Probe station

A probe station holds a wafer or chip on a flat chuck under a microscope and carries manipulators that bring electrical probes and optical fibers into contact with, or just above, the device. It makes measurement possible before a device is packaged: lasers, photodiodes, modulators and complete photonic integrated circuits are tested where they were fabricated, with probes landing on pads typically 50–100 µm across. Manual stations are positioned by hand; automatic probers step a wafer die by die under software control, the basis of wafer-level test.

Components

• Chuck. A vacuum chuck holds the wafer or die flat. Chucks are often temperature-controlled by a thermoelectric cooler or a fluid loop, because many photonic quantities drift with temperature: a silicon ring resonance near 1550 nm moves about 67 pm/K, so a chuck held to ±0.1 K still allows ±6.7 pm of drift. It often doubles as the back contact.
• Microscope and camera. A long-working-distance microscope looks down on the probes, and on optical stations a side camera shows the gap between fiber tip and chip surface.
• DC probes. Tungsten or beryllium-copper needles on manipulators, or a probe card with many needles in a fixed pattern, carry bias and read currents through a source-measure unit.
• RF probes. Coplanar ground-signal-ground (GSG) probes with a pitch typically between 100 and 150 µm carry signals to tens of gigahertz from a network analyzer or bit-error tester; the pad layout on the chip must match the probe pitch.
• Optical probes. Single fibers or a fiber array on motorized and piezo stages, angled to match grating couplers for surface coupling, or lensed fibers aligned to the facet for edge-coupled chips in die-level test.

Alignment by peak search

An optical probe is aligned by moving it while monitoring a signal and maximizing it: the power transmitted through a loopback waveguide to an output fiber, or the photocurrent of an on-chip photodiode read through a DC probe. A coarse raster or spiral scan over a window of tens of micrometers finds the first light, and a finer gradient search finds the peak. The search must be finer than the coupling tolerance. For a grating coupler whose beam matches a fiber mode of 10.4 µm diameter, a Gaussian overlap estimate gives 0.04 dB of loss at 0.5 µm of lateral offset, 0.16 dB at 1 µm and 1.0 dB at 2.5 µm, so a final step of a fraction of a micrometer is needed to hold the landing-to-landing scatter near 0.1 dB. Fiber height is set separately, close enough for good coupling and far enough to clear wafer bow and particles. The test flow for a whole wafer is set out in Wafer-Level Optical Testing.

Calibration

Every probe has a response of its own. For RF measurements, the network analyzer is calibrated to the probe tips with an impedance standard substrate, a ceramic chip carrying short, open, load and through structures at the probe pitch, so that the reference plane moves from the instrument cables to the tips. Pad parasitics on the device itself are then removed by de-embedding with open and short dummy pads on the wafer. On the optical side, the loss of fibers, connectors and couplers is removed with reference structures, such as back-to-back grating couplers, measured on the same die.

Repeatability and pitfalls

• Contact resistance. Each landing of a needle produces a slightly different contact, typically a fraction of an ohm on clean gold pads and more on aluminum or oxidized metal. A 0.5 Ω contact on a laser with 5 Ω of series resistance is a 10% error, which four-wire (Kelvin) probing with separate force and sense needles avoids.
• Fiber height and angle. Drift in the fiber's height or angle changes the coupling and its spectral peak. A crash into the wafer breaks the fiber tip and can scratch the device.
• Polarization. With standard single-mode fiber, moving the probe changes the polarization reaching a polarization-selective grating coupler; polarization-maintaining fiber or a polarization adjustment at each die avoids this.
• Light, shielding and condensation. Room light adds photocurrent to dark-current measurements, so low-current work is done in a dark, shielded enclosure, which is also purged with dry gas when the chuck runs below the dew point.

Common questions

What is the difference between a probe station and a prober?

The terms overlap. "Probe station" usually means a manual or semi-automatic bench instrument for engineering measurements; "prober" or "wafer prober" usually means an automatic system that loads wafers and steps through every die for production test.

Can a probe station measure edge-emitting lasers?

Yes, at the die or bar level. Edge-emitting lasers have no surface coupler, so they are tested after cleaving, with a DC needle on the top contact, the chuck as the back contact, and a photodetector or fiber at the facet.

What GSG pitch should pads be designed for?

The pitch of the probes that will be used, commonly 100, 125 or 150 µm. Smaller pitches allow smaller pads with less capacitance; larger pitches are more robust and cheaper.

References: L. A. Coldren, S. W. Corzine and M. L. Mašanović, Diode Lasers and Photonic Integrated Circuits, 2nd ed. (Wiley, 2012); L. Chrostowski and M. Hochberg, Silicon Photonics Design: From Devices to Systems (Cambridge University Press, 2015); D. M. Pozar, Microwave Engineering, 4th ed. (Wiley, 2012).