Context

I am working on a Free Space Optical Communication (FSOC) project. Compute nodes will connect through ethernet to a media converter which will have a custom Small Form-factor Pluggable (SFP). This bidirectional SFP will send 1550nm wavelength data through free space to a receiving custom SFP. The receiving SFP will interface with a media converter that is connected by ethernet to another compute node. The compute nodes should be able to communicate as if they were connected by ethernet. End target is communication at 1km at 1gb/s with total cost of bidirectional communication node under $1000 and under 1 kg. Constraints may turn out to be impossible and need tuning. Goal is to put these on drones to make a mesh network.

Week 2 Goals

Finish last week’s goal of reading I2C data. Get a gadget that reads the SFP from the breakout board.
Determine needs to create a mesh network (how many edges should each node support). Rough idea of how node bidirectional edges would work on a drone platform.
Make an initial calculation of size, weight, power, cost, and performance of the system. Interested in trade-off size of receiving lens and expected performance.

Results

Goals were wildly optimistic and was only able to accomplish creating a FSOC SFP simulator to help analyze SWAP-C.

From ‘What is the SWAP-C’ to ‘Is this Feasible’

I wanted to understand the dependencies on different components and what was the true ‘bottle neck’ for making a auto-alignment Free Space Optical Communication system work using custom SFPs.

I followed the same approach from last week to help me understand the components of the system, by building a simple html/css/js simulator. Initially thought the limiting factor would be how far the optical signal can reach and the sensitivity of the receiver. Turns out the limiting factor is if the beam can stay on the receiver array and not fall off.

This needs more iteration. Still need to add more base physics to the simulation so better constraints can be found. Also need cost of goods, better selection of possible components through maybe a deep research prompt in Gemini. Served it’s purpose for this week’s goals and provides a good tech asset for further development. Repo is here. Read the max distance explained to understand where the failure point is in the config set up.

TLDR: Alignment is the limiting factor (can the beam stay pointed at the receiver), not whether enough optical power can get to the receiver.

Simulator found 15m is the current max distance with a cheap bill of materials on 2 drones in non-calm wind conditions. Try out the simulator here.

I initially wanted to understand an SFP TOSA to ROSA link, but realized that it wouldn’t be accurate in detailing if the project was feasible. I needed to include auto alignment as the system would need a way to “point” at the other nodes without manual intervention. The 2 nodes would start close and pointed at each other to initialize, then as pulled apart there would be a data transfer to ensure alignment. The thought is there will be an array of photoreceptors on the Rx node, and a data stream would be sent back to the Tx node ‘how far off center’ the beam was. This would allow the Tx node to then update some kind of mirror or gimbal to remain aligned.

Auto-Alignment System

This required research on if it was possible to send 2 data channels over the same 1550nm wavelength.

Landed on 2 possibilities:

Sending alignment data back on a “pilot channel”
Using a secondary wavelength to send this alignment data

Pilot Channel

The idea is that while the main laser carries the high-speed data, a small subcarrier or modulation tone (the “pilot”) rides along to send low-rate control info, like alignment feedback or link health. On the hardware side, you’d need a way to inject and detect that tone — usually a wavelength coupler or an RF subcarrier modulator on the transmit side, and a photodiode plus DSP or analog filter to peel it off on the receive side. That pilot data could come from an array of photodiodes measuring where the beam’s hitting, so the receiver can tell the transmitter how far off it is. The transmitter would then tweak a MEMS mirror or gimbal to stay locked in, all while keeping the main data stream uninterrupted.

The upside of using a pilot channel for a 1 Gb/s FSOC link is that it gives you continuous alignment feedback without needing a separate optical path or camera system — you get real-time pointing corrections baked into the same beam. It’s also relatively lightweight in terms of bandwidth, since the pilot only occupies a narrow spectral or modulation slice and can coexist with the primary data stream.

The downsides are that embedding the pilot tone can slightly reduce SNR and modulation depth available for the main data, especially if the optical power budget is tight. The system also becomes more complex — you now need precise filtering and demodulation to separate the pilot signal without corrupting the gigabit payload. Finally, the control loop latency and crosstalk between the pilot feedback and high-speed modulation can introduce jitter or alignment oscillations if not carefully tuned.

Secondary Wavelength

Using a secondary wavelength for the pilot channel means running a separate optical carrier — still around 1550 nm but offset by a few nanometers — dedicated to alignment or telemetry data. The main 1 Gb/s data stays on the primary wavelength, while the secondary beam carries a much lower-rate control signal used for alignment, status, or link maintenance. The two are combined and separated using wavelength-division multiplexers (WDMs), so each path can be optimized independently — one for clean data modulation, the other for robust beaconing and tracking. This makes it easier to filter and detect the pilot without interfering with the main signal and allows different detectors or optics to be tuned for stability vs. speed.

The pros are cleaner isolation between data and control channels, better noise immunity, and more flexibility to adjust pilot power or modulation without touching the data path. It also simplifies DSP since the pilot can be detected with low-speed, high-sensitivity hardware. The cons are increased optical complexity — you now need WDM filters, possibly dual-band lenses or coatings, and two precisely aligned laser sources. There’s also a small hit to optical efficiency and alignment difficulty since both wavelengths must share the same beam path and focus geometry, which can drift with temperature or optical element dispersion.

Analysis

Used ChatGPT to discuss the 2 different options, pilot channel or alternative wavelength and it kept suggesting using an alternative wavelength. To me that seems more complicated, how you need 2 different systems tuned, and all the downsides of the same channel didn’t seem insurmountable. Turns out the big issue is all the difficulty in getting a clean signal at 1 gb/s over the air. Any type of ‘messing’ with this reduces the ability for this to work.

I want to extend the simulator to take all this discussion and analysis into account. It seems like the simulator will be greatly extended to help actually understand sub components and hopefully turn out new bottle necks.

ChatGPT printed “With unlimited optical power, you can crank the link margin up, but your pilot still shares the same photodiode and TIA. At high photocurrents, the TIA’s linearity and thermal drift become the new bottlenecks — so extra power helps SNR but can worsen distortion unless you redesign the front-end bias and gain stages.”. This is the type of realization I want the simulation to surface, so there is a deeper level the simulator needs to get to.

Conclusion

Initially the simulator provided a distance of 1.3km which seemed like great news, then I realize we hadn’t added simulating gimbal and pointing jitter. Then the max distance went down to under 0.48km, and then to 15m.

Simulator reveals that pointing is the most difficult part. Need to iterate and figure out what is keeping it down to 15m. Maybe it is a bug, but maybe that is how hard it is to stay pointed at an array on 2 drones.

Next Week

I don’t want to end up bike-shedding and playing software simulator so will take up working on the I2C reader with some hardware this week. Want to be able to read SFP’s response as my purchase acts as the “master” for information.

Appendix

After messing with ChatGPT for a while created a good document explaining the different components of the systems and their limitations. Link. Full document below. Will likely iterate on it more with ChatGPT to get more technical and engineering, then will apply more and more of the concepts to the simulator. This should help with exploring position solution configurations and where bottle necks arise.

1. Introduction — What 1 Gb/s Free-Space Optical Communication (FSOC) Really Means

Free-Space Optical Communication (FSOC) is, at its core, a laser link that uses light instead of electrical signals or radio waves to transmit data through open air. Instead of a fiber guiding the light, the medium is the atmosphere itself — meaning the beam must be precisely aimed, and the air must stay clear enough for photons to arrive intact.

At 1 Gb/s, we’re essentially building a wireless optical Ethernet link equivalent to what a 1000BASE-LX fiber transceiver does inside a data center — but without the fiber. The system must take binary digital data (1s and 0s), modulate a laser at gigabit-per-second speeds, launch that light across tens or hundreds of meters, and then convert it back to an electrical signal at the receiver, all while maintaining synchronization and low bit-error rates.

In your case, this FSOC system isn’t generating Ethernet frames directly — it’s taking the electrical differential signals from a media converter or SFP module (via its 20-pin connector) and transmitting those over free space. That means the data format and timing are identical to a 1000BASE-LX optical fiber link: 1.25 GBd NRZ signaling, line-coded using 8b/10b encoding. The challenge isn’t digital logic — it’s reproducing the physical-layer performance of fiber in air.

Why 1550 nm

FSOC systems almost universally use light in the 1550 nm band, part of the infrared spectrum. There are three main reasons:

Telecom component availability: Decades of fiber-optic telecom research have made 1550 nm lasers, photodiodes, and WDM components cheap and high-performance.

Atmospheric transmission window: The atmosphere is relatively transparent around 1550 nm, with minimal absorption by water vapor or CO₂ compared to shorter wavelengths.

Eye safety: At 1550 nm, light is absorbed mostly by the cornea rather than the retina, dramatically raising the permissible exposure limit (PEL) for human eyes.

This last point — eye safety — drives major design decisions.

Eye Safety: What It Means Physically

Human eyes focus visible and near-infrared light onto the retina. At wavelengths below ~1400 nm, photons pass through the cornea and lens and can directly burn retinal tissue, causing permanent blindness. Above ~1400 nm (including 1550 nm), water in the cornea absorbs the radiation first, so the energy doesn’t reach the retina — the damage risk is limited to corneal heating, which is much less sensitive.

Because of this, regulatory standards (IEC 60825-1) impose strict power limits for any optical system that could expose a human eye to a laser beam. For example, a Class 1 laser product at 1550 nm is limited to roughly 10 mW continuous optical power (depending on beam diameter and divergence). That’s why most commercial FSOC units operate below this level, spreading the beam slightly to stay eye-safe.

If Eye Safety Didn’t Matter

If we remove the eye-safety constraint — as in certain research, military, or closed-environment systems — several aspects of FSOC design change:

Transmit Power: The optical transmitter could output tens or even hundreds of milliwatts. This improves the link budget, making it possible to close longer distances or tolerate more atmospheric loss.

Beam Divergence: Engineers could tighten beam divergence dramatically (sub-milliradian), increasing irradiance on the receiver without worrying about unsafe exposure levels.

Aperture Size: The receiver’s optical aperture could be smaller while maintaining the same received power, simplifying optics.

Link Robustness: Higher power provides more margin against scintillation (rapid amplitude fading due to turbulence) and misalignment.

Thermal and Electrical Limits Become Dominant: Without eye-safety as the limiting factor, the next bottlenecks are component heating, photodiode saturation, and amplifier linearity.

However — and this becomes critical for our later sections — even with unlimited optical power, other constraints don’t disappear. Once your receiver photodiode and transimpedance amplifier (TIA) are driven close to their linear limits, extra photons stop helping and start distorting the signal. Also, high instantaneous optical intensity can exacerbate nonlinearities, mode competition in the laser, and intermodulation when you superimpose pilot signals.

