Chromatic dispersion (CD) and polarization mode dispersion (PMD) are the two dispersion mechanisms that limit how far a high-speed signal can travel down a fiber before it becomes unreadable. Both mechanisms do fundamentally the same thing to a signal: they spread a sharp pulse of light out in time as it travels, so that neighboring pulses start to overlap and the receiver can no longer tell where one bit ends and the next begins. Where they differ is in cause, behavior, and how predictable each one is, and understanding those differences is the core of the CD, PMD, and SA knowledge area in this certification.
Chromatic dispersion happens because light is never perfectly one wavelength. Every optical source, even a narrow-linewidth laser, emits a small spread of wavelengths, and different wavelengths travel through glass fiber at slightly different speeds. That speed difference, accumulated over kilometers of fiber, causes the faster and slower wavelength components of a single pulse to arrive at slightly different times, spreading the pulse. CD is a deterministic, well-behaved property: for a given fiber type at a given wavelength, the CD coefficient is stable, predictable, and essentially the same every time it is measured, which is why it can be compensated for with a high degree of confidence.
Polarization mode dispersion happens for a completely different reason. Ideal singlemode fiber has a perfectly circular, symmetric core, in which light polarized in any direction travels at the same speed. Real fiber is never perfectly circular or symmetric. Manufacturing imperfections, cabling stress, bending, and temperature changes all introduce tiny amounts of birefringence, meaning the fiber's core behaves slightly differently depending on the polarization state of the light passing through it. Light entering the fiber splits its energy across two polarization axes that now travel at very slightly different speeds, arriving at the far end with a time offset called differential group delay (DGD). Because the physical stresses causing this birefringence vary randomly along the fiber's length and change with temperature and mechanical disturbance, PMD is a statistical, not deterministic, quantity, and that distinction drives everything about how it is tested and how much it can be trusted from one measurement to the next.
What Causes Chromatic Dispersion
CD in a singlemode fiber is the combination of two effects: material dispersion, caused by the way the glass itself refracts different wavelengths at different speeds, and waveguide dispersion, caused by the way the fiber's core and cladding geometry guides different wavelengths slightly differently. These two effects partially cancel each other out at a specific wavelength called the zero-dispersion wavelength, which for standard singlemode fiber compliant with ITU-T G.652 sits around 1310 nm. This is precisely why older 1310 nm systems experienced minimal CD, while systems that moved to the lower-attenuation 1550 nm window for long-haul reach ran into a meaningful CD coefficient, typically around 16 to 18 ps/(nm·km) for standard G.652 fiber at 1550 nm.
Fiber manufacturers responded to this problem by engineering the waveguide dispersion component to shift the zero-dispersion point. Dispersion-shifted fiber (ITU-T G.653) moves the zero-dispersion wavelength into the 1550 nm window, minimizing CD exactly where long-haul systems wanted to operate, though this created its own problems with nonlinear effects in WDM systems because operating near zero dispersion makes channels more susceptible to four-wave mixing. Non-zero dispersion-shifted fiber (ITU-T G.655) strikes a middle ground, maintaining a small but nonzero amount of dispersion across the operating band specifically to suppress those nonlinear effects while still keeping CD low enough to manage.
What Causes Polarization Mode Dispersion
PMD arises from the same category of physical imperfection regardless of its specific source: anything that makes the fiber's cross-section slightly non-circular or introduces asymmetric stress creates birefringence. During manufacturing, even high-quality fiber has some residual ellipticity in the core from the draw process. During cabling and installation, bending, twisting, lateral pressure from other cables or hardware, and temperature-driven expansion and contraction of buffer coatings and cable jackets all add mechanical stress that changes the fiber's birefringence along its length. Because these stress points occur somewhat randomly along a real cable route and their effects combine through a statistical process called mode coupling, the total PMD of a long fiber span does not simply add up linearly with length the way attenuation does. Instead, PMD accumulates with the square root of length, which is why the PMD coefficient for fiber is expressed in picoseconds per root-kilometer rather than picoseconds per kilometer.
This square-root scaling is a direct mathematical consequence of PMD's random-walk nature, and it means that a fiber with a given PMD coefficient becomes proportionally more tolerable on longer spans than a naive linear scaling would suggest, but it also means the exact value depends on the specific pattern of stress points along that specific fiber, which is unique to each installed cable and not predictable from fiber type alone the way CD is.
Specifications and Tolerance by Network Type
Because CD and PMD tolerance both scale with bit rate, but scale differently as modulation formats and forward error correction change, the maximum tolerated dispersion for a given network varies considerably depending on exactly what system is running on the fiber. A common rule of thumb is that the maximum tolerable dispersion-induced pulse spreading should stay within roughly ten percent of the bit period, and system vendors publish specific tolerance numbers derived from that principle combined with their particular receiver design and forward error correction strength.
