Transmission Basics  |  Components  |  Applying Fiber Optic Technology  |  System Performance  

Multichannel CATV Links and Backreflection

The general knowledge has been that backreflection hurt the performance of a link because the reflected light gets into the laser cavity, disturbs the standing optical wave, and creates noise. Some lasers are very susceptible to backreflection due to the design of the laser chip itself. Most often the determining factor is how tightly the fiber is coupled to the laser chip. In a low power laser potentially only 5-10% of the laser power is coupled into the fiber. This means that only 5-10% of the backreflection would be coupled to the laser chip, making the laser relatively immune to backreflections. On the other hand, a high-power laser may have 50-70% of the laser chip output coupled to the fiber. This means that 50-70 % of the backreflection is coupled to the laser cavity. "Laser Backreflection — The Bane of Good Performance" describes the overall impact on backreflection on lasers. This article discusses the problem as it specifically applies to multichannel CATV links. The type of signal being carried on the fiber optic link has a great deal to do with how susceptible the link might be to backreflections. The most sensitive case, not surprisingly, is broadband analog, such as multichannel CATV links. Let's first look at what happens to a fiber optic link that carries a single tone when it is exposed to varying degrees of backreflection. Our test case will be a broadband analog fiber optic link incorporating an isolated DFB laser that is transmitting a single 100 MHz tone. Figure 1A shows the single tone after it is transmitted through the link. The noise floor is about 76 dB below the carrier and the 2nd and 3rd harmonics are 56 dB and 68 dB down. Figure 1B shows a single weak backreflection. The noise floor is now about 75 dB down and the relative amplitude of the 3rd harmonic has increased by 3 dB. In Figure 1C, two strong backreflections have been added. The noise floor has now increased by about 14 dB and the 2nd and 3rd harmonics are now 44 dB and 60 dB down. Notice the multiple sidebands that have formed around the fundamental and especially the 2nd harmonic. Figure 1D looks at the same basic data over a wider frequency range. The periodic nature of the noise floor is a sure tip off that backreflections are affecting the system. The spacing between the peaks in the noise floor is related to the physical distance, and thus the time delay, between the reflecting elements in the system.

Figure 1 - Backreflection in a Single Channel Broadband Analog Fiber Optic Link

Now let's look at a broadband analog optic link that is carrying ten CATV signals to see the effects of backreflection. Figure 2A shows the signal through the link with minimal backreflections. The CNR is about 65 dB and the 2nd harmonic is about 60 dB down. In Figure 2B, a single backreflection is added. The noise floor increases 5 to 10 dB. In Figure 2C, two moderate backreflections are added to the optical path. The noise floor has increased dramatically to reduce the CNR from 65 dB originally to about 28 dB. In Figure 2D, two strong backreflections are added to the optical path. The desired signal is now almost totally lost in the noise floor with a CNR of less than 10 dB.

Figure 2 - Backreflection in a 10 Channel Broadband Analog Fiber Optic Link

Even though we used an isolated laser, multiple backreflections had a very dramatic effect on the quality of the signal transported over the fiber optic link. The noise increased significantly. Second, the multiple backreflections changed the distortion characteristics of the fiber optic link. Both the 2nd and the 3rd harmonics were affected, but the 2nd harmonic showed the larger effect. The bottom line is that backreflections anywhere in a fiber optic system hurt performance. The only question is how much? For analog systems, a good rule of thumb is to ensure that all backreflections in the fiber path are more than 55 dB down. For digital systems, backreflections should be kept at least 45 dB down. Higher levels will lead to unpredictable system performance.


Backreflections can be observed by monitoring the photodiode servo-loop for disturbances. To do this, place the end of the laser pigtail in glycerin, which will eliminate virtually all backreflections. Then note the output of the servo loop at that time. Afterwards, connect the laser pigtail to the system. Any significant perturbations noted are backreflections of sufficient amplitude to disturb the standing wave in the laser cavity. This directly observes laser backreflections. A less direct test for laser backreflections involves testing at frequencies where backreflections will occur at an exact multiple of the bit time. Basically, this procedure calculates the round trip transit time to the potential reflection interfaces in the system. It is generally easiest to measure the spacing between the high interference points when using this method. Often, the laser pigtail is made very short so that the first reflection occurs at a frequency higher than any frequency being transmitted by the system. However, longer fiber segments in the system will yield a low fundamental interference frequency and harmonics that will clutter the spectrum. Designing laser-based systems for low backreflections remains the only practical strategy.