In short: removing eye-safety limits buys you margin in the power budget, but it doesn’t eliminate the precision, timing, or linearity challenges that define high-speed FSOC.

2. Digital Signaling and Timing Physics

To understand how gigabit optical links actually move data, we need to start with how electrical and optical systems represent digital information physically.

What is a Symbol?

A symbol is the smallest unit of signal that carries information. In a binary system, each symbol can represent a ‘1’ or a ‘0’. In higher-order modulation formats (like QPSK or 16-QAM used in RF systems), a symbol can represent multiple bits. But in our FSOC system, we’re using binary non-return-to-zero (NRZ) signaling — meaning each bit is represented by a single voltage or light intensity level that stays constant for the duration of one bit period.

At 1 Gb/s, each bit period is 1 nanosecond (1e-9 s). The transmitter must toggle the light intensity cleanly between two levels every nanosecond, and the receiver must detect those transitions accurately despite all noise, jitter, and distortion.

What is 1000BASE-X and 8b/10b Encoding?

1000BASE-X is the physical layer standard for Gigabit Ethernet over optical fiber. It runs at 1.25 gigabaud (GBd) — that’s 1.25 billion symbols per second. The extra 25% overhead (1.25 vs 1.0) comes from 8b/10b encoding, a line code that maps 8 data bits into 10 transmitted bits. This encoding ensures:

DC balance: roughly equal numbers of ones and zeros, keeping the average optical power stable.

Transition density: frequent 0→1 or 1→0 transitions so the receiver’s clock and data recovery (CDR) circuit can stay synchronized.

Error detection: a few invalid 10-bit combinations can be used to detect transmission faults.

What is NRZ Signaling?

Non-Return-to-Zero (NRZ) means the signal doesn’t reset to a neutral level between bits. A ‘1’ might be represented by a high optical power (laser on), and a ‘0’ by a low power (laser partially off). The laser intensity directly follows the binary pattern. NRZ is efficient and simple, but it’s sensitive to inter-symbol interference (ISI) — when slow transitions or limited bandwidth cause one bit’s level to smear into the next.

Symbol Rate vs. Bandwidth

The symbol rate (baud rate) defines how fast the signal changes. To reproduce those changes, the system must have enough analog bandwidth — roughly half the symbol rate for NRZ. So a 1.25 GBd link requires around 700 MHz to 1 GHz of clean analog bandwidth from the laser driver, optical path, and receiver electronics. That bandwidth ensures sharp rise and fall times and preserves the shape of the signal eye.

Bandwidth in this context doesn’t mean how many bits per second you send — it refers to how fast the system’s voltage or optical intensity can change. Limited bandwidth means slower edges, which blur transitions and cause ISI.

What is an Eye Diagram?

If you overlay thousands of bits of the received signal on an oscilloscope, you get an eye diagram — a pattern that literally looks like an eye. The open space in the center shows where the receiver can safely sample a bit without confusion. The height of the eye shows amplitude margin (SNR), and the width shows timing margin (jitter tolerance). When distortion or noise increases, the eye “closes,” leading to bit errors.

What is Jitter and Timing Stability?

Jitter is the deviation of bit transition timing from ideal positions. It comes from noise, power supply ripple, temperature drift, and imperfect phase-locked loops (PLLs). Too much jitter means the receiver samples a bit before or after the correct moment, leading to errors.

In gigabit systems, total jitter must stay below about 0.25 unit intervals (UI) — that’s less than 250 picoseconds of total timing uncertainty. This constraint drives every part of the transmitter and receiver design, from the laser driver’s rise/fall time to the TIA’s phase response.

Physics of Fast Transitions — Why It’s Hard

To switch optical power at GHz speeds, the laser driver and photodiode circuits fight several physical effects:

Carrier lifetime: electrons and holes in the laser’s active region don’t respond instantly; their recombination sets a natural modulation bandwidth.

Parasitic capacitance and inductance: every wire, bond pad, and trace adds reactive elements that slow voltage/current transitions (RC and LC time constants).

Material limits: the laser’s junction and the photodiode’s depletion layer have finite transit times for carriers; pushing too fast leads to amplitude roll-off and phase delay.

In fiber optics, these effects are carefully managed with impedance matching and high-speed driver ICs. In FSOC, the same rules apply — you’re just coupling the optical signal into free space instead of a fiber core.

How the Pilot Channel Affects Digital Signaling

Now that we’ve described how gigabit signaling works, we can understand how a pilot tone or secondary wavelength would interact with it:

Same-Wavelength Pilot: Adds a low-frequency modulation on top of the gigabit NRZ waveform. This slightly distorts the optical intensity levels, introducing low-frequency amplitude ripple. That ripple shifts the average optical power, narrowing the eye opening and adding deterministic jitter.

Secondary-Wavelength Pilot: Uses a separate laser and wavelength, so the gigabit signal’s waveform remains untouched. The pilot light can carry alignment or telemetry data independently, detected by a separate photodiode. This avoids corrupting the main eye diagram entirely.

In short, 1 Gb/s signaling pushes every analog and timing parameter close to its physical limits. Adding a same-wavelength pilot modulates the very signal the receiver depends on for clock recovery, while a secondary wavelength keeps that timing channel clean at the cost of extra optics.

Next, we’ll dive into the laser physics and driver electronics — how p–n junctions actually generate coherent light, what “bias above threshold” means physically, and how these electrical-to-optical conversions behave when you try to add a pilot tone or second wavelength.

3. Laser Physics and Driver Electronics

This section builds the physics from the p–n junction up, then maps it to real transceiver blocks (DFB/EML lasers, drivers, bias tees, impedance control). We’ll finish with how a same-wavelength pilot vs a secondary-wavelength pilot perturbs each piece.

3.1 How a semiconductor laser makes coherent light

p–n (often p–i–n) junction and recombination

A laser diode is a forward-biased semiconductor junction. Injected electrons (from n-side) and holes (from p-side) recombine in an active region (often a quantum well) and emit photons. Two regimes exist:

Below threshold current (I_th): emission is spontaneous (LED-like), broadband, low coherence.

At/above I_th: stimulated emission dominates inside an optical cavity (cleaved/DBR/DFB mirrors). Photons stimulate more identical photons along the same mode → narrow spectrum, high coherence.

Why bias above threshold? Crossing threshold each bit would be slow and noisy. Keeping a DC bias just above I_th holds the gain medium in the lasing regime so small-signal modulation is fast and linear(ish). This minimizes turn-on delay, reduces relative intensity noise (RIN), and shrinks pattern dependence.

Cavity, modes, DFB vs EML

DFB (Distributed Feedback) laser: grating in the cavity selects a single longitudinal mode → narrow linewidth at 1550 nm; good for direct intensity modulation up to multi‑GHz.

EML (Electro-Absorption Modulated Laser): a DFB supplies CW light; a separate EA modulator imposes intensity. EMLs improve high-speed eye quality (lower chirp), at cost/complexity.

External modulator (MZM/LiNbO₃): even cleaner intensity/phase control and very low chirp; bigger, pricier, and needs a separate laser.

Key laser parameters

Threshold current (I_th): where gain equals cavity loss.

Slope efficiency (mW/mA): optical power per modulation current above bias.

Linewidth / phase noise: frequency noise from spontaneous emission.

RIN (Relative Intensity Noise): intrinsic intensity fluctuations (dB/Hz).

Chirp: intensity modulation induces refractive index change → frequency (phase) modulation; in dispersive media chirp broadens pulses. (Free space has negligible chromatic dispersion vs fiber, so chirp is much less harmful for FSOC, but it still participates in AM↔PM conversion through components.)

3.2 From bits to photons: bias + modulation

Bias point and extinction ratio (ER)

Bias current (I_B): set just above I_th for speed and linearity.

Modulation current (i_m(t)): superimposed on I_B to form NRZ ones/zeros.

Extinction ratio (ER) = P_1/P_0 (often in dB). Higher ER → taller eye. Too high a swing can push nonlinearity, thermal load, and chirp.

Rise/fall limits and parasitics

Carrier lifetime + junction capacitance + bond-wire inductance + package/PCB transmission lines set a practical bandwidth. You fight this with controlled-impedance (50/100 Ω) launches, very short interconnects, and drivers with GHz gain‑bandwidth.

Driver and bias-tee topology

The driver supplies fast AC swing; a bias tee (inductor + capacitor network) injects DC bias without loading the high-speed path.

Impedance matching (to the SFP electrical interface and the laser package) prevents reflections that cause ISI and deterministic jitter.

3.3 Noise and linearity through the E/O chain

Nonlinearity: The L–I curve is not perfectly linear; large swings produce harmonics and intermodulation (IMD) when more than one tone exists.

Thermal drift: I_th rises with temperature; slope efficiency falls. Systems use APC (automatic power control) and ATC (automatic temperature control/Thermoelectric cooler) to hold output power and wavelength steady.

AM–PM conversion: Packaging and optics convert some amplitude changes into phase (and vice versa), creating additional timing error paths.

3.4 What a same‑wavelength pilot does here

Consider the optical power leaving the transmitter:

P(t) = P_data(t) [1 + m_p cos(2π f_p t)]

with pilot modulation index (m_p << 1) and pilot frequency (f_p) in the kHz–MHz range.

ER tax & eye ripple: The pilot eats a sliver of dynamic range. Even at 1–2% index, you superimpose low‑frequency ripple on the NRZ envelope → vertical eye closure.

Intermodulation: Nonlinear L–I and driver/TIA nonlinearity mix the pilot with the NRZ spectrum → sidebands that fold into baseband after square‑law detection at the receiver, raising the noise floor.

Bias interaction: If the pilot is implemented as bias dither, it perturbs I_B around threshold, changing instantaneous slope efficiency and RIN.

Power isn’t a silver bullet: Even if eye safety is irrelevant and you raise total optical power, the front-end linearity of PD/TIA clamps the benefit; too much optical current compresses the receiver and worsens IMD.

Mitigations (same-λ): keep m_p ≤ 1–2%, choose f_p away from control and CDR loop bandwidths, and ensure clean pilot notch at the receiver with low group‑delay ripple.

3.5 What a secondary‑wavelength pilot does here

The data laser’s bias, ER, and swing remain untouched. No AM pilot rides on the payload.

The pilot uses a second DFB (e.g., ±2–4 nm from data) combined with a WDM. You can choose any pilot modulation (AM/FM/phase dither) and any index needed for robust tracking without altering the data laser’s operating point.

