Note · Microwave photonics
How a Mach–Zehnder modulator driver sets the limits of a radio-over-fiber link
A radio-over-fiber (RoF) link carries an RF signal as the intensity of light. You put a microwave signal on an optical carrier at one end, send it through fiber, and recover it with a photodiode at the other end. On paper the fiber is almost lossless and has huge bandwidth. In practice, the link's gain, noise figure, and dynamic range are set mostly by two components: the modulator and the circuit that drives it.
This note walks through why, using the most common setup: a Mach–Zehnder modulator (MZM) with direct detection.
The modulator's transfer curve
An MZM splits light into two arms, shifts the phase of one relative to the other with an applied voltage, and recombines them. The output intensity follows a raised cosine:
Here Vπ is the voltage that swings the output from fully on to fully off, and φ is the bias phase. For an analog link you bias at quadrature (φ = π/2), where the curve is steepest and the second-order distortion cancels. Around that point, a small RF voltage maps to a nearly linear change in light intensity.
Why Vπ dominates the link gain
The slope of the transfer curve at quadrature is proportional to π/Vπ. For an intensity-modulated link with direct detection, the small-signal RF gain scales as
where Idc is the average photocurrent and R is the source and load resistance. The square is what matters: halving Vπ buys 6 dB of link gain, the same as doubling the received optical power. Many passive analog links show net loss, often 20 dB or more, so this improvement has a direct effect on noise figure.
This is also where the driver comes in. In silicon depletion-mode modulators, Vπ·L is typically a few V·cm. A short device has a large Vπ, and a long one has more RF loss and harder velocity matching. The driver has to supply enough swing to use the modulator well without overloading it.
What the driver has to do
| Requirement | Why it matters | Typical design response |
|---|---|---|
| Large output swing | Link gain depends on Vdrive/Vπ | Stacked or cascoded output stages to get past the CMOS breakdown limit |
| Wide, flat bandwidth | The traveling-wave electrode is a lossy transmission line | Distributed or segmented drivers, matched termination |
| Linearity | Driver distortion adds to the modulator's own cosine nonlinearity | Back off from compression, consider predistortion |
| Low noise | Driver noise is amplified by the same link gain as the signal | Low-noise front stage ahead of the power stage |
Dynamic range: where the cosine catches up with you
Quadrature bias cancels even-order distortion, but the cosine still produces third-order intermodulation. That sets the link's spurious-free dynamic range (SFDR), usually quoted in dB·Hz2/3 so you can compare links measured at different noise bandwidths. A driver that compresses before the modulator does will reduce SFDR on its own. A good co-design keeps the driver's third-order intercept well above the modulator's.
Takeaways
The fiber is rarely the limiting part of an RoF link. The modulator's Vπ sets the gain, its cosine response sets the linearity, and the driver decides how much of both you actually get.
Further reading
- C. H. Cox III, Analog Optical Links: Theory and Practice, Cambridge University Press, 2004.
Questions or corrections? Get in touch through the contact links on the contact section.