Chromatic Dispersion (CD) vs PMD: What They Are and When to Test Fibre Dispersion
A fibre link can have plenty of optical power and a perfect loss budget and still fail at 10 Gbit/s and above. The usual culprit is dispersion: the spreading of light pulses as they travel down the fibre until neighbouring bits overlap and the receiver can no longer tell a 1 from a 0. There are two kinds that matter in single-mode fibre: chromatic dispersion (CD) and polarisation mode dispersion (PMD). This guide explains both, the limits for each bit rate, and when and how to test them before upgrading a route.
What is dispersion?
A laser sends data as short pulses of light. Over distance each pulse gets wider, because different parts of it arrive at slightly different times. At low bit rates the pulses are far apart, so a little spreading does no harm. At 10G, 40G and above, the pulses are much closer together, so the same spreading makes them overlap and causes bit errors. Dispersion is measured in picoseconds (ps), and the question is always how big it is compared with the bit period: 400 ps at 2.5 Gbit/s but only 100 ps at 10 Gbit/s.
CD vs PMD at a glance
| Chromatic dispersion (CD) | Polarisation mode dispersion (PMD) | |
|---|---|---|
| Cause | Different wavelengths (colours) in the laser pulse travel at different speeds in the glass | The two polarisation states of light travel at slightly different speeds because the fibre core is not perfectly round or is under stress |
| Unit | ps/nm for a link; ps/(nm·km) for the fibre coefficient | ps (differential group delay) for a link; ps/√km for the coefficient |
| Grows with length | In direct proportion to length | With the square root of length |
| Stable over time? | Yes, predictable and fixed for a given fibre and wavelength | No, it changes randomly with temperature, wind and handling, so it is treated statistically |
| Can it be compensated? | Yes, with dispersion-compensating fibre or modules, or electronically in coherent receivers | Hard to compensate optically; coherent receivers handle it electronically |
| Test standard | IEC 60793-1-42, TIA FOTP-175 | IEC 60793-1-48, TIA FOTP-124A, ITU-T G.650.2 |
Chromatic dispersion (CD)
No laser emits a single pure wavelength; every pulse contains a narrow spread of colours. In glass, different wavelengths travel at slightly different speeds, so the colours separate over distance and the pulse widens. This is chromatic dispersion.
Standard ITU-T G.652.D single-mode fibre has its zero-dispersion wavelength between 1300 and 1324 nm, so CD is close to zero in the O-band. At 1550 nm, where loss is lowest, CD is about 17 ps/(nm·km). Multiply by the length to get the total for a link: an 80 km route at 1550 nm has about 17 × 80 = 1,360 ps/nm. Non-zero dispersion-shifted fibre (ITU-T G.655) was designed with much lower CD at 1550 nm for long DWDM routes.
The tolerance of a transmitter falls with the square of the bit rate, so going four times faster cuts the tolerated CD by about sixteen times:
| Bit rate (NRZ, no compensation) | Approximate CD tolerance | Approximate reach on G.652 fibre at 1550 nm |
|---|---|---|
| 2.5 Gbit/s | About 16,000 ps/nm | Roughly 900 km |
| 10 Gbit/s | About 1,000 ps/nm | Roughly 60 km |
| 40 Gbit/s | About 60 ps/nm | Only a few km |
| 100G/400G coherent | Tens of thousands of ps/nm, handled in DSP | Not CD-limited on normal routes |
These are approximate figures for directly modulated NRZ optics; always use the CD tolerance on your transceiver's data sheet. CD is stable and predictable, so it can be corrected with dispersion-compensating fibre or fibre Bragg grating modules at amplifier sites, and modern coherent 100G and 400G systems remove it digitally in the receiver.
Polarisation mode dispersion (PMD)
Light in single-mode fibre travels in two polarisation states at right angles to each other. In a perfect, perfectly round fibre they travel at the same speed. Real fibre cores are very slightly oval, and bending, twisting, pressure and temperature add stress, so one polarisation runs a little faster than the other. The time difference between them is the differential group delay (DGD), and the average DGD over time and wavelength is the link's PMD.