Thermal and aging control are per‑laser; data and pilot can have independent APC/ATC loops.

Trade: extra parts (DFB #2, WDMs, possibly a second PD/TIA on Rx) and small chromatic focus/reflectivity differences to manage in the collimator path.

3.6 Practical design checklist (Tx side)

Pick the laser: DFB for compactness and adequate speed; EML if you want lower chirp and cleaner eyes at the cost of BOM.

Set bias: I_B ≈ I_th + 10–30% for margin over temperature.

Target ER: ~6–9 dB for 1.25 GBd IM/DD is common; verify with link‑budget and receiver sensitivity.

Control parasitics: 50/100 Ω lines, sub‑mm bondwires, minimal stubs; simulate S‑parameters.

Thermal: Heatsinking or TEC for wavelength/power stability; place thermistor close to the can.

If same‑λ pilot: design the driver to inject a tiny, spectrally clean dither; validate no spurs near the CDR loop bandwidth, and characterize eye with pilot on/off across temperature.

If secondary‑λ: spec WDM IL (<1 dB) and isolation (>30–40 dB); ensure fiberized WDM → single shared collimator stack-up for repeatable co‑boresight.

Up next: Section 4 — Optical Launch and Collimation (Gaussian beams, divergence, coatings, pointing actuators), with explicit notes on how same‑λ vs secondary‑λ pilots impact optical design and alignment.

4. Optical Launch and Collimation

Once the laser output is generated and modulated, the next stage is getting that light into free space cleanly, stably, and efficiently. This section walks through beam physics, optical design, and the impact of using a same-wavelength versus secondary-wavelength pilot.

4.1 Gaussian beams and divergence

Beam fundamentals

A laser doesn’t emit a perfectly parallel beam — it emits a Gaussian beam that naturally diverges over distance. The beam’s waist (the narrowest point) and divergence angle (spread in radians) are connected by:

θ = λ / (π w₀)

where λ is wavelength (≈1550 nm) and w₀ is the waist radius. A smaller waist (tighter focus) gives higher intensity but faster divergence. Engineers balance these to achieve the desired spot size at the receiver.

For FSOC, beams are often expanded through a collimator or small telescope to lower divergence (sub-milliradian). For example, a 5 mm waist gives a divergence of about 0.1 mrad — meaning 10 cm of spread per kilometer of distance.

Why divergence matters

Too wide: power density drops fast, reducing received SNR.

Too narrow: micro-radian pointing errors cause the link to miss entirely.

Atmospheric turbulence: causes the beam centroid to wander and the intensity to fluctuate (scintillation). Larger apertures average these effects out.

Beam quality and M²

Real lasers deviate from an ideal Gaussian beam. The M² factor quantifies this — an M² of 1.0 is diffraction-limited, while 1.2–1.5 is typical for DFBs. Higher M² means larger divergence for a given waist, reducing coupling efficiency.

4.2 Collimation optics and coatings

Lenses and expansion

To reduce divergence, designers use collimating lenses or multi-element beam expanders (Keplerian or Galilean telescopes). The ratio of expansion determines beam size and divergence:

θ_out = θ_in / M

where M is the magnification ratio. Expanding a 3 mm beam by 5× yields ~0.2 mrad → 0.04 mrad divergence.

Anti-reflection (AR) coatings

Every air-glass surface reflects a few percent of light (Fresnel reflection). AR coatings — typically quarter-wave stacks tuned for 1550 nm — reduce this to <0.5% reflection, improving launch efficiency and preventing unwanted back-reflections that could destabilize the laser.

Alignment tolerances

Collimation alignment is critical: a 0.1° misalignment at the transmitter can produce centimeter-level spot shifts at the receiver over tens of meters. Mounts often use kinematic adjusters (tip/tilt screws) or MEMS mirrors for fine tuning.

4.3 Optical launch alignment and pointing control

Static alignment

In a fixed link, optical heads can be mechanically aligned using precision jigs or cameras. The outgoing beam is centered on the receiver aperture at a reference distance.

Dynamic alignment (auto-pointing)

For mobile or long-range links, you need active pointing control. Sensors (quad photodiodes, camera arrays, or separate pilot detectors) feed back alignment errors to actuators (gimbals or MEMS mirrors) that keep the beam centered.

Typical control loops run at tens to hundreds of Hz — fast enough to reject slow drift and turbulence but not so fast that they chase scintillation noise.

Control loop coupling

If the same optical carrier carries both data and pilot, the feedback signal must be extracted without disturbing the data. If you use a secondary wavelength, you can fully separate the control optics from the data optics, simplifying both loops.

4.4 How same-wavelength vs. secondary-wavelength pilots impact optics

Same-wavelength pilot

Perfect co-alignment: pilot and data always share the same optical axis and turbulence state.

Single set of optics: no WDMs or dichroic filters needed.

Drawbacks: The pilot tone modulates the main optical envelope, so the receiver’s photodiode and TIA must handle both data and pilot simultaneously. Any alignment-sensing or beacon function extracted from that signal can’t easily be filtered optically — only electronically, which adds noise and group delay.

Secondary-wavelength pilot

Independent power control: you can crank up pilot power without touching data amplitude.

WDM separation: each wavelength can have its own detector and optics tuned for its power range and bandwidth.

Optical isolation: a simple dichroic mirror or WDM coupler can combine/separate beams with >30–40 dB isolation.

Drawbacks: small chromatic differences — focus, refraction index, and AR coating reflectivity — can cause a few microradians of offset between beams. Temperature and glass dispersion make this drift slowly unless carefully compensated.

Thermal and chromatic considerations

At 1550 ± 5 nm, silica and typical IR coatings change refractive index by ~1×10⁻⁵ per nm. Over multiple elements, this can cause measurable pointing drift as temperature changes. Designers minimize this by:

Using achromatic doublets or low-dispersion glasses.

Designing the pilot wavelength within a few nm of the data wavelength.

Periodically recalibrating alignment using a wide-field detector.

4.5 Optomechanical tolerances and material physics

Thermal expansion

Mounts and housings expand with temperature (ΔL = α·L·ΔT). For aluminum (α ≈ 23×10⁻⁶ /°C), a 10 cm optical base shifts by 23 µm per 10°C — several beam diameters at long range. High-stability systems use Invar, carbon fiber, or active feedback to compensate.

Vibration and damping

Vibration introduces micro-radian misalignment that modulates received intensity. Engineers use stiff, damped mounts and often isolate optics from the main chassis using elastomers or flexures.

Optical cleanliness

At 1550 nm, even small dust particles can scatter light strongly at small angles. Optical surfaces must be AR-coated and clean; otherwise, speckle and scattering degrade the link.

4.6 Summary: Optical section pros and cons

Aspect Same-wavelength pilot Secondary-wavelength pilot
Alignment fidelity Perfect co-propagation Slight chromatic offset (correctable)
Optical complexity Minimal Requires WDMs / dichroics
Noise coupling Pilot adds amplitude ripple into data Clean separation between paths
Power control Shared with data Independent tuning possible
Thermal sensitivity Single thermal path Slight dispersion-induced drift
Implementation cost Lower Higher (extra optics, laser, PD)

Up next: Section 5 — Propagation Through Atmosphere, where we’ll explain turbulence physics (Cn², scintillation, coherence length) and compare how same- and secondary-wavelength pilots respond to fading and beam wander.

5. Propagation Through Atmosphere

At this point, the light has left the transmitter’s optics and is traveling through air — which is far from a perfect medium. Unlike fiber, the refractive index of air fluctuates continuously with temperature, pressure, and humidity. These variations introduce both random amplitude changes (scintillation) and beam direction wander (angle-of-arrival fluctuations) that define the main physical challenges in FSOC.

5.1 Refractive index variations and turbulence physics

Structure constant (Cn²)

Atmospheric turbulence is quantified by the refractive index structure constant, denoted (C_n^2), with units m⁻²⁄³. It measures how strongly the refractive index varies spatially. Typical values:

Good conditions: (10^{-16}) to (10^{-15}) m⁻²⁄³

Moderate turbulence: (10^{-14}) m⁻²⁄³

Bad (urban/heat haze): (10^{-13}) m⁻²⁄³ or higher

Larger (C_n^2) means stronger fluctuations and worse link stability.

Kolmogorov turbulence spectrum

Turbulent eddies follow a Kolmogorov cascade: energy transfers from large scales (meters) to small ones (millimeters). These refractive index cells act like moving lenses that bend and focus the beam differently along its path, producing speckle-like intensity patterns at the receiver.

Fried coherence length (r₀)

The Fried parameter, (r_0), describes the spatial coherence length — the aperture size over which the phase front remains roughly flat. For visible/IR light:

[ r_0 ≈ 0.185 (λ^2 / C_n^2 L)^{3/5} ]

For example, at 1550 nm over 1 km with (C_n^2 = 10^{-14}), (r_0) ≈ 5 cm. That means apertures larger than ~5 cm don’t gain coherent addition — parts of the beam see different phase distortions.

5.2 Scintillation and fading

Intensity fluctuations

When a beam passes through these cells, random focusing and defocusing cause the intensity at the receiver to vary rapidly — known as scintillation. It’s similar to how stars twinkle when viewed through the atmosphere.

The scintillation index, (σ_I^2), quantifies normalized intensity variance:

[ σ_I^2 = rac{⟨I^2⟩ – ⟨I⟩^2}{⟨I⟩^2} ]

Weak turbulence: (σ_I^2 < 0.3)

Strong turbulence: (σ_I^2 > 1.0)

When (σ_I^2) approaches 1, deep fades can momentarily drive the signal to near-zero, breaking the link unless you have margin or error correction.

Beam wander

Large-scale eddies (centimeter to meter scale) deflect the entire beam by micro-radians — called angle-of-arrival jitter. At 100 m range, a 10 µrad deflection = 1 mm shift at the receiver. Over longer distances, these shifts can cause significant signal loss unless the receiver’s aperture or pointing loop compensates.

Temporal dynamics

Turbulence patterns are carried by the wind, so fading occurs on timescales of milliseconds to seconds. FSOC receivers often use automatic gain control (AGC) and forward error correction (FEC) to smooth over these fluctuations.

5.3 Atmospheric absorption and scattering

While turbulence dominates short links (<2 km), molecular absorption and aerosol scattering become important over long paths.

Absorption: caused by water vapor, CO₂, and ozone. Fortunately, 1550 nm sits in a “window” of low absorption.

Scattering: from fog, dust, and haze. Mie scattering (particle sizes ≈ wavelength) is the main loss mechanism in fog; it scales roughly as (e^{-αL}), where α is the attenuation coefficient.

