Erbium-Doped Fiber Amplifier & CW Laser
The goal of this project is to construct a high power fiber laser using double-clad fiber co-doped with erbium and ytterbium. Then to optimize the fiber laser, altering the length of the gain medium and the output coupling ratio to maximize output power.
- 1 Background
- 2 Fiber Laser Construction
- 3 Controlling the Pump Laser
- 4 Initial Results
- 5 Low Power Experimental Setup
- 6 Refined Experimental Setup
- 7 Single Longitudinal Mode Operation
- 8 Self-Heterodyne Linewidth Measurement
- 9 Single Longitudinal Mode Results
- 10 Future Improvements
- 11 Honors College Thesis
Optical fiber comes in two common types: single-mode and multimode, each with their own advantages and drawbacks. Single-mode fiber allows for higher beam quality with low propagation loss, but requires single-mode pump sources, which tend to be lower power and more expensive. Multimode fiber allows the use of higher power and less expensive multimode pump sources, but sacrifice beam quality and propagation loss in the process.
Double-clad fiber, however, offers the best of both fibers. Like its name suggest, double-clad fiber has two claddings, along with single-mode core. Double-clad fiber offers high beam quality with low propagation loss of the signal in the core, but supports higher power and less expensive multimode pumping of the inner cladding.
Fiber Laser Construction
Our double-clad fiber ring laser, like all ring lasers, uses a positive feedback loop into a gain medium to achieve lasing. A 10 W 975 nm multimode laser diode is coupled into the inner cladding of the active fiber us a multimode pump coupler. The active fiber is co-doped with erbium and ytterbium. The erbium absorbs the pump light and emits 1550 nm light through the core, while the ytterbium only serves to increase the efficiency of the energy transfer. The signal light propagates through the core and into a coupler, where the majority of the light is sent through the feedback loop. An optical isolator in the feedback loop ensures light only propagates in one direction to reduce any possible loss. The feedback light is then sent into another end of the multimode pump coupler, where it is coupled into the core of the erbium-ytterbium doped fiber (EYDF) where it causes stimulated emission. The remaining portion of the signal light that is not returned as feedback exits the system as the output of the laser.
Controlling the Pump Laser
To power our pump laser, we are using a laser diode driver which drives 0-15 A based on a supplied 0-10 V signal. To control the signal, we created a graphical user interface using Python, with the modules Tkinter and PySerial, as well as an Arduino Duemilanove. The GUI allows us to set a 0-8 V signal (so as not to surpass the 12 A current limit of the diode laser) with 8-bit resolution using a digital-to-analog converter. All of these components were then built into a box to allow for simple operation of the laser diode.
After constructing the fiber laser, using an 80/20 output coupler and 1.5 m of active fiber, we were able to measure the output of the laser in response to the pump laser power.
We discovered the laser had an output power of 140 mW with the pump laser at 10 W. But when the pump laser is at 7.1 W, the laser has an output power of 190 mW. We expect we are saturating the erbium-ytterbium co-doped fiber, but that would result in a gain curve which plateaus, not decreases. Investigating other high power lasers, we found a common loss mechanism is Stimulated Brillouin Scattering (SBS). SBS is a nonlinear loss mechanism that increases with pump power and temperature. This means that once we hit max gain on our EYDF, the increased effects of SBS force our laser's output power to decrease. To test this, we switched out the 80/20 2x1 output coupler with a 90/10 2x2 output coupler, which has an additional output for back-reflected light. If there was a significant amount of Stimulated Brillouin Scattering, the amount of light being back-reflected would match the drop in power. However, measuring the light out of the back-reflected endwe only detected 1.5 mW out of the 10% back-reflection end. Meaning there is only 15 mW of total back-reflection, and the loss cannot be due to SBS.
Now using the 90/10 output coupler, we also implemented a mode scrambler to increase the amount of pump light crossing the core in the active fiber. While the gain curve still has a similar shape, with the peak output power occurring near 7.1 W of pump power, the maximum output power improved drastically to over 1 W.
In an attempt to further increase the output power of the laser, we increased the size of the active fiber to 3 m. Our initial results were promising, measuring near 1.25 W of output power, however before we could characterize the output, the 10 W pump laser burned out.
