Case studies > EDFA for long-wavelength signal

Case Study: Erbium-doped Fiber Amplifier for a Long-Wavelength Signal

Key questions:

Why is ASE a substantial challenge for the high-gain amplification of a long-wavelength signal?
What is the effect of the fiber length?
How can the situation be optimized?

Design Goal

We want to design an erbium-doped fiber amplifier which can amplify a weak input signal (100 μW, super-Gaussian spectrum with 5 nm bandwidth) at a relatively long wavelength of 1580 nm, where the erbium gain is well below its maximum. The signal output power should be around 100 mW.

Initial Attempt

We first try with a single amplifier stage.

We use the RP Fiber Power software. In its Power Form “Fiber amplifier for continuous-wave signals”, we easily enter the parameters of signal input and amplifier:

Power Form inputs for a single stage Er-doped fiber amplifier. — Figure 1: Part of the Power Form, where we specify details of the input signals and the amplifier stages. In the bottom area, you also see some outputs.

We choose the commercial fiber “Liekki Er40-4-125” (as one of many dozens for which we have data) and have to decide on a fiber length and the pump power at 980 nm. We find that a length around 2.5 m is sufficient to absorb most of the pump power.

We then soon encounter a serious problem: We get only a quite low signal power out. For example, for 100 mW pump power in backward direction we get only 5.46 mW signal out, while producing 90.8 mW amplified spontaneous emission (ASE) in forward direction and 47.7 mW ASE in backward direction:

powers of signal, pump and ASE as well as excitation level inside the fiber — Figure 2: Powers of signal, backward pump, forward and backward ASE along with the erbium excitation level versus position in the fiber.

Essentially, the problem is that the amplifier gain is highest around 1530–1560 nm, while at 1580 nm (central wavelength of our signal) we have substantially less gain (see Figure 3). So we produce a lot of ASE at these shorter wavelengths before we reach a sufficiently high signal gain.

spectrum of ASE powers with signal-pass gain spectrum — Figure 3: The output spectra of forward and backward ASE, and the single-pass gain spectrum. The spectral region of the signal is indicated in blue.

The first attempt to solve this problem could be changing the fiber length and pump power. We use one of the helpful variation diagrams in the Power Form, where we can, for example, vary the fiber length in some range for a given pump power. That way, we quickly find that we can indeed achieve our goal of 100 mW signal output power, but with a fiber length of 8 m and a pump power as high as 1 W! This is highly inefficient, with more than half of the pump power converted to ASE rather than to signal power. Changing the pumping direction and fiber type also do not result in significant improvements.

Further, it may be rather surprising that a fiber length of 8 m is better, as we already got most of the pump light absorbed in 2.5 m. So we inspect that situation (8 m fiber length, 1 W pump power):

powers and excitation levels vs. position in 8 m long fiber — Figure 4: Optical powers in the 8 m long amplifier with 1 W pump power.

Although the pump power is largely exhausted already in the middle of the fiber, there is still substantial erbium excitation and signal amplification in the left half – but why? This is because backward ASE is reabsorbed there, now acting as pump light for the still longer-wavelength signal! (For the long signal wavelength, there is still some positive gain despite the moderate erbium excitation level, while shorter-wavelength ASE is reabsorbed.)

So we have a solution for our design goal, but one which works in a somewhat strange and not fully satisfactory way. However, there is a better solution:

Using a Second Amplifier Stage

We now realize the amplifier with two consecutive stages. That gives us the opportunity to suppress ASE with a bandpass filter between the stages: we prevent most of the ASE light generated in the first stage from entering the second stage, where it would be amplified further.

This is easy to try with the Power Form:

Enable the second stage and set its parameters (same fiber type, same length (now back to 2.5 m).
For the first stage, we can now turn the pump power down to 30 mW, as this mainly means less ASE output while not much reducing the signal power.
At the end of the first stage, we insert a bandpass filter with a Gaussian transmission curve. See the Losses at output input of the first stage, where we define a wavelength-dependent loss function (variable l for wavelength):

bandpass filter as a Power Form input — Figure 5: Entering a wavelength-dependent expression to represent a bandpass filter between the two stages.

For the second stage, a pump power of 190 mW turns out to be sufficient, as now we don't have serious gain saturation by ASE.

This results in the wanted signal output power of 104 mW, with a total fiber length of 5 m and a combined pump power of only 220 mW (rather than the 8 m, 1 W pumped single stage amplifier). Further, only 3.6 mW of forward ASE and 2.1 mW of backward ASE are generated in the second stage (in the first stage: 5.5 mW and 2.1 mW).

We can also inspect the evolution of signal power on a logarithmic scale:

evolution of signal power over the entire two stage amplifier — Figure 6: Evolution of signal power in the dual-stage amplifier. We see the gain of each stage and the coupling losses at each of the three interfaces. (In the middle, we have moderate signal power losses at the bandpass filter.)

See also the ASE output spectra of the first stage, with and without the bandpass filter:

spectrum of ASE powers, showing filtered and unfiltered powers — Figure 7: The ASE output spectra of the first amplifier stage (with and without filter).

Although the bandpass filter does not remove a very high ASE power, it suppresses the further strong amplification of the short-wavelength ASE while causing only a 0.37-dB loss for the signal.

Conclusions

The RP Fiber Power software is an invaluable tool for such work – very powerful and at the same time pretty easy to use!

You can learn various things from this study:

Particularly when trying to strongly amplify signals away from the wavelength of maximum gain, we can run into limitations by ASE getting too strong.
To some extent, we can sometimes mitigate that by optimizing the fiber length, which affects the shape of the net gain spectrum.
With a two-stage amplifier design and intermediate spectral filtering, we can get substantially further.

With a suitable simulation software, you can easily analyze such aspects and optimize your amplifier design.

See also our encyclopedia articles on erbium-doped fiber amplifiers and amplified spontaneous emission.