Rain/snow: liquid droplets and snowflakes also scatter and refract light, causing temporary outages or large power penalties.

Link budgets usually include 2–3 dB margin for clear air, 10–20 dB for haze, and total outage if visibility <100 m.

5.4 Mitigating turbulence effects

Aperture averaging

A larger receiver aperture averages over multiple speckle cells, reducing scintillation depth roughly as 1/√N, where N is the number of independent speckle regions across the aperture.

Adaptive optics

High-end systems use wavefront sensors and deformable mirrors to correct the distorted phase front in real time — restoring a sharp beam. This is complex and usually reserved for astronomical or long-haul FSOC systems.

Spatial and temporal diversity

You can transmit the same data on multiple beams (spatial diversity) or use interleaved time slots to reduce fade correlation. Combining receivers improves robustness but increases mechanical and optical complexity.

5.5 Pilot vs. secondary-wavelength response in turbulence

Same-wavelength pilot

The pilot and data experience identical turbulence (since they are the same photons). This makes the pilot an excellent real-time channel probe — you can use its amplitude or phase to infer the state of the link for adaptive optics or pointing.

However, since the pilot is encoded on the same carrier, the amplitude fluctuations also appear in the data signal, making AGC and CDR loops interact. Strong scintillation can cause the pilot’s subcarrier to beat with data, modulating the noise floor.

Secondary-wavelength pilot

Slightly different wavelengths (~±2–4 nm) experience nearly identical turbulence, but not perfectly identical — because refractive index fluctuations are wavelength-dependent (weakly, via dispersion). In practice, the correlation is >0.99, so both fade together.

The key advantage: because the pilot is on a separate detector chain, you can design its bandwidth and AGC specifically for low-rate tracking. Even when the data channel saturates or fades momentarily, the pilot receiver can maintain lock, helping reacquire alignment.

Summary of atmospheric interaction

Effect Same-wavelength pilot Secondary-wavelength pilot
Turbulence correlation Perfect (same photons) ~99% correlated
Fading behavior Shared fades; pilot adds amplitude modulation to data Independent detection; pilot unaffected by data load
Recovery after fade Faster CDR reacquisition, but tighter coupling Easier alignment recovery; less jitter coupling
Implementation Simple (no WDM) More parts, but more robust control

Up next: Section 6 — Photodiode and Receiver Front-End Physics, where we’ll dive into photoelectric conversion, TIA noise, and how same- vs. secondary-wavelength pilots interact with the receiver’s linearity and filtering.

6. Photodiode and Receiver Front-End Physics

At the receiver, the optical signal must be converted back into an electrical current — a process that hinges on quantum mechanics and semiconductor physics. This section explains how photodiodes and transimpedance amplifiers (TIAs) work, what their physical limits are, and how a same- versus secondary-wavelength pilot interacts with these limits.

6.1 Photoelectric conversion fundamentals

The photoelectric effect in semiconductors

When a photon with energy E = hν (where h is Planck’s constant and ν is optical frequency) strikes a semiconductor, it can excite an electron from the valence band to the conduction band — provided E exceeds the bandgap energy Eg. For 1550 nm light (≈0.8 eV), this means we need materials with Eg < 0.8 eV, such as InGaAs, which is the standard material for infrared photodiodes.

The number of photons converted to electrons per unit time defines the responsivity R(λ):

R(λ) = I_ph / P_opt = η q / (hν)

where η is the quantum efficiency (fraction of photons that generate carriers) and q is electron charge. Typical InGaAs responsivity at 1550 nm is 0.8–1.0 A/W.

6.2 Photodiode types and physics limits

PIN photodiode

A p–i–n structure includes a lightly doped intrinsic region between p and n layers. The electric field in the intrinsic layer quickly sweeps photo-generated carriers to the contacts, producing current proportional to incident optical power.

Key tradeoffs:

Speed vs. sensitivity: Thinner intrinsic region → faster carrier transit → higher bandwidth, but smaller active volume → less responsivity.

Capacitance: The depletion capacitance (Cj) limits the RC bandwidth. Bandwidth f₃dB ≈ 1/(2πR_L Cj). Typical high-speed PDs have 0.1–0.2 pF junction capacitance.

Avalanche photodiode (APD)

An APD adds a high-field multiplication region. Each carrier can create secondary carriers via impact ionization, yielding internal gain M (often 5–10×). However, multiplication noise (excess noise factor F(M)) and temperature sensitivity make APDs more complex.

Saturation and linearity

When the optical power is too high, carriers accumulate faster than they can be swept out, distorting the current waveform. This space-charge effect flattens the response — the PD no longer outputs a current proportional to light. Saturation current is usually tens of milliamps for high-speed PDs. Beyond that, signal distortion and heating occur.

6.3 Transimpedance amplifier (TIA) design

The PD’s photocurrent (microamps to milliamps) is converted to voltage by the TIA, which must provide wide bandwidth (~1 GHz), low noise, and large dynamic range.

Core elements

Feedback resistor (Rf): sets transimpedance gain (Vout/Iin). Typical range: 1–10 kΩ for 1 Gb/s systems.

Input noise: includes resistor Johnson noise, amplifier current noise, and shot noise from the PD. Total input-referred current noise often <10 pA/√Hz.

Gain–bandwidth tradeoff: Higher Rf gives more gain but lowers bandwidth due to the RC pole with PD capacitance.

Automatic gain control (AGC): Some TIAs vary gain to handle signal fading (common in FSOC to deal with turbulence).

Dynamic range

TIAs have a linear range — typically a few hundred millivolts at output. If the photocurrent drives the amplifier beyond this, clipping or compression occurs. Optical overdrive can thus introduce distortion and bit errors, even if the photodiode itself is unsaturated.

6.4 Noise sources and SNR

The total noise current (in) combines several components:

in² = 2qIdcB + 4kTB/Rf + i_amp²B

where:

2qIdcB: shot noise from dark current and signal current.

4kTB/Rf: thermal noise from feedback resistor.

i_amp²B: amplifier internal current noise.

For high SNR, the signal photocurrent (Isig = RPopt) must dominate these noise terms. In FSOC, optical fades cause Isig to fluctuate, temporarily reducing SNR and increasing bit-error rate (BER).

6.5 Pilot interaction with the receiver

Same-wavelength pilot

When a pilot tone rides on the same carrier, the PD sees:

I(t) = I_data(t) [1 + m_p cos(2π f_p t)]

After square-law detection, the mixing term between the pilot and the NRZ data produces sum and difference frequencies. Low-frequency components (below ~10 MHz) fall into the TIA’s passband, appearing as baseline ripple or amplitude modulation noise. The TIA must either:

Include a notch filter at the pilot frequency, or

Handle the pilot ripple in its AGC loop.

Each adds group-delay distortion and small phase shift near the notch frequency, which contributes to deterministic jitter at the slicer.

If the optical power is high (since eye-safety isn’t limiting), the pilot’s low-frequency modulation can also drive the PD/TIA pair closer to compression, reducing linear headroom for data peaks.

Secondary-wavelength pilot

When the pilot arrives on its own wavelength, it is separated by a WDM filter before the PD. The data PD/TIA chain never sees the pilot — so no extra ripple, no notch filters, and no group-delay issues. The pilot signal goes to a separate, low-bandwidth PD optimized for high sensitivity and alignment feedback.

This allows completely independent gain, bandwidth, and filter design between the data and control channels.

Practical note

In same-λ systems, engineers sometimes intentionally overdesign the TIA linear range and bandwidth (e.g., 2× margin) to ensure the pilot ripple doesn’t distort the data. This adds cost and power consumption but can make the single-wavelength approach feasible.

6.6 Temperature and drift considerations

Both PD responsivity and TIA gain vary with temperature:

InGaAs PDs: Responsivity changes about –0.2%/°C.

TIA electronics: Gain drifts due to resistor and transistor temperature coefficients.

Over wide temperature swings (–40 to +85 °C), these effects can cause 1–2 dB gain error unless compensated with calibration or feedback.

Pilot signals (either same or separate wavelength) can help here: the pilot amplitude can serve as a calibration reference to normalize received power across temperature and component aging.

6.7 Summary: Receiver front-end

Feature Same-wavelength pilot Secondary-wavelength pilot
Electrical noise coupling Pilot adds low-frequency AM ripple to TIA input Pilot isolated on separate PD/TIA
Optical overdrive risk Shared between data and pilot Independent power control
Filtering complexity Needs notch or DSP compensation Simpler analog chain
Alignment feedback bandwidth Limited by data TIA BW Tunable separately
Calibration role Can use pilot amplitude as bias reference Easier to isolate, lower impact

Up next: Section 7 — AC Coupling, DC Balance, and Clock Recovery, where we’ll explore how data encoding interacts with receiver coupling, PLLs, and why pilot tones can cause baseline wander or CDR instability in high-speed FSOC links.

7. AC Coupling, DC Balance, and Clock Recovery

By the time the signal reaches the receiver’s analog front end and is amplified, it must be passed through coupling and recovery stages that prepare it for digital slicing. These circuits are highly sensitive to DC levels and timing. This section explains what AC coupling and DC balance are, how clock recovery works, and why adding a pilot tone at the same wavelength can cause baseline wander and jitter.

7.1 AC coupling — blocking DC offsets

Most high-speed optical receivers use AC coupling capacitors between amplifier stages. These capacitors block DC offsets and slow drifts from bias circuits, preventing them from shifting the amplifier’s operating point. The result is a high-pass filter with a cutoff frequency (f_c = 1/(2πRC)), where R is the input resistance of the following stage and C is the coupling capacitance.

In practice, (f_c) is chosen to be well below the data rate (e.g., 10–100 kHz for 1 Gb/s) so that it passes the gigabit data spectrum (hundreds of MHz) but rejects DC components.

However, this means slow variations in average signal level — such as long runs of identical bits or low-frequency amplitude modulation — will be attenuated and appear as baseline wander (slow drift of the eye’s vertical center).

7.2 DC balance and 8b/10b encoding

To minimize baseline wander, high-speed links enforce DC balance — roughly equal numbers of 1s and 0s over short time intervals. The 8b/10b encoding scheme used in 1000BASE-X ensures this by mapping each 8-bit data byte into a 10-bit symbol with controlled disparity (difference between ones and zeros).

This DC balance keeps the average optical power — and therefore the receiver’s average photocurrent — constant, allowing AC coupling to maintain a steady baseline. It also guarantees enough bit transitions (0→1 or 1→0) so the clock and data recovery (CDR) circuit can lock to the data timing.

