2.1

Photodiode

The detector used in this work is an InGaAs/InP high speed PIN photodiode chip (PD40X1, see Figure 1). The detection area as well as the GSG connection pads are located on the front side of the chip. On the backside of the InP substrate, the light entry area features a monolithically integrated spherical lens with anti-reflective coating reducing reflection of the optical interface below 1% at the specified wavelength. The position of the lens with respect to the top side detection area is aligned with photolithographic precision, better than ± 2 μm. The lens enables high optical coupling efficiency, increased alignment tolerances and can be optimized for specific optical illumination system requirements. The photodiode offers a responsivity of R_PD = 0.82 A/W at wavelength λ = 1550 nm and T = 25 °C without polarization dependent losses while the small front side active area provides a low capacitance of C_PD = 40 fF and high 3 dB-bandwidth of f_3db,PD = 38 GHz measured on 50 Ω load on a probe station¹.

Figure 1.

56 Gbd InGaAs/InP photodiode chip PD40X1, Left: front side with GSG pad layout. The vertical illuminated active element is in the centre of the chip, Right: backside with monolithically integrated, spherical lens with anti-reflective coating.

2.2

Fabrication

The fully tested photodiode chip PD40X1 is flip-chip soldered on an Al₂O₃ ceramic carrier using an eutectic solder process. The ceramic includes a resistive broadband 50 Ω matching circuit as well as a bias-T circuitry. This sub-assembly is soldered into the package body. Then, the optical fiber is actively aligned to the photodiode to optimize responsivity and linearity. The optical feedthrough is hermetically sealed. After functional testing, the lid of the package is hermetically sealed, and its hermeticity is tested in a gross and fine leak test. The resulting packaged photodiode PQS40A-L is shown in Figure 2.

Figure 2.

Packaged RF photodiode PQS40A-L.

2.3

Packaged Photodiode Description

PQS40A-L features an optical single mode fiber pigtail with an FC/APC connector. The DC coupled RF output is provided through a K-connector. The photodiode reverse bias voltage is applied through the internal bias-T (see Figure 3 for the schematic pin configuration). The package is specifically optimized for small size and low weight (< 8 g without fiber pigtail). The dimensions are shown in Figure 3.

Figure 3.

Left: Schematic pin configuration of PQS40A-L. Right: Mechanical dimensions (in μm) of PQS40A-L. The package height is 9.1 μm.

3.1

Responsivity

Figure 4.

Responsivity measurement result of PQS40A-L as a function of temperature.

Responsivity of the packaged photodiode is measured using a continuous wave (CW) laser with wavelength of λ = 1550 nm. The optical power is controlled and calibrated in front of the optical connector of PQS40A-L using an Agilent N7752A optical attenuator. In order to measure the temperature dependence, the device under test (DUT) is placed in a Vötsch VT4002 climate chamber. We obtain a responsivity of R = 0.75 A/W at wavelength λ = 1550 nm, T = 25 °C and reverse bias voltage V_B = 5 V. Compared to the responsivity of PD40X1, we estimate the coupling efficiency between the FC/APC connector and photodiode chip to be around 91%. The responsivity decreases at lower temperature. The optical return loss (ORL) is better than 40 dB over the entire temperature range, indicating stable optical coupling.

3.2

S-Parameter

The S-parameters of PQS40A-L are measured by the setup shown in Figure 5. The electrical RF signal from the vector network analyser (VNA) is used to generate an optically modulated signal using a LiNbO₃ modulator. The modulated optical signal at wavelength λ = 1550 nm has an extinction ratio (ER) around 14 dB. The optical signal is transmitted through an optical attenuator to the optical input of PQS40A-L (port #2). The average optical input power is approximately P_in ~ -3 dBm. The electrical RF output signal (port #1) is directly connected to the VNA, while the bias voltage is applied through an external power supply to the DC bias pin (port #3). In order to measure the temperature dependence, the DUT is placed in a climate chamber, which allows to control the temperature between -40 °C and 90 °C in ambient atmosphere.

Figure 5.

O/E S-parameter measurement setup.

The frequency response (S₁₂) at wavelength λ =1550 nm and electrical return loss (S₁₁) are shown in Figure 6. The response is flat over the entire frequency up to 38 GHz with a 3 dB-bandwidth above 40 GHz. The broadband response of this component in combination with its high responsivity enables operation in C-, X-, Ku- and the entire Ka-band (26.5 GHz to 40 GHz) as well as in the lower portion of the Q-band (37.5 GHz to 42.5 GHz). The electrical return loss Mag(S₁₁) remains below -5 dB up to 34 GHz.

Figure 6.

Frequency response (S₁₂) and electrical return loss (S₁₁) measurement of PQS40A-L. Left: As a function of applied DC bias voltage V_B. Right: As function of temperature.

The frequency response of PQS40A-L shows only a small bias voltage dependence since the photodiode chip is fully depleted above V_B = 1.0 V, meaning the device capacitance and transit time does not change significantly under higher reverse bias voltage. The temperature dependence of the frequency response is small.

3.3

Linearity

We define the linear regime of the photodiode by measuring the 1 dB compression point at a fixed RF frequency. The linear region is determined as a function of the RF output vs. the RF input power where the RF output power at the fundamental frequency decreases by 1 dB or 10 % from its linear relationship with respect to the RF input power. Above this point the photodiode goes into optical saturation and generates higher order harmonics and non-linearities. The optical RF input power is a function of its ER and is proportional to the average optical input power which is itself proportional to the average photocurrent generated by the photodiode. Therefore, for a fixed ER, it is common to describe the linearity of the photodiode by the average photocurrent at the 1 dB compression point. We call this photocurrent the optical saturation current I_s¹⁶.

The linearity of PQS40A-L is measured using the setup shown in Figure 7. The VNA generates an electrical RF signal at a fixed frequency which drives a LiNbO₃ modulator to generate an