For a 10 Gbps SONET/SDH system at OC-192/STM-64, a commonly cited maximum CD tolerance is around 1000 to 1200 ps/nm, corresponding to reach of roughly 60 to 80 kilometers of standard G.652 fiber before compensation becomes necessary. Moving to 40 Gbps reduces that tolerance by roughly a factor of sixteen, often down to somewhere in the range of a few hundred ps/nm depending on modulation format, which can bring the uncompensated reach down to just a handful of kilometers. Advanced coherent modulation formats used at 100 Gbps and beyond behave very differently: because coherent receivers use digital signal processing to electronically compensate for chromatic dispersion after detection, modern 100G and higher coherent systems can tolerate enormous accumulated CD, sometimes tens of thousands of ps/nm, effectively removing CD as a practical limiting factor on many coherent long-haul routes even though it remains a limiting factor on older, simpler systems running on the very same fiber.
PMD tolerance follows a similar bit-rate-driven pattern but with tighter margins because the acceptable outage probability for PMD-induced errors is extremely low. For 10 Gbps SONET/SDH systems, a commonly cited maximum mean PMD (DGD) value is around 10 ps, derived from keeping instantaneous DGD below roughly 30 ps for 99.999 percent of the time given PMD's statistically random behavior. For 10 Gigabit Ethernet, which typically requires much higher reliability, the accepted mean PMD tolerance is tighter still, often cited around 5 ps. Moving to 40 Gbps tightens PMD tolerance to roughly 2 to 2.5 ps mean DGD in many system specifications, and 100 Gbps coherent systems, like their handling of CD, benefit from digital signal processing that can compensate for a meaningful amount of PMD electronically, though PMD compensation in the digital domain is generally less complete than CD compensation, so PMD remains a real limiting factor even on modern coherent systems in ways CD often does not.
Test Methods for CD and PMD
Chromatic dispersion is measured in the field using one of a small number of standardized methods, each defined by its own TIA/EIA test procedure. The pulse-delay method sends short pulses at multiple wavelengths and directly measures the arrival time differences. The modulated phase-shift method modulates a source at a fixed frequency across multiple wavelengths and measures the resulting phase shift at the receiver, from which the group delay and dispersion coefficient are calculated. The differential phase-shift method is a related variant used especially for shorter fiber lengths where direct pulse-delay timing becomes less practical. All three methods ultimately produce the same output: a curve of relative group delay versus wavelength, from which the CD coefficient in ps/(nm·km) is derived as the slope of that curve, and because this behavior is deterministic, a single well-executed measurement is generally trustworthy and repeatable.
PMD measurement is fundamentally different because it is measuring a statistical quantity. Common test methods include the fixed analyzer method, which scans wavelength while monitoring the output polarization state and derives PMD from the resulting spectral pattern, and the Jones matrix eigenanalysis method, which directly measures the polarization transfer function of the fiber at multiple wavelengths and calculates DGD from the eigenvalues of that transfer function. Both approaches ultimately compute PMD as a mean DGD value averaged across a defined wavelength range, typically spanning tens of nanometers around the wavelength of interest, specifically because any single-wavelength DGD reading is just one random sample of a statistically varying quantity, and averaging across wavelength gives a more representative and stable estimate of the fiber's overall PMD.
Compensating for CD
Because CD is deterministic and predictable, it is the more straightforward of the two mechanisms to compensate for. Dispersion-compensating fiber (DCF) is a specialty fiber with a large negative dispersion coefficient, spliced or connectorized into the link in a length calculated to cancel out the accumulated positive dispersion of the transmission fiber preceding it; this was the dominant compensation method for years on 10 Gbps and 40 Gbps systems. Fiber Bragg grating-based compensators achieve a similar result using a different physical mechanism, reflecting different wavelengths with different delays to flatten out the accumulated dispersion, generally in a smaller physical footprint than a spool of DCF. Tunable dispersion compensators allow the amount of compensation to be adjusted electronically or mechanically after installation, which matters on networks where the exact path length or fiber type a signal will traverse is not fixed. Modern coherent transceivers add a further option: digital signal processing in the receiver can electronically compensate for large amounts of accumulated CD after the optical signal has already been converted to the electrical domain, which is why many current 100G and higher coherent systems require little or no optical-domain CD compensation hardware at all, in sharp contrast to the compensation-heavy designs common on older high-speed non-coherent systems.
Why PMD Testing Shows More Variability Than CD Testing
CD testing on a given fiber, repeated under different conditions, produces essentially the same result every time, because CD depends on the glass composition and waveguide geometry of the fiber, both of which are stable physical properties that do not meaningfully change with temperature, time, or handling on any timescale relevant to field testing. PMD testing on the same fiber, repeated under different conditions, can produce noticeably different results from one measurement to the next, and this is not a sign of instrument error or poor technique; it is an accurate reflection of PMD's underlying physical nature.
Because PMD arises from birefringence caused by mechanical stress and temperature, and because the specific pattern of that stress along a real installed fiber shifts with temperature changes, mechanical disturbance, and even the passage of time, the instantaneous DGD of a fiber measured at one moment can differ from the DGD measured an hour later or a day later under different environmental conditions, even though nothing about the fiber's fundamental construction has changed. This is why PMD is reported and specified as a mean or statistical value rather than a single fixed number, and why proper PMD test methods average across a wavelength range rather than relying on one measurement at one wavelength. A single PMD reading taken in isolation, without context about measurement conditions or repeatability, tells a technician much less than a single CD reading does, and understanding this asymmetry between the two mechanisms is one of the most consistently tested concepts in the CFOS/FC exam.