7.3 Clock and Data Recovery (CDR)

The CDR reconstructs the original clock by aligning a local oscillator to the transitions of the incoming data stream. A phase-locked loop (PLL) adjusts its phase to minimize the time difference between expected and actual data transitions.

PLL structure and dynamics

Phase detector: compares incoming data transitions to local clock edges.

Loop filter: averages out noise and determines how fast the loop responds (bandwidth typically tens of MHz).

Voltage-controlled oscillator (VCO): generates the recovered clock.

The recovered clock is then used to sample data at the optimal point in each bit period — the center of the eye.

If the signal’s transition density drops (e.g., too many consecutive identical bits), the PLL loses phase reference and drifts, increasing jitter.

7.4 How pilot tones interfere with coupling and CDR

Same-wavelength pilot

Adding a low-frequency pilot (kHz–MHz range) amplitude-modulates the optical signal, changing its average level over time. After AC coupling, these slow variations cause baseline wander: the effective “zero” level of the data eye moves up and down, squeezing the vertical opening.

In the frequency domain, the pilot sits near or below the AC coupling cutoff. Some of it is filtered, but the remaining component introduces slow amplitude oscillations. The CDR interprets these as pattern-dependent jitter because the data threshold shifts while the PLL tries to follow it.

If the pilot frequency is close to the PLL’s loop bandwidth, it can directly modulate the recovered clock phase, creating periodic jitter sidebands on the recovered data timing.

Secondary-wavelength pilot

Because the data path is unaffected by the pilot, its AC coupling and CDR circuits see only pure 8b/10b data. The baseline remains stable, and the PLL locks to consistent transitions without low-frequency interference. The pilot’s own receiver can use a slow PLL or digital demodulator tuned to its kHz-range modulation for tracking or telemetry, fully independent of the data path.

7.5 Engineering trade-offs

Concept Same-wavelength pilot Secondary-wavelength pilot
Baseline wander Introduced by low-frequency pilot ripple None — pilot isolated
Transition density Unaffected by pilot, but eye center moves Unaffected and stable
PLL/CDR stability Can introduce periodic jitter near loop BW Stable recovered clock
Coupling design Requires careful RC and notch placement Simpler and standard
Implementation cost Lower Higher (extra optics and PDs)

7.6 Physical intuition

To visualize the effect: imagine the data eye on an oscilloscope. The pilot adds a slow sinusoidal tilt to the eye’s baseline. Every few microseconds, the eye slides up and down. If you freeze the scope at random moments, the slicing level shifts — that’s baseline wander. The PLL, trying to keep up, injects timing shifts, which become jitter.

When the pilot is moved to a separate wavelength, the data eye becomes rock-solid again — no tilt, no baseline breathing. All control information now travels on an optically isolated carrier that can be processed at a slower rate, leaving the gigabit timing chain pristine.

Up next: Section 8 — Equalization and ISI Mitigation, where we’ll cover how limited analog bandwidth blurs bit transitions (inter-symbol interference) and how equalizers restore the eye — plus how pilot modulation interacts with these filters.

8. Equalization and ISI Mitigation

At 1.25 GBd, even small analog imperfections blur consecutive bits together. This section explores how limited bandwidth, parasitics, and filtering produce inter‑symbol interference (ISI) and how engineers counter it through equalization — and how a pilot channel interacts with those equalizers.

8.1 What causes ISI

Bandwidth limitation

Every amplifier, laser driver, and TIA behaves like a low‑pass filter. When the cutoff frequency approaches the data rate, the edges of each bit slow down. Instead of clean, rectangular pulses, you get overlapping transitions where one bit bleeds into the next.

Group delay distortion

Real filters rarely have perfectly flat phase. The phase response bends near the cutoff, meaning different frequency components of a pulse arrive at slightly different times. This group delay variation shifts edges and introduces deterministic jitter.

Reflections and impedance mismatch

In high‑speed traces (like the SFP differential lines or PCB microstrip feeding the laser), any impedance mismatch creates reflections. Those echoes superimpose delayed copies of previous bits — another form of ISI. Good 100 Ω differential control and short stubs are mandatory.

8.2 Equalization techniques

Continuous‑time linear equalizer (CTLE)

A CTLE is an analog filter (usually a high‑frequency boost network) that compensates for low‑pass roll‑off. By adding a small zero above the signal band, it tilts the frequency response upward, restoring high‑frequency components lost through the channel.

Feed‑forward equalizer (FFE)

An FFE applies weighted, delayed versions of previous bits to cancel post‑cursor ISI. Implemented in DSP or analog tap circuits, it corrects predictable distortion patterns.

Decision‑feedback equalizer (DFE)

A DFE uses past detected bits to subtract their estimated ISI contribution from the current sample. It’s powerful for cleaning long tails of ISI but sensitive to error propagation.

8.3 How ISI shows up in FSOC

Although FSOC avoids fiber dispersion, the electrical parts still limit bandwidth:

Laser driver and PD/TIA: finite rise/fall times (~100–200 ps) mean energy from one bit spills into the next.

Long cables or flex between SFP and optics: even a few centimeters add parasitic capacitance and reflections.

Filtering for pilot tones: if a notch or low‑pass is inserted to separate the pilot, it adds more group‑delay ripple.

ISI manifests as eye closure — the eye height (amplitude margin) shrinks and the eye width (timing margin) narrows. At 1 Gb/s, 100 ps of extra delay or 10% amplitude tilt can push BER from 10⁻¹² to 10⁻⁶.

8.4 Pilot tone interaction with equalizers

Same‑wavelength pilot

Frequency interference: The pilot adds a low‑frequency component that equalizers weren’t designed to handle. CTLE/FFE structures may over‑boost or under‑boost around the pilot frequency, introducing phase ripple into the data band.

Notch side‑effects: If a notch filter removes the pilot before equalization, its steep slope causes phase distortion near its corner frequency. The CTLE then amplifies that distortion, worsening deterministic jitter.

Adaptive algorithms: In DSP‑based equalizers, adaptation loops assume white data spectra; the pilot’s periodic energy can trick them into mis‑adjusting tap weights.

Secondary‑wavelength pilot

Because the pilot is filtered optically before detection, the data equalizer sees a clean NRZ signal. There’s no need for pre‑notching or compensation, so equalization converges normally. The pilot’s own low‑bandwidth receiver can have a simple integrator or slow AGC rather than a GHz‑class equalizer.

8.5 Engineering perspective

Effect Same‑wavelength pilot Secondary‑wavelength pilot
ISI coupling Pilot ripple alters equalizer adaptation No effect on data EQ
Filter design Requires precise notch and flat group delay Standard wideband chain
DSP convergence Disturbed by deterministic pilot tone Stable convergence
Implementation cost Lower Higher (extra optics)
Data eye margin Slightly reduced due to ripple Maximum possible

8.6 Physical intuition

Think of the data spectrum as a wide plateau from 0 Hz to ~700 MHz. Equalizers flatten that plateau. A same‑λ pilot adds a tall narrow spike at, say, 500 kHz. To remove it, you carve a notch — but that notch also dents the plateau edges, warping timing. The secondary‑λ method moves the spike to another optical band entirely, leaving the data plateau untouched.

Up next: Section 9 — Automatic Gain Control (AGC) and Dynamic Range, where we’ll explain how receivers manage power swings from atmospheric fades, and how pilot placement influences gain loops and saturation behavior.

9. Automatic Gain Control (AGC) and Dynamic Range

As the FSOC beam travels through turbulent air, its received power can fluctuate by orders of magnitude. To maintain a constant signal amplitude and avoid saturating or underdriving subsequent stages, receivers use Automatic Gain Control (AGC). This section explains the physics of dynamic range, how AGC circuits work, and how pilot placement affects gain stability.

9.1 Dynamic range and link fluctuations

Why optical power fluctuates

Turbulence, scintillation, and pointing jitter cause the received optical power to vary over time. In free space, a 20–30 dB fade range is typical over seconds in moderate conditions. The receiver must still produce a stable logic-level signal for the CDR.

Definition of dynamic range

The receiver’s dynamic range (DR) is the ratio between its maximum linear input power and its minimum detectable power (MDP). For optical receivers:

DR = 10 log10 (Pmax / Pmin)

Typical values: 30–40 dB. Beyond the upper limit, the PD/TIA saturates; below the lower limit, noise dominates.

9.2 AGC principles

Analog AGC loops

An AGC loop senses the output amplitude (from the TIA or limiting amplifier) and adjusts gain to keep it constant. It usually consists of:

Detector: rectifies the output to a DC level proportional to signal amplitude.

Error amplifier: compares to a reference level.

Variable-gain element: adjusts the TIA or postamp gain.

The control bandwidth of the AGC loop (often 1–100 kHz) must be slower than data transitions (GHz) but fast enough to track fades.

Digital AGC

In DSP-based receivers, the AGC can adjust scaling factors in real time. It’s common in software-defined modems or FPGAs with ADC front-ends.

9.3 Interaction between AGC and pilot tones

Same-wavelength pilot

Because the pilot modulates the same optical power as the data, the AGC sees both together. Depending on loop bandwidth, the pilot can beat with the AGC:

If the pilot frequency lies within the AGC loop bandwidth, the loop attempts to cancel it, producing gain oscillations.

If just outside, the AGC partially tracks it, leading to amplitude ripple or compression.

During strong fades, the AGC cranks up gain, but when the pilot reappears at high amplitude, it can drive the loop into saturation or oscillation.

To mitigate this, designers restrict the AGC bandwidth to be well below the pilot frequency (e.g., AGC cutoff ~100 Hz, pilot ~10 kHz) so the pilot appears as a steady average rather than a modulation.

Secondary-wavelength pilot

The data AGC now only tracks the data signal; the pilot has its own low-speed receiver and AGC loop optimized for slow changes (alignment, temperature drift). This decoupling makes control much simpler — no crosstalk between gain loops.

It also allows independent dynamic range optimization: the data channel can saturate slightly during bright bursts while the pilot receiver stays linear, preserving tracking data for feedback.

9.4 Saturation physics in PD/TIA stages

When the received optical power exceeds the PD/TIA’s linear range, the instantaneous photocurrent drives the amplifier into compression. The physics behind this:

Space-charge limitation: high carrier density reduces the electric field in the PD’s depletion region, slowing collection.

Amplifier headroom: TIAs have finite supply voltage; large currents push the output into rail limits.

Thermal drift: high photocurrent causes local heating, raising dark current and lowering responsivity.

Even short overdrive events can cause baseline shifts as the AGC or bias circuits recover. The result is transient jitter and burst errors.

A same-λ pilot increases total average power at the PD, slightly reducing available headroom. A secondary-λ pilot uses a separate PD, so each operates closer to its optimal bias point.

