An ITU channel is written in terahertz, a transponder is trimmed in megahertz
Everything about a dense wavelength-division multiplexed line is planned in terahertz. The channel plan, the amplifier passband, the filter centres inside a wavelength-selective switch — all terahertz, all anchored to one agreed number. Then you open the transponder datasheet and the tolerances change unit: linewidth in megahertz, allowed offset from the channel centre in megahertz, locker stability in megahertz per day. Nothing in that second list can be checked against the first until one of them is rewritten. Put the grid frequency in the left-hand field and the megahertz equivalent appears as you type.
What each unit is actually doing here
One anchor, then arithmetic
A slot has a width as well as a centre
Megahertz is the error budget
From the channel plan to the transponder configuration
Enter the nominal centre exactly as the plan prints it
Grid frequencies are quoted to two or three decimals — 192.10, 193.10, 193.15. Type it with a dot or a comma, whichever your locale gives you; stray spaces are stripped before the value is read.
Set the megahertz figure beside the laser’s tolerance
With the centre expressed in megahertz, an offset limit of ±2 500 MHz or a locker drift of 250 MHz per day sits in the same column and can be subtracted directly rather than reconverted at every step.
Copy the bare number into the line-system template
The copy button on either field puts the digits alone on the clipboard — no unit, no thousands spaces — which is what a provisioning template or a NETCONF payload expects. Ctrl+C inside a field does the same thing.
Reverse it when the measurement comes back in megahertz
An optical spectrum analyser or a coherent receiver’s frequency-offset readout reports in MHz. Press the swap arrows (↔), or simply type into the right-hand field, to see where that lands on the terahertz plan. Either selector can be searched if a figure arrives in GHz instead.
C-band grid points, channel by channel
Vendor channel numbers follow one rule everywhere: multiply the frequency in terahertz by ten, then subtract 1 900. The wavelengths below are vacuum values derived from the frequency, which is the order the standard insists on — frequency is the specification, wavelength is the approximation.
| Channel | Nominal centre | Same value in MHz | Vacuum wavelength | Position on the plan |
|---|---|---|---|---|
| C21 | 192.10 THz | 192 100 000 MHz | 1560.61 nm | Long-wavelength end of the usual 100 GHz plan |
| C27 | 192.70 THz | 192 700 000 MHz | 1555.75 nm | Well inside the erbium amplifier’s flat region |
| C31 | 193.10 THz | 193 100 000 MHz | 1552.52 nm | The anchor every other grid point is measured from |
| C31.5 | 193.15 THz | 193 150 000 MHz | 1552.12 nm | First 50 GHz point above the anchor |
| C34 | 193.40 THz | 193 400 000 MHz | 1550.12 nm | The channel people mean when they say “1550 nm” |
| C40 | 194.00 THz | 194 000 000 MHz | 1545.32 nm | A frequent default on C-band test sets |
| C50 | 195.00 THz | 195 000 000 MHz | 1537.40 nm | Amplifier gain starts tilting here on long spans |
| C61 | 196.10 THz | 196 100 000 MHz | 1528.77 nm | Short-wavelength end of the standard plan |
Slot widths handled as plain numbers
Enter 0.0125 and the reading comes back as 12 500 MHz, so a flexible-grid slot count turns into a megahertz width without reaching for a second calculator.
Decimals kept for a sub-slot offset
Up to eight decimal places survive the conversion, so a centre trimmed to 193.10125 THz still resolves as 193 101 250 MHz instead of collapsing back onto the nominal grid point.
Spacing checked from either side
Search the selectors for GHz when a spacing is quoted that way, and the same page settles whether two centres are one slot apart or two without abandoning the channel plan.
Questions from the optical line-up bench
Why is the whole grid anchored at 193.1 THz?
Because a shared reference was needed and that one is convenient in both directions. In wavelength terms it is 1552.52 nm, comfortably inside silica’s low-loss window and near the middle of the gain an erbium-doped amplifier can hold flat. In frequency terms it is a round figure, so adding steps of 0.1, 0.05 or 0.00625 THz produces centres that stay tidy however far you walk from the anchor. Had the plan been anchored in nanometres instead, every step would have produced an awkward wavelength and no two vendors would have rounded it identically.
How do 50 GHz and 100 GHz spacings map onto channel numbers?
The common vendor convention is channel = frequency in THz × 10 − 1 900, so 192.1 THz is C21, 193.1 THz is C31 and 196.1 THz is C61. On a 100 GHz plan the numbers step by one, because 0.1 THz × 10 = 1. Halve the spacing to 50 GHz and the intermediate centres fall on half numbers — C31.5 at 193.15 THz — which is why some equipment lists channels as 31.5 while other systems renumber the whole plan to keep integers. Expressed in megahertz nothing is ambiguous: neighbours sit 100 000 MHz or 50 000 MHz apart.
Why is laser linewidth specified in megahertz on a terahertz channel?
Linewidth measures how badly the carrier phase wanders, and it is many orders of magnitude smaller than the channel it lives in — writing it in terahertz would be a string of leading zeros. A few megahertz is perfectly acceptable for direct detection, where only received power matters. Coherent formats are far less forgiving: the receiver’s carrier-phase estimator has to follow that wander, and the tolerance tightens as the constellation grows. That is why transmitters intended for 16QAM and 64QAM use external-cavity or narrow-linewidth sources specified below 100 kHz rather than the few-megahertz distributed-feedback part that would happily serve a 10G link.
What does 50 GHz of spacing look like in nanometres?
About 0.40 nm near 1550 nm — but only near 1550 nm, and that caveat is the whole point. Wavelength and frequency are reciprocal, so a fixed frequency step maps onto a wavelength step that grows as you move towards longer wavelengths. Around 1550 nm the exchange rate is roughly 125 GHz per nanometre, making 100 GHz about 0.80 nm and 12.5 GHz about 0.10 nm. Define a plan in nanometres and the spacing quietly changes across the band; define it in terahertz and every slot really is identical, which is precisely why the grid is written in frequency.
How far does a laser drift, and what holds it on channel?
An uncooled distributed-feedback die moves roughly 0.1 nm for every degree of junction temperature, which near 1550 nm is about 12 500 MHz per °C. Two or three degrees of unmanaged drift would therefore walk a channel straight into its neighbour’s slot. That is why every DWDM transmitter carries a thermoelectric cooler and, above it, a wavelength locker referenced to an etalon: together they typically hold the centre inside about ±0.02 nm, near enough ±2 500 MHz, with day-to-day stability quoted around 0.002 nm — roughly 250 MHz — over 24 hours.
No comments yet. Be the first to comment!