S
Seed account
Short RF-over-fiber links mostly care about gain, noise figure and RIN. Once the fiber run reaches tens of kilometres at 1550 nm, two fiber effects start to dominate the design: dispersion-induced RF power fading and stimulated Brillouin scattering (SBS). Both are easy to estimate on paper.
1. Dispersion-induced power fading
With ordinary double-sideband intensity modulation, the optical spectrum has a carrier and two sidebands at plus and minus f_RF. Chromatic dispersion gives the two sidebands opposite phase shifts relative to the carrier. At the photodiode each sideband beats with the carrier, and the two beat products add as vectors. When their relative phase reaches 180 degrees they cancel.
For a chirp-free modulator (for example a push-pull MZM at quadrature) the detected RF power varies as:
P_RF proportional to cos^2(pi * D * L * lambda^2 * f_RF^2 / c)
Nulls occur at:
f_k = sqrt(c * (2k + 1) / (2 * D * L * lambda^2)), k = 0, 1, 2, ...
Numbers for standard single-mode fiber (D = 17 ps/(nm km)) at 1550 nm:
- 5 km: first null at about 27.1 GHz
- 10 km: about 19.2 GHz
- 20 km: about 13.6 GHz (next nulls at about 23.5 and 30.3 GHz)
- 50 km: about 8.6 GHz
Note the 1/sqrt(L) dependence: doubling the length only lowers the null frequency by a factor of 1.41. Below the first null the fading is gradual. At 20 km, a 6 GHz signal loses only about 0.4 dB, but a 10 GHz signal already loses about 3.7 dB, and the loss is very steep near the null.
Directly modulated lasers make this worse and less predictable. Laser chirp (linewidth enhancement factor typically 2 to 5 for DFB lasers) shifts the null positions, usually to lower frequencies for standard fiber with positive dispersion, and the shift depends on bias and modulation depth. Measure the S21 of the actual link over temperature rather than trusting the formula.
Ways around it
- Use the O-band (around 1310 nm), where standard fiber dispersion is near zero. You pay a little more fiber loss (about 0.33 dB/km instead of about 0.2 dB/km) and lose the option of EDFAs, but for links up to a few tens of kilometres this is often the simplest fix.
- Single-sideband modulation: a dual-drive MZM driven through a 90 degree hybrid, or an optical filter that removes one sideband. With only one sideband there is nothing to cancel against, and the fading disappears (dispersion then only adds a frequency-dependent phase, i.e. delay).
- Optical carrier-suppressed or filtered schemes for mm-wave generation.
- Dispersion compensation (DCF or chirped fiber Bragg gratings) when you already have a long-haul infrastructure.
- Transport an IF and upconvert at the remote end, or digitize (digital RoF/CPRI-style) if the latency and bandwidth budget allow.
2. Stimulated Brillouin scattering
Analog links want high received optical power, because RF link gain scales with the square of photocurrent. But a narrow-linewidth laser launched into long fiber hits the SBS threshold, beyond which power is reflected back toward the transmitter instead of propagating. Symptoms: received power stops increasing with launch power, back-reflected power rises sharply, and the link noise floor climbs because the Brillouin process converts into extra intensity noise.
A classic estimate (Smith):
P_th = about 21 * A_eff / (g_B * L_eff)
with A_eff about 80 um^2, g_B about 5e-11 m/W, and L_eff = (1 - exp(-alpha L))/alpha.
For 20 km of standard fiber at 0.2 dB/km, L_eff is about 13 km and P_th comes out around 2.6 mW (about +4 dBm). Practical onset for a sub-MHz-linewidth DFB is commonly a few dB either side of this, depending on fiber type, polarization and laser linewidth; measure it with a circulator and a power meter on the back-reflection.
Ways around it
- Keep launch power below the measured threshold, with margin over temperature.
- Broaden the laser linewidth with a small dither (phase or frequency modulation). This raises the threshold but the added phase noise interacts with dispersion and can convert to intensity noise, so check the RF noise floor after doing it.
- Use the RF modulation itself: a strongly modulated carrier spreads energy and raises the effective threshold somewhat, but do not count on it at low modulation index.
- Split the launch over multiple fibers or wavelengths.
3. Design summary
For a 1550 nm analog link longer than about 10 km, put three items into the link budget early: the first fading null for your modulation format, the SBS-limited launch power, and the resulting maximum photocurrent. Those three often decide the architecture before any RF component is chosen.