9.5 Design strategies for wide dynamic range

Logarithmic amplifiers: use nonlinear gain compression to extend DR without feedback loops.

Dual-gain TIAs: switch gain ranges based on input power (fast AGC equivalent).

Optical attenuation control: motorized variable attenuators or liquid crystal shutters can stabilize input power.

Digital gain correction: measure average amplitude in the FPGA/ASIC and scale accordingly.

9.6 Engineering comparison

Feature Same-wavelength pilot Secondary-wavelength pilot
AGC stability Possible beating with pilot frequency Independent, stable loop
Gain range interaction Shared gain dynamics between pilot and data Isolated per channel
Overdrive recovery Longer recovery (shared PD/TIA) Localized recovery (data only)
Implementation simplicity One optical path Two paths, more components
Robustness in turbulence Moderate High (decoupled loops)

Up next: Section 10 — Alignment Sensing and Control Loops, where we’ll integrate everything: how the pilot (same or secondary λ) drives feedback to mirrors or gimbals for beam tracking, and how those control loops interact with the data link.

10. Alignment Sensing and Control Loops

Even with perfect electronics, FSOC links fail if the beam isn’t pointed precisely at the receiver. Alignment systems keep the optical axes of both nodes locked on each other, compensating for drift, vibration, and atmospheric wander. This section explains how those control loops work and how the pilot channel choice affects them.

10.1 The pointing problem

Beam geometry and alignment tolerance

With a divergence of 0.1 mrad over a 100 m link, the beam spot is only ~10 mm wide. A lateral misalignment of even a few millimeters (tens of microradians) can cause severe signal loss. Since buildings flex, mounts expand, and air shimmers, closed-loop pointing correction is mandatory for stable FSOC links.

Sources of misalignment

Mechanical drift: thermal expansion or vibration shifts optics.

Atmospheric refraction: temperature gradients bend the beam path.

Turbulence: introduces random beam wander on the order of micro- to milliradians.

Platform motion: for mobile or drone-based FSOC, attitude errors dominate.

10.2 Alignment sensing architectures

Quadrant photodiode (QPD)

A QPD has four segments. By comparing photocurrents (A, B, C, D), you get horizontal and vertical error signals:

[ E_x = (A + D) – (B + C) ]
[ E_y = (A + B) – (C + D) ]

When the beam is centered, both errors are zero. Offsets in either axis drive actuators until balance is restored.

Position-sensitive detector (PSD)

A continuous analog detector that outputs a voltage proportional to the beam centroid position. Useful for fine, continuous tracking.

Focal-plane array (FPA) camera

Provides full 2D imaging for coarse acquisition. Used to find the partner node initially, then hand off to faster sensors like QPDs.

10.3 Actuation mechanisms

Gimbal or pan-tilt stage

Larger optics often use a two-axis motorized gimbal for coarse alignment. Response bandwidths are low (1–10 Hz), suitable for slow drift.

MEMS mirror

Micro-electromechanical mirrors provide fine, high-speed steering (up to kHz). They correct for turbulence and small vibrations.

Fast steering lens (piezo or voice coil)

Moves the optical axis through refractive shift; less common but effective for compact heads.

10.4 Control loop structure

A typical auto-alignment system is a cascaded control loop:

Error sensing: QPD measures beam displacement.

Signal conditioning: low-pass filter removes scintillation and noise.

Controller: PID or lead-lag compensator computes correction.

Actuator: mirror/gimbal re-centers beam.

Loop bandwidths are usually 50–500 Hz — fast enough to counter building sway and low-frequency turbulence.

Stability considerations

The control loop must avoid chasing scintillation (fast amplitude noise). Filtering out frequencies above the Greenwood frequency (~tens of Hz) prevents overreaction to transient speckle motion.

10.5 Pilot-driven feedback

Same-wavelength pilot

The pilot tone is embedded in the main beam; the QPD detects both data and pilot light simultaneously. You can demodulate the pilot frequency component from the QPD signal to isolate the alignment data. This approach guarantees perfect co-boresight — the feedback truly represents the same optical path as the data.

However, it adds electronic complexity: the alignment detector must include narrowband demodulation filters to extract the pilot tone, and any pilot-induced amplitude modulation slightly modulates the data intensity. Strong turbulence may also cause pilot amplitude fluctuations that couple into the data path.

Secondary-wavelength pilot

A secondary wavelength (separate laser) acts as a beacon. A dichroic or WDM filter splits it to a dedicated QPD or camera. This lets the data receiver ignore tracking light entirely, preventing cross-coupling. The pilot beam can be run at higher average power and wider beam diameter for better visibility, while the data beam stays tightly collimated.

The downside is the potential for chromatic misalignment: different refractive indices or coatings cause the pilot and data beams to focus or refract differently by a few microradians. Calibration and temperature compensation minimize this.

10.6 Control-loop latency and interaction with data path

If the same-wavelength pilot is used, its alignment corrections directly influence the optical axis of the data — but any delay or phase lag in the control loop can cause beam jitter, which appears as received amplitude modulation in the data signal. Proper loop tuning (phase margin >45°) and damping are essential.

In the secondary-wavelength case, the alignment loop runs independently, but because the beams are nearly co-linear, its corrections indirectly stabilize the data beam with slightly lower precision. The separation of feedback paths makes the system more modular and easier to debug.

10.7 System-level trade-offs

Aspect Same-wavelength pilot Secondary-wavelength pilot
Alignment accuracy Perfect co-boresight Slight chromatic offset (correctable)
Control complexity Requires demodulation and shared optics Separate optical chain, simpler control
Noise coupling Pilot modulation seen in data None in data chain
Beacon visibility Limited to data power budget Can run at high power
Mechanical simplicity Single aperture Two-beam combination needed

10.8 Practical example

In a 100 m rooftop link:

A QPD receives both data and pilot at 1550 nm.

The pilot is a 10 kHz subcarrier at 1% modulation depth.

PID control runs at 200 Hz bandwidth to correct drift.

If turbulence increases, the pilot amplitude fluctuates. The AGC compensates for average power, but residual amplitude modulation still adds <0.1 dB intensity noise to the data. Switching to a secondary 1552 nm pilot with a separate detector eliminates this entirely, allowing 3–5 dB more SNR margin.

Up next: Section 11 — Manufacturing, Calibration, and Test, where we’ll discuss alignment calibration procedures, pilot depth verification, and how each architecture scales for production and field maintenance.

11. Manufacturing, Calibration, and Test

FSOC systems are precision optoelectronic instruments. Once the optical, electrical, and control subsystems are defined, the practical challenge becomes building them repeatedly and verifying performance under real conditions. This section explains how manufacturing calibration works and how same-wavelength versus secondary-wavelength pilots affect production complexity and long-term maintainability.

11.1 Optical alignment and calibration procedures

Coarse and fine alignment

Coarse alignment: mechanical assembly positions the laser, collimator, and receiver apertures within tens of microradians using jigs or reference targets.

Fine alignment: uses a QPD or FPA camera with live feedback while maximizing received optical power.

Lock-in detection: if a pilot tone is present, a synchronous detector measures pilot amplitude to guide the alignment process — extremely sensitive for fine tuning.

For same-wavelength pilots, this lock-in approach is inherent; the pilot tone is already modulating the data laser. For secondary-wavelength pilots, the alignment system uses a separate beacon detector and WDM splitter.

Co-boresight calibration

When using two wavelengths, mechanical fixtures and alignment microscopes ensure both lasers (data and pilot) share the same optical axis. During factory calibration, a far-field camera measures the relative beam centers at multiple distances. Any offset >10 µrad is corrected with tip/tilt adjustments or dichroic alignment screws.

11.2 Electrical calibration

Bias and modulation current setup

Laser bias (I_B): adjusted above threshold for desired output (e.g., 10 mA + 20%).

Modulation current: tuned for target extinction ratio (6–9 dB typical).

Pilot injection: set for 1–2% modulation index, verified with spectrum analyzer or optical oscilloscope.

Receiver chain characterization

Dark current measurement: ensures PD leakage <1 nA.

TIA gain and bandwidth: verified with modulated light source or network analyzer.

AGC calibration: adjust reference level so output amplitude is constant for ±20 dB input power swing.

For secondary-λ systems, both data and pilot chains undergo these tests separately, each optimized for its power and bandwidth target.

11.3 Environmental and vibration testing

Thermal testing

Systems are cycled from –40°C to +85°C. Laser output power, wavelength, and receiver sensitivity are logged. The goal is <1 dB performance drift. Secondary-wavelength systems may require recalibration of chromatic alignment after large thermal swings.

Vibration and shock

Gimbals and optics are tested with random vibration profiles up to several g RMS.

QPD alignment loops are observed for control stability under vibration.

Same-wavelength pilots simplify these tests since there’s only one beam to monitor; dual-wavelength systems must verify both beams remain coaligned.

Atmospheric emulation

FSOC links are tested in turbulence chambers or long optical paths with controlled refractive index fluctuations (heated air columns). Pilot-driven control loops are exercised to validate tracking algorithms.

11.4 Production and field test metrics

Metric Typical Specification Test Method
Optical output power 10–100 mW Optical power meter
Extinction ratio 6–9 dB Optical oscilloscope or BER tester
Pilot modulation index 1–2% Spectrum analyzer at pilot frequency
Beam divergence <0.2 mrad Far-field camera
Pointing accuracy <20 µrad QPD readout
BER @ 1 Gb/s <10⁻¹² BERT analyzer

11.5 Maintenance and field recalibration

Over time, mechanical creep or thermal cycling may shift alignment by a few microradians. Maintenance procedures typically include:

Optical power check: verify received power vs. baseline.

Pilot amplitude verification: ensures proper beacon strength.

Auto-realignment sequence: gimbal performs slow raster scan to re-maximize pilot signal.

Bias and gain recalibration: laser bias current and AGC reference are readjusted for nominal SNR.

In same-λ systems, recalibration is faster — one detector and beam. In secondary-λ systems, pilot alignment is verified first, then data alignment fine-tuned.

11.6 Manufacturing trade-offs

Aspect Same-wavelength pilot Secondary-wavelength pilot
Optical alignment Simpler (one beam) Dual-beam co-boresight required
Electrical test One driver chain Two independent laser drivers
Calibration steps Fewer More (dual wavelength setup)
Production time Faster Slower (~1.2–1.5× longer)
Field servicing Easier (single loop) More robust but complex

11.7 Long-term reliability

Both architectures can meet telecom-grade MTBFs (>100,000 hours) if designed properly. However:

Same-λ pilots concentrate optical stress on one diode, potentially shortening lifetime under high power.