PMD is random: it changes as the cable warms, cools, moves in the wind or is handled, so it is described statistically and grows with the square root of length. The fibre's quality is given as a PMD coefficient in ps/√km. ITU-T G.652.D specifies a link design value (PMDQ) of no more than 0.20 ps/√km, and modern fibre is usually well below that. Fibre made before the mid-1990s can be several times higher, which is why older routes are the first to fail during upgrades.
| PMD coefficient (ps/√km) | 25 km | 100 km | 400 km |
|---|---|---|---|
| 0.1 (modern G.652.D, typical) | 0.5 ps | 1.0 ps | 2.0 ps |
| 0.2 (G.652.D link design limit) | 1.0 ps | 2.0 ps | 4.0 ps |
| 0.5 (older fibre) | 2.5 ps | 5.0 ps | 10.0 ps |
| 1.0 (poor legacy fibre) | 5.0 ps | 10.0 ps | 20.0 ps |
A common rule for NRZ systems is that the mean link PMD should stay below about 10% of the bit period, so that the rare moments when DGD peaks do not cause outages:
| Bit rate | Bit period | Typical maximum mean link PMD |
|---|---|---|
| 2.5 Gbit/s | 400 ps | about 40 ps |
| 10 Gbit/s | 100 ps | about 10 ps |
| 40 Gbit/s (NRZ) | 25 ps | about 2.5 ps |
| 100G and 400G coherent | About 30 ps per symbol at 32 GBaud | Much higher; the receiver DSP compensates tens of ps |
So a 100 km route built with older fibre at 1.0 ps/√km already sits at the 10 Gbit/s limit, while the same route on modern G.652.D fibre has a large margin. Coherent 100G and 400G receivers tolerate far more PMD, but older 10G and 40G services, and many access and metro links, still depend on low PMD.
When should you test CD and PMD?
- Before upgrading a route to 10G, 40G or 100G, especially fibre that is more than 15 to 20 years old or whose records are lost.
- At acceptance of new long-haul or metro routes, so the fibre owner has baseline figures for future leasing and upgrades.
- Before adding DWDM channels, since the longest-wavelength and fastest channels are the most sensitive.
- When a link has good power and loss but still shows bit errors, or errors that come and go with weather on aerial routes.
- When leasing dark fibre, to choose the best fibres in a cable for high-speed services.
How to test CD and PMD: step by step
Dispersion testing is the last stage of fibre characterisation. Do the basic tests first, because a dirty connector or a bad splice gives misleading dispersion readings.
| Step | What to do | Tool |
|---|---|---|
| 1. Clean and inspect | Every connector in the path, at both ends; a dirty or damaged end-face spoils every later reading | Fibre inspection probe |
| 2. Confirm loss and continuity | Measure end-to-end loss and check the link against its loss budget | Light source and power meter (OLTS) |
| 3. Characterise with an OTDR | Find splices, bends, high-loss events and reflections, and confirm length | OTDR |
| 4. Measure CD | Record total dispersion (ps/nm) and the zero-dispersion wavelength and slope across the band you plan to use | CD analyser |
| 5. Measure PMD | Record mean DGD (ps) and the PMD coefficient (ps/√km); test each fibre you will light, as fibres in the same cable can differ | PMD analyser |
| 6. Compare and report | Compare the results with the transceiver or system limits and keep the report for the route records | Test platform software |
For PMD, the interferometric method (TIA FOTP-124A, also in IEC 60793-1-48) is widely used in the field because it is fast and copes with fibre that moves during the test. The EXFO FTB-5500B PMD analyzer uses this method, completes a measurement in under 5 seconds for any PMD range, and can test through EDFA amplifiers, so long amplified routes do not have to be broken into sections. It runs on EXFO's FTB platforms such as the FTB-4 Pro and FTB-2 Pro, and pairs with EXFO's FTB-5800 CD analyzer for a full dispersion test from one platform.
For the earlier steps, use a fibre inspection probe, a loss test set such as the EXFO FOT-300, and an OTDR such as the FTBx-750C metro and long-haul module or the MaxTester 715B. Our guides to connector cleaning, insertion-loss testing and OTDR testing cover these in detail.
