonator length was important. To achieve this, I placed the grating close to the diode mount, and in order for the beam to pass the mount, I attached a 45 degree dielectric mirror on it; see the pics:

 

With some careful tinkering with adjustments, more than 40mW in SLM operation could be achieved:

Here, as always, superimposed grey shows noise in the light output, and we see that in zones where there is no noise there is clean single mode operation.  I found the alignment to be very critical, the stability depends very much also on the collimator adjustemt, and it seems that this is sensitive to like 1/100 turn! This can probably pushed up even a bit more, but I didn't try more than 115mA in order not to risk the diode.

Here a brief movie (2MB) showing single mode operation from the CCD spectrum analyzer at around 40mW (mode jumps indicated by popping sounds):

The scanning interferometer shows a linewidth of less than approx. 10Mhz, so the coherence length should be at least several tens of meters:

Due to a foolish mistake, I eventually killed this diode. The second one proved to be similar, except that had a substantially higher efficiency:

Again, stable SLM operation beyond 40mW could be achieved after a very careful adjustment of the collimator:


11)Results for the Opnext HL45023TG 80mW/445nm laser diode

TBA


12) Results for the Opnext HL63603TG 120mW/638nm laser diode (3.8mm case)

I tested one of those and it ran very stable in ECDL operation. This diode is pretty cheap on ebay and also have a wavelength close to HeNe, a slight disadvantage is the 3.8mm housing. Here a graph of output power in free running and in ECDL mode with grating G3, both with Lauschmann collimator for 650nm (focus point at 2m):

This diode proved very stable in excess of 60mW! Probably part of the reason is a very short resonator, of about 1cm. This was possible by a construction using a 12mm brass tube mount for 3.8mm diodes sold on Laserfreak.de. It was embedded in one of my standard aluminum mounts, and the overall mechanical stability could probably be improved by a dedicated design, but for the sake of demonstration this would do:

The mode scan shows very good stability until approx 120mA:

Here a brief movie (1.7MB) showing single mode operation from the CCD spectrum analyzer at around 70mW (mode jumps indicated by popping sounds): 

 


13) Results for the Mitsubishi ML520G54/110mW 638nm laser diode

This diode has been a good surprise: I tested 3 samples and all ran very stable SLM in ECDL operation. They are pretty cheap on ebay and also have a wavelength close to HeNe, so they are a very good value. Here a graph of output power in free running and in ECDL mode with grating G3, both with Lauschmann collimator for 450nm (focus at 2m):

The mode scan looks pretty good:

I used this diode in a laser setup based on the commercial laser described here.
Here a brief movie (2.3MB) showing single mode operation from the CCD spectrum analyzer at around 60mW (mode jumps indicated by popping sounds):


14) Results for the Osram PL520 50mW/520nm laser diode (3.8mm case)

This is the first relatively cheap green diode and obviously I was eager to test it. Free running it is hopelessly running multimode, but in ECDL mode it runs very well and stable to 40mW and beyond. Here a graph of output power in free running and in ECDL mode with grating G3, both with Lauschmann collimator for 450nm (focus at 2m):

The first run was only moderately good:

It turned out that the diode runs very stable when using a 2400l/mm grating G6:

The major benefit is that due to the larger deflection angle, the resonator length can be minimzed to about 1.5cm.That was possible by using the 3.8mm diode mount made by Guido (a mechanical masterpiece, many thanks!), which has a bevelled corner so that the grating can be very close to the diode while the beam is not obstructed: 

Here a movie (2.7MB) showing single mode operation from the CCD spectrum analyzer at around 40mW (mode jumps indicated by popping sounds):


15) Results for the Mitsubishi LPC-836 300mW/655nm laser diode

This diode runs strongly multimode even in ECDL setup, and despite some tinkering, I didn't get it any better than that:

Probably the huge gain that these diode have makes the single longitudinal mode practically impossible, even at lowest currents. A pity, as these diodes have excellent beam data.


Stay tuned for further results.


Return to home page

Vers 3/1-2014