Secondary-λ systems spread thermal load across two lasers, improving lifetime at the cost of more parts.

Thermal management, clean optics, and sealed enclosures dominate reliability far more than pilot architecture.

Up next: Section 12 — System-Level Comparison and Final Recommendations, where we’ll consolidate all trade-offs across subsystems (optical, electrical, and control) to determine which pilot approach offers the best balance of robustness, manufacturability, and performance.

12. System-Level Comparison and Final Recommendations

This final section consolidates all the preceding discussions — optical, electrical, control, and manufacturing — to evaluate both pilot-channel strategies at the system level. The goal is to understand which approach provides the best combination of performance, robustness, and scalability for a 1 Gb/s FSOC link built from SFP-derived electronics.

12.1 Subsystem comparison summary

Subsystem Same-wavelength pilot Secondary-wavelength pilot
Laser & Driver Single laser; small AM dither affects ER and RIN; risk of intermodulation Two independent lasers; data chain unaffected; higher part count
Optical Path Simplest optics; perfect co-alignment Needs WDM or dichroic; small chromatic offset risk
Receiver Front-End Shared PD/TIA; potential compression and pilot ripple Separate PD/TIA; isolated, clean data path
Equalization & CDR Pilot ripple can add deterministic jitter and baseline wander No interference; data CDR fully stable
AGC & Dynamic Range Pilot interacts with gain loop; possible beating Decoupled loops; stable under fades
Alignment & Control Perfect co-boresight but shared feedback; coupling into data possible Independent beacon feedback; higher optical complexity
Manufacturing & Calibration Fast assembly; simpler test Longer alignment and test time (dual beams)
Field Reliability Fewer parts, but higher risk of mutual interference More parts, but cleaner modular operation
Power Budgeting Pilot steals small % of optical power Full data power preserved; pilot power adjustable

12.2 Performance analysis

Signal integrity

At 1 Gb/s, the same-wavelength pilot must be extremely shallow (<2% modulation index) to avoid closing the data eye. Even that small ripple can contribute 0.1–0.3 dB penalty in SNR and visible deterministic jitter. The secondary-wavelength pilot completely removes this effect, preserving the clean 8b/10b eye diagram.

Robustness

In atmospheric turbulence or strong fades, the ability to crank pilot power without touching the data path makes the secondary-λ system more robust. Even if the data link momentarily saturates or drops, the pilot receiver can maintain alignment lock for rapid reacquisition.

Manufacturability

The same-λ design wins here: fewer lasers, no WDM alignment, single calibration chain. For small-scale or low-cost prototypes, this simplicity is valuable. The secondary-λ approach requires precise dichroic assembly and longer calibration but scales better for production volume once fixtures are built.

Power and thermal load

Running two lasers doubles electrical and thermal load, but allows each to operate at lower stress for longer lifetime. A single high-power diode (same-λ) runs hotter and must dissipate more heat to maintain wavelength stability.

12.3 System-level trade-offs

Attribute Same-wavelength pilot Secondary-wavelength pilot
Complexity Lowest Higher (extra components)
Optical isolation None (shared) Full isolation
Performance stability Moderate; sensitive to tuning High; independent control
Jitter/eye penalty 0.1–0.3 dB None measurable
Manufacturing effort Simple (single path) 20–50% longer calibration
Field calibration Fast (single loop) Requires dual verification
Resilience in turbulence Moderate Excellent
Scalability to higher data rates (>2.5 Gb/s) Limited by intermodulation Scales cleanly

12.4 Recommended approach by application

Application Type Optimal Pilot Scheme Rationale
Low-cost short links (<100 m) Same-wavelength pilot Simplest architecture; easy to manufacture and align; sufficient SNR margin
Mid-range industrial links (100–500 m) Secondary-wavelength pilot Better stability in temperature and turbulence; manageable alignment complexity
Long-range or moving platforms (>500 m) Secondary-wavelength pilot Independent beacon tracking essential; higher optical power flexibility
Lab development or proof-of-concept Same-wavelength pilot Fast integration with minimal optics; good for demonstration
Telecom-grade deployment Secondary-wavelength pilot Robustness, scalability, and lower long-term jitter outweigh extra parts

12.5 Engineering conclusion

At the fundamental level, both designs transmit the same 1.25 GBd NRZ waveform through air. The difference lies in how cleanly they separate control and data physics. The same-wavelength approach merges everything — data, pilot, and feedback — into one optical carrier. This minimizes cost but couples all noise and control loops together. The secondary-wavelength approach explicitly separates these domains, allowing each to be optimized independently.

For a system derived from an SFP media converter, where the electrical interface already provides precise timing and the challenge is optical stability, the secondary-wavelength pilot architecture is the more robust choice. It adds modest optical complexity but yields a cleaner, more maintainable, and higher-performance FSOC link — especially when eye-safety is not a limiting factor.

12.6 Closing perspective

A pilot channel in FSOC is not just a telemetry feature; it’s a bridge between optical physics and control systems. Whether implemented as a subtle subcarrier or a dedicated beacon, it transforms a static laser link into a self-stabilizing, adaptive communication node. The engineering decision between same- and secondary-wavelength pilots comes down to whether your system prioritizes simplicity or isolation. In the long run, as data rates and environmental demands rise, wavelength separation remains the path to higher robustness, scalability, and precision.

Executive Summary and System Overview

Free-Space Optical Communication (FSOC) at 1 Gb/s pushes electronic, optical, and mechanical systems to their limits. This document walked through each stage — from digital signaling physics to beam alignment and calibration — to evaluate how same-wavelength and secondary-wavelength pilot channels affect real-world FSOC performance.

Summary of Core Insights

Digital & Optical Fundamentals: FSOC systems carrying 1.25 GBd NRZ data require ~1 GHz of clean bandwidth, tight jitter control (<250 ps), and precise optical alignment (tens of microradians). Removing eye-safety limits boosts SNR but doesn’t eliminate receiver linearity or jitter constraints.

Laser & Driver Physics: A same-wavelength pilot adds a small amplitude ripple to the main laser current. Even at 1–2% modulation, this can cause measurable eye closure. A secondary wavelength avoids this entirely by using a separate DFB and WDM.

Atmospheric Propagation: Turbulence and scintillation dominate FSOC loss. Independent pilot control allows the beacon to run brighter and wider for robust tracking, while data remains narrow and optimized for bandwidth.

Receiver & CDR Integrity: Shared-wavelength pilots can inject low-frequency noise, notch-filter distortion, and baseline wander into the CDR loop. Separate wavelengths maintain isolation and timing purity.

AGC & Control Stability: Same-λ pilots share gain dynamics with the data signal, risking AGC oscillations. Separate pilots use independent AGC and feedback, simplifying tuning.

Alignment & Manufacturing: One beam is easier to build and calibrate; two beams yield better stability and diagnostics but require more optical fixtures and test time.

Strategic Recommendation

For experimental or low-cost systems, a same-wavelength pilot offers simplicity and rapid prototyping. For long-range, high-availability, or scalable FSOC networks, a secondary-wavelength pilot is decisively superior — enabling independent optimization of data fidelity, tracking robustness, and thermal stability.

In other words: use one laser to prove the concept; use two to deploy a product.

System Architecture Diagram (Conceptual Overview)
        ┌───────────────────────────────┐
        │        Transmitter Node       │
        │                               │
        │  ┌──────────────┐             │
SFP TX → │  Laser Driver  │───► DFB @1550 nm (Data) ─┐
        │  (Bias + Mod.)  │             │             │
        │  └──────────────┘             │             │
        │                               │     ┌───────▼────────┐
        │                     Optional  │     │ Collimator /   │
        │   ┌──────────────┐   Pilot    │     │ Beam Expander  │
        │   Pilot Driver   │──► DFB @1552 nm ─┘  │   │
        │   (Low-rate Mod.)│                     │   │
        │   └──────────────┘                     │   │
        │                                       └───┘
        │                               Outgoing Free-space Beam
        └───────────────────────────────┘

                        ▼
              (Atmospheric Path)
                        ▼

        ┌───────────────────────────────┐
        │         Receiver Node         │
        │                               │
        │   ┌──────────────┐            │
        │   Data PD + TIA  │◄─── WDM ───┤
        │   (1 GHz BW)     │            │
        │   └──────────────┘            │
        │   ┌──────────────┐            │
        │   Pilot PD + TIA │◄───        │
        │   (kHz BW)       │            │
        │   └──────────────┘            │
        │                               │
        │  Control Loops → MEMS Mirror  │
        │  for Fine Alignment            │
        └───────────────────────────────┘
Closing Perspective

FSOC design sits at the intersection of photonics, RF engineering, and control theory. Every millivolt, microradian, and microsecond matters. The decision to separate pilot and data wavelengths is more than an optical choice — it defines how modular, debuggable, and scalable the system becomes.

The same-wavelength pilot represents elegance through simplicity. The secondary-wavelength pilot represents precision through separation. The right answer depends on whether your challenge is to prove that FSOC works — or to make it work everywhere, every day.

Visual Summary Diagram — Pilot Architectures Side-by-Side

The diagram below compares the overall data and control paths for the two FSOC pilot approaches. Each block represents a functional subsystem, showing where data, pilot, and feedback signals diverge or combine.
        ┌────────────────────────────────────────────────────────────────────────┐
        │        SAME-WAVELENGTH PILOT ARCHITECTURE (Single 1550 nm Laser)       │
        └────────────────────────────────────────────────────────────────────────┘

        SFP TX Data  ──►  Laser Driver  ──►  DFB Laser @1550 nm  ──►  Collimator
                                         │
                                         ▼
                                   Low-frequency Pilot
                                   (Amplitude Modulation)

        ┌─────────────────────────────── Free-Space Path ───────────────────────────────┐
        │   Atmospheric Turbulence, Beam Wander, Scintillation, Alignment Feedback     │
        └──────────────────────────────────────────────────────────────────────────────┘

        Received Optical Power  ──►  Shared PD/TIA  ──►  AGC  ──►  CDR + Equalizer
                                         │                         │
                                         ▼                         ▼
                              Pilot Demodulation (kHz–MHz)     Recovered 1.25 GBd Data
                                         │
                                         ▼
                            Alignment Feedback → MEMS / Gimbal Control

  ► Advantages: simple optics, perfect co-boresight.
  ► Drawbacks: shared linearity, pilot ripple in data, notch filter complexity.