Low Power Experimental Setup
After the 10 W pump laser burned out, we set up two 1 W diode lasers to pump our system. We set up one of the laser diodes on its own temperature controller and diode driver. For the other laser diode, we re-coded the Python GUI so it would only drive the 1.7 A needed to run the 1 W laser diode. The multimode pump coupler has two input ports, which we connected both pump lasers to, as can be seen below. We may investigate the effects of backwards pumping in the future.
The maximum output power measured was 297 mW at 2 W of pump power. When pumping near threshold, we noticed through the IR viewer that the second fusion splice pulses(the splice connecting the active fiber the dump WDM). Analyzing the output, we measured that the output also pulses with a frequency of 44 kHz. Although we have not yet been able to measure the amplitude of the pulsing, we have determined that our laser is self-pulsing. Self-pulsing is commonly caused by either low pump power or a long cavity length. When the pump is not able to maintain a complete population inversion, then the gain medium acts like a saturable absorber. When the cavity length is long, the light pulse will have more time until it is fed back into the gain medium, allowing the pump time to start saturating the gain medium again. Our laser has the right setup for self-pulsing, as the cavity is nearly 12.5 m long. However, once we pump the laser with more than 600 mW, the pulsing is suppressed. While an interesting result, self-pulsing should not be a problem in the future, as our setup can supply more than enough power to suppress the pulsing.
Refined Experimental Setup
To improve the laser, several new components were implemented into the fiber ring laser, including a new pump laser. The two 1 W 975 nm diodes were replaced with a 10 W 915 nm laser diode.
The setup works just as before, however a second WDM was included to dump any excess 915 nm pump light out of the system. The isolator was replaced with an optical circulator, which allows the implementation of a fiber Bragg grating (FBG). An FBG is a selective reflector which reflects particular wavelengths of light, and transmits all others. Thus, the linewidth is reduced, as only the light reflected by the FBG is amplified. Another advantage of the fiber Bragg grating is that it allows the laser to be tuned. By stretching the FBG, longer wavelengths are reflected. We were able to stretch the FBg and tune the output of the fiber laser from 1550 nm to 1560 nm. The experimental results for the refined setup can be found below.
The laser had a maximum output of 1.384 W and a measured linewidth of 0.63 nm. However, the monochromator used to measure the linewidth only has a resolution of around 0.5 nm, while the FBG has a bandwidth of 0.1 nm, so the linewidth of the laser is most likely near 0.1 nm as well. We also measured the beam quality factor () of 1.10, almost competely single-mode.
Single Longitudinal Mode Operation
Because of the large length of the fiber ring laser, there can be many different longitudinal modes which the system can lase in. Without the fiber Bragg grating, there are approximately 270,000 modes supported by the laser. Including the FBG cuts down the amount of modes significantly, however there are still about 430 modes supported. Therefore, to reduce the linewidth of the laser even further, the free spectral range (FSR) of the laser must be increased drastically.
For a ring cavity, the FSR is given by where c is the speed of light, n is the index of refraction of the optical fiber, and d is the length of the ring cavity. A shorter cavity will have a greater FSR, however it will have less gain as the active fiber will be shorter. Therefore to increase the FSR without decreasing the length of doped fiber, a multi-ring cavity design is used.
In a multi-ring cavity, small cavities are built into the large cavity to increase the FSR. The effective FSR of the cavity is equal to the least common multiple of the inner cavities' FSR and the large cavity's FSR. Thus, single longitudinal mode operation can be implemented for even a very large cavity through the use of several small cavities.
To construct the multi-ring cavity laser, 50/50 output couplers were incorporated into the ring laser. Two of the fiber ends were terminated at short distances to create small rings within the cavity to maximize the FSR.
To achieve single longitudinal mode operation, the polarization must also be taken into account. Each different mode has a different polarization, so for true single-mode operation, there can be only one polarization state allowed. To do so, an in-fiber polarizer was built into the system. Before the polarizer, a polarization controller was used to change the polarization of the laser to match that of the polarizer, so that a maximum amount of light could be put through the laser. As well, the light which travels through the inner loops must match the polarization of the light in the large loop when it is combined. Therefore, polarization controllers were put into each inner cavity so that all light would have the same polarization. The layout of the multi-ring cavity fiber laser can be seen below.