Reading the results
- CD: compare the total ps/nm at your operating wavelength with the transceiver's tolerance. If it is over, use dispersion compensation, move to a wavelength with lower CD, or use coherent optics.
- PMD: compare the mean DGD in ps with the bit-rate limit above, and the coefficient in ps/√km with the fibre specification. Test each fibre: in an old cable some fibres may pass at 10G while others do not.
- Keep the reports: results tied to fibre numbers and dates make later upgrades and fault finding much quicker.
Troubleshooting dispersion problems
| Symptom | Possible dispersion cause | What to check |
|---|---|---|
| 10G link works at short length but errors on a long route | Total CD beyond the optic's tolerance | CD in ps/nm against the transceiver data sheet; consider dispersion compensation or a different optic |
| Errors come and go with the time of day or weather on an aerial route | PMD changing as the cable moves and heats up | PMD on that fibre; try another fibre in the cable with lower PMD |
| One fibre in a cable fails at 10G but others pass | High PMD on that particular fibre | PMD of each fibre; reserve the best fibres for high-speed services |
| Link fails after a 2.5G to 10G upgrade on old fibre | PMD or CD that was fine at 2.5G is now over the 10G limit | CD and PMD of the route before the upgrade |
| Optical power and loss are fine but there are still bit errors | Dispersion, not loss, is the limit | CD and PMD, then a service test |
Keeping dispersion low on new routes
The easiest way to avoid PMD problems is to start with low-PMD fibre and install it carefully. Litech manufactures single-mode G.652.D duct, direct-buried and ADSS optical fibre cables in Pakistan. Respecting the cable's bend radius, pulling tension and crush limits avoids adding the stress that raises PMD, and choosing the right fibre type is covered in G.652.D vs G.657. For DWDM routes, see our guide to CWDM vs DWDM.
Key takeaway: CD is predictable, grows with length and is about 17 ps/(nm·km) at 1550 nm on G.652.D fibre; PMD is random, grows with the square root of length, and should stay below about 10% of the bit period. Test both before upgrading old fibre to 10G or faster, after cleaning, loss and OTDR checks.
Frequently asked questions
Does dispersion matter at 1 Gbit/s?
Rarely. At 1 Gbit/s the bit period is 1,000 ps, so typical CD and PMD on access and metro links are far below the limits. Dispersion testing becomes important from 10 Gbit/s upwards and on long routes.
Can an OTDR measure PMD?
A standard OTDR measures loss, length and events, not PMD. You need a PMD analyser for an accurate PMD figure. Some OTDR-based tools estimate CD, but a dedicated analyser is the reference method.
Why is PMD different each time I test?
Because PMD is random and changes with temperature and movement, especially on aerial cable. Small differences between tests are normal; the mean value over wavelength is what you compare with the limit.
Does multimode fibre have PMD?
Multimode links are limited mainly by modal dispersion, which is set by the fibre's bandwidth grade (OM3, OM4, OM5). CD and PMD testing applies to single-mode fibre.
Is dispersion a problem for 100G and 400G?
Coherent 100G and 400G systems compensate CD and PMD electronically, so they tolerate far more. Non-coherent high-speed optics, such as short-reach 100G PAM4 transceivers, are still sensitive, so check the optic's specification.
Need dispersion test equipment or fibre cable?
Litech Pakistan (Private) Limited supplies EXFO test equipment, including PMD analysers, OTDRs, loss test sets, inspection probes and test platforms, and manufactures single-mode optical fibre cables in Pakistan. Message us on WhatsApp +92 313 7776766 with your route length, fibre age and target bit rate for advice and a quotation.
Sources: ITU-T G.652, Characteristics of a single-mode optical fibre and cable; ITU-T G.655 (non-zero dispersion-shifted fibre); ITU-T G.650.2 (test methods for polarisation-related attributes); IEC 60793-1-42 (chromatic dispersion) and IEC 60793-1-48 (PMD); TIA FOTP-124A and FOTP-175; EXFO FTB-5500B specification.