-----------------------------------------------------------------------------------------

        ┌────────────────────────────────────────────────────────────────────────┐
        │     SECONDARY-WAVELENGTH PILOT ARCHITECTURE (Dual 1550/1552 nm Beams)  │
        └────────────────────────────────────────────────────────────────────────┘

        SFP TX Data ──►  Laser Driver  ──►  DFB Laser @1550 nm  ──►  Collimator / WDM ──┐
                                                                                       │
                                      Pilot Generator ──►  DFB @1552 nm  ─────────────┘

        ┌─────────────────────────────── Free-Space Path ───────────────────────────────┐
        │     Co-propagating but Independent Data & Pilot Beams Through Atmosphere     │
        └──────────────────────────────────────────────────────────────────────────────┘

        ► Data Channel (1550 nm):
             └─► Data PD/TIA (1 GHz BW) ─► AGC ─► CDR ─► Recovered 1 Gb/s Signal

        ► Pilot Channel (1552 nm):
             └─► Pilot PD/TIA (kHz BW) ─► Pilot Demodulator ─► Feedback Controller
                                          │
                                          ▼
                                     MEMS / Gimbal Mirror

  ► Advantages: full isolation, independent gain loops, no data jitter coupling.
  ► Drawbacks: extra lasers, WDMs, and calibration alignment.

-----------------------------------------------------------------------------------------

**Key Insight:**  The same-wavelength design merges data and control in one photonic path, minimizing parts but coupling noise.  
The secondary-wavelength design separates them, maximizing robustness and scalability.
This dual architecture visualization completes the FSOC pilot analysis. Both configurations are functionally proven, but their suitability depends on design priorities:

Simplicity and speed to prototype → Same-wavelength pilot

Performance, reliability, and scalability → Secondary-wavelength pilot

Aspect	Same-wavelength pilot	Secondary-wavelength pilot
Alignment fidelity	Perfect co-propagation	Slight chromatic offset (correctable)
Optical complexity	Minimal	Requires WDMs / dichroics
Noise coupling	Pilot adds amplitude ripple into data	Clean separation between paths
Power control	Shared with data	Independent tuning possible
Thermal sensitivity	Single thermal path	Slight dispersion-induced drift
Implementation cost	Lower	Higher (extra optics, laser, PD)

Effect	Same-wavelength pilot	Secondary-wavelength pilot
Turbulence correlation	Perfect (same photons)	~99% correlated
Fading behavior	Shared fades; pilot adds amplitude modulation to data	Independent detection; pilot unaffected by data load
Recovery after fade	Faster CDR reacquisition, but tighter coupling	Easier alignment recovery; less jitter coupling
Implementation	Simple (no WDM)	More parts, but more robust control

Feature	Same-wavelength pilot	Secondary-wavelength pilot
Electrical noise coupling	Pilot adds low-frequency AM ripple to TIA input	Pilot isolated on separate PD/TIA
Optical overdrive risk	Shared between data and pilot	Independent power control
Filtering complexity	Needs notch or DSP compensation	Simpler analog chain
Alignment feedback bandwidth	Limited by data TIA BW	Tunable separately
Calibration role	Can use pilot amplitude as bias reference	Easier to isolate, lower impact

Concept	Same-wavelength pilot	Secondary-wavelength pilot
Baseline wander	Introduced by low-frequency pilot ripple	None — pilot isolated
Transition density	Unaffected by pilot, but eye center moves	Unaffected and stable
PLL/CDR stability	Can introduce periodic jitter near loop BW	Stable recovered clock
Coupling design	Requires careful RC and notch placement	Simpler and standard
Implementation cost	Lower	Higher (extra optics and PDs)

Effect	Same‑wavelength pilot	Secondary‑wavelength pilot
ISI coupling	Pilot ripple alters equalizer adaptation	No effect on data EQ
Filter design	Requires precise notch and flat group delay	Standard wideband chain
DSP convergence	Disturbed by deterministic pilot tone	Stable convergence
Implementation cost	Lower	Higher (extra optics)
Data eye margin	Slightly reduced due to ripple	Maximum possible

Feature	Same-wavelength pilot	Secondary-wavelength pilot
AGC stability	Possible beating with pilot frequency	Independent, stable loop
Gain range interaction	Shared gain dynamics between pilot and data	Isolated per channel
Overdrive recovery	Longer recovery (shared PD/TIA)	Localized recovery (data only)
Implementation simplicity	One optical path	Two paths, more components
Robustness in turbulence	Moderate	High (decoupled loops)

Aspect	Same-wavelength pilot	Secondary-wavelength pilot
Alignment accuracy	Perfect co-boresight	Slight chromatic offset (correctable)
Control complexity	Requires demodulation and shared optics	Separate optical chain, simpler control
Noise coupling	Pilot modulation seen in data	None in data chain
Beacon visibility	Limited to data power budget	Can run at high power
Mechanical simplicity	Single aperture	Two-beam combination needed

Metric	Typical Specification	Test Method
Optical output power	10–100 mW	Optical power meter
Extinction ratio	6–9 dB	Optical oscilloscope or BER tester
Pilot modulation index	1–2%	Spectrum analyzer at pilot frequency
Beam divergence	<0.2 mrad	Far-field camera
Pointing accuracy	<20 µrad	QPD readout
BER @ 1 Gb/s	<10⁻¹²	BERT analyzer

Aspect	Same-wavelength pilot	Secondary-wavelength pilot
Optical alignment	Simpler (one beam)	Dual-beam co-boresight required
Electrical test	One driver chain	Two independent laser drivers
Calibration steps	Fewer	More (dual wavelength setup)
Production time	Faster	Slower (~1.2–1.5× longer)
Field servicing	Easier (single loop)	More robust but complex

Subsystem	Same-wavelength pilot	Secondary-wavelength pilot
Laser & Driver	Single laser; small AM dither affects ER and RIN; risk of intermodulation	Two independent lasers; data chain unaffected; higher part count
Optical Path	Simplest optics; perfect co-alignment	Needs WDM or dichroic; small chromatic offset risk
Receiver Front-End	Shared PD/TIA; potential compression and pilot ripple	Separate PD/TIA; isolated, clean data path
Equalization & CDR	Pilot ripple can add deterministic jitter and baseline wander	No interference; data CDR fully stable
AGC & Dynamic Range	Pilot interacts with gain loop; possible beating	Decoupled loops; stable under fades
Alignment & Control	Perfect co-boresight but shared feedback; coupling into data possible	Independent beacon feedback; higher optical complexity
Manufacturing & Calibration	Fast assembly; simpler test	Longer alignment and test time (dual beams)
Field Reliability	Fewer parts, but higher risk of mutual interference	More parts, but cleaner modular operation
Power Budgeting	Pilot steals small % of optical power	Full data power preserved; pilot power adjustable

Attribute	Same-wavelength pilot	Secondary-wavelength pilot
Complexity	Lowest	Higher (extra components)
Optical isolation	None (shared)	Full isolation
Performance stability	Moderate; sensitive to tuning	High; independent control
Jitter/eye penalty	0.1–0.3 dB	None measurable
Manufacturing effort	Simple (single path)	20–50% longer calibration
Field calibration	Fast (single loop)	Requires dual verification
Resilience in turbulence	Moderate	Excellent
Scalability to higher data rates (>2.5 Gb/s)	Limited by intermodulation	Scales cleanly

Application Type	Optimal Pilot Scheme	Rationale
Low-cost short links (<100 m)	Same-wavelength pilot	Simplest architecture; easy to manufacture and align; sufficient SNR margin
Mid-range industrial links (100–500 m)	Secondary-wavelength pilot	Better stability in temperature and turbulence; manageable alignment complexity
Long-range or moving platforms (>500 m)	Secondary-wavelength pilot	Independent beacon tracking essential; higher optical power flexibility
Lab development or proof-of-concept	Same-wavelength pilot	Fast integration with minimal optics; good for demonstration
Telecom-grade deployment	Secondary-wavelength pilot	Robustness, scalability, and lower long-term jitter outweigh extra parts

Week 2 – Initial Calculation of SWaP-C

Context

Week 2 Goals

Results

From ‘What is the SWAP-C’ to ‘Is this Feasible’

Auto-Alignment System

Conclusion

Next Week

Appendix

1. Introduction — What 1 Gb/s Free-Space Optical Communication (FSOC) Really Means

Why 1550 nm

Eye Safety: What It Means Physically

If Eye Safety Didn’t Matter

2. Digital Signaling and Timing Physics

What is a Symbol?

What is 1000BASE-X and 8b/10b Encoding?

What is NRZ Signaling?

Symbol Rate vs. Bandwidth

What is an Eye Diagram?

What is Jitter and Timing Stability?

Physics of Fast Transitions — Why It’s Hard

How the Pilot Channel Affects Digital Signaling

3. Laser Physics and Driver Electronics

3.1 How a semiconductor laser makes coherent light

p–n (often p–i–n) junction and recombination

Cavity, modes, DFB vs EML

Key laser parameters

3.2 From bits to photons: bias + modulation

Bias point and extinction ratio (ER)

Rise/fall limits and parasitics

Driver and bias-tee topology

3.3 Noise and linearity through the E/O chain

3.4 What a same‑wavelength pilot does here

3.5 What a secondary‑wavelength pilot does here

3.6 Practical design checklist (Tx side)

4. Optical Launch and Collimation

4.1 Gaussian beams and divergence

Beam fundamentals

Why divergence matters

Beam quality and M²

4.2 Collimation optics and coatings

Lenses and expansion

Anti-reflection (AR) coatings

Alignment tolerances

4.3 Optical launch alignment and pointing control

Static alignment

Dynamic alignment (auto-pointing)

Control loop coupling

4.4 How same-wavelength vs. secondary-wavelength pilots impact optics

Same-wavelength pilot

Secondary-wavelength pilot

Thermal and chromatic considerations

4.5 Optomechanical tolerances and material physics

Thermal expansion

Vibration and damping

Optical cleanliness

4.6 Summary: Optical section pros and cons

5. Propagation Through Atmosphere

5.1 Refractive index variations and turbulence physics

Structure constant (Cn²)

Kolmogorov turbulence spectrum

Fried coherence length (r₀)

5.2 Scintillation and fading

Intensity fluctuations

Beam wander

Temporal dynamics

5.3 Atmospheric absorption and scattering

5.4 Mitigating turbulence effects

Aperture averaging

Adaptive optics

Spatial and temporal diversity

5.5 Pilot vs. secondary-wavelength response in turbulence

Same-wavelength pilot

Secondary-wavelength pilot

Summary of atmospheric interaction

6. Photodiode and Receiver Front-End Physics

6.1 Photoelectric conversion fundamentals

The photoelectric effect in semiconductors

6.2 Photodiode types and physics limits

PIN photodiode