Self-Heterodyne Linewidth Measurement
Because of the resolution limit of the monochromator, the linewidth of the laser could not be measured accurately. Therefore, a self-heterodyne interferometer was made for linewidth measurements.
The laser is split into two paths using a 50/50 coupler. The first path travels through an accous-optic modulator, which adds 80 MHz to the frequency of the light. The second path travels through a long piece of fiber, 20.56 km in this case, which is longer than the coherence length. Therefore, when the two signals are re-combined, they interfere as if they were two different sources. There is also a polarization controller to change the polarization of one leg to match the other, so that they can interfere correctly. The two paths are combined with another 50/50 coupler, and are put onto a photodetector. Because of the 80 MHz added to one path, the photodetector reads the beat frequency. Taking the Fourier transform of the signal creates a signal in frequency space, centered around 80 MHz. The full width at half max of the Fourier transform is the FWHM of the laser. The resolution of the self-heterodyne interferometer is limited by the delay in the second path. With the 20.56 km of fiber, the self-heterodyne interferometer has a resolution limit of approximately 10 kHz.
Single Longitudinal Mode Results
The laser was characterized after each inner cavity was added to the original configuration. The fiber laser was put into a Fabry-Perot cavity, where the transmission was recorded. As well, the power was characterized, and a self-heterodyne linewidth measurement was taken.
Without any additional cavities, the single cavity fiber laser had an output power of 900 mW. The transmission through the Fabry-Perot for the single cavity laser was very noisy. Because of the many different modes being lased, the output rarely appeared periodic. The linewidth could not be measured using the self-heterodyne method, however it is safe to assume it would be the same as the fiber Bragg grating. Therefore the linewidth was 12.5 GHz, or 0.1 nm. Thus the laser had a coherence length of only 7.63 mm.
With one additional cavity, the output power of the fiber ring laser dropped to 351 mW. However, the transmission response through the Fabry-Perot improved dramatically. A much more periodic structure appeared, and only a few side modes appeared to exist. The linewidth also drastically improved, and was measured to be 240 kHZ, or 1.92 fm. And by simply adding one inner cavity, the coherence length was improved to 398 m.
With two additional inner cavities, the output power increased to 383 mW. This was quite a confusing result, as adding inner cavities had always reduced the output power. One possible cause is that with one ring, some modes still existed which had a different polarization than the in-fiber polarizer. By adding a second ring, these modes were eliminated, so that only modes with the correct polarization existed, and these modes had more stimulated emission than before. The transmission response through the Fabry-Perot improved as well, as even fewer side modes were observed. The linewidth was measured to be 48 kHz, or 385 am, meaning the laser had a coherence length of 1.99 km.
With three additional inner cavities, the output power decreased to 302 mW. The transmission response through the Fabry-Perot improved so that all side modes were suppressed, and the output appeared to be single mode. As well, the linewidth improved as well, and was measured to be 11 kHz, or 88.2 am. Thus, the laser had a coherence length of 8.68 km.
To optimize the fiber laser, the output power could be increased through several methods. Using RP Fiber Power, the optimal length of EYDF for 10 W 915 nm pump power was determined to be ~12 m. However, for 10 W 975 nm pump power, it is only ~ 7 m. As well, different output couplers could be used to make the inner cavities. 50/50 couplers were originally used to make the inner cavities because the paper which described the process used 50/50 couplers. However, it is possible that 99/1 coupler could achieve the same increase in FSR, while reducing the output power by a smaller amount. As well, the optimal output coupling ratio of the fiber laser could have been changed by the addition of the inner cavities. In addition, tuning of the fiber laser could be explored using an uncoated fiber Bragg grating, rather than the coated FBG used, by stretching the FBG to change the reflected wavelength. Finally, while the inner cavities have limited the output to only a single-longitudinal, the fiber laser is still very unstable, and methods should be explored to increase stability. The most likely cause is the splices of the double-clad fibers, which were made with a machine not designed to splice double-clad fiber. Therefore, there is most likely some heating in these splices which causes the output of the laser to be unstable.