If you have measured fading nulls with DMLs at different bias points, posting the data would help others calibrate the chirp effect.
1. Dispersion-induced power fading
With ordinary double-sideband intensity modulation, the optical spectrum has a carrier and two sidebands at plus and minus f_RF. Chromatic dispersion gives the two sidebands opposite phase shifts relative to the carrier. At the photodiode each sideband beats with the carrier, and the two beat products add as vectors. When their relative phase reaches 180 degrees they cancel.
For a chirp-free modulator (for example a push-pull MZM at quadrature) the detected RF power varies as:
P_RF proportional to cos^2(pi * D * L * lambda^2 * f_RF^2 / c)
Nulls occur at:
f_k = sqrt(c * (2k + 1) / (2 * D * L * lambda^2)), k = 0, 1, 2, ...
Numbers for standard single-mode fiber (D = 17 ps/(nm km)) at 1550 nm:
- 5 km: first null at about 27.1 GHz
- 10 km: about 19.2 GHz
- 20 km: about 13.6 GHz (next nulls at about 23.5 and 30.3 GHz)
- 50 km: about 8.6 GHz
Note the 1/sqrt(L) dependence: doubling the length only lowers the null frequency by a factor of 1.41. Below the first null the fading is gradual. At 20 km, a 6 GHz signal loses only about 0.4 dB, but a 10 GHz signal already loses about 3.7 dB, and the loss is very steep near the null.
Directly modulated lasers make this worse and less predictable. Laser chirp (linewidth enhancement factor typically 2 to 5 for DFB lasers) shifts the null positions, usually to lower frequencies for standard fiber with positive dispersion, and the shift depends on bias and modulation depth. Measure the S21 of the actual link over temperature rather than trusting the formula.
Ways around it
- Use the O-band (around 1310 nm), where standard fiber dispersion is near zero. You pay a little more fiber loss (about 0.33 dB/km instead of about 0.2 dB/km) and lose the option of EDFAs, but for links up to a few tens of kilometres this is often the simplest fix.
- Single-sideband modulation: a dual-drive MZM driven through a 90 degree hybrid, or an optical filter that removes one sideband. With only one sideband there is nothing to cancel against, and the fading disappears (dispersion then only adds a frequency-dependent phase, i.e. delay).
- Optical carrier-suppressed or filtered schemes for mm-wave generation.
- Dispersion compensation (DCF or chirped fiber Bragg gratings) when you already have a long-haul infrastructure.
- Transport an IF and upconvert at the remote end, or digitize (digital RoF/CPRI-style) if the latency and bandwidth budget allow.
2. Stimulated Brillouin scattering
Analog links want high received optical power, because RF link gain scales with the square of photocurrent. But a narrow-linewidth laser launched into long fiber hits the SBS threshold, beyond which power is reflected back toward the transmitter instead of propagating. Symptoms: received power stops increasing with launch power, back-reflected power rises sharply, and the link noise floor climbs because the Brillouin process converts into extra intensity noise.
A classic estimate (Smith):
P_th = about 21 * A_eff / (g_B * L_eff)
with A_eff about 80 um^2, g_B about 5e-11 m/W, and L_eff = (1 - exp(-alpha L))/alpha.
For 20 km of standard fiber at 0.2 dB/km, L_eff is about 13 km and P_th comes out around 2.6 mW (about +4 dBm). Practical onset for a sub-MHz-linewidth DFB is commonly a few dB either side of this, depending on fiber type, polarization and laser linewidth; measure it with a circulator and a power meter on the back-reflection.
Ways around it
- Keep launch power below the measured threshold, with margin over temperature.
- Broaden the laser linewidth with a small dither (phase or frequency modulation). This raises the threshold but the added phase noise interacts with dispersion and can convert to intensity noise, so check the RF noise floor after doing it.
- Use the RF modulation itself: a strongly modulated carrier spreads energy and raises the effective threshold somewhat, but do not count on it at low modulation index.
- Split the launch over multiple fibers or wavelengths.
3. Design summary
For a 1550 nm analog link longer than about 10 km, put three items into the link budget early: the first fading null for your modulation format, the SBS-limited launch power, and the resulting maximum photocurrent. Those three often decide the architecture before any RF component is chosen.
If you have measured fading nulls with DMLs at different bias points, posting the data would help others calibrate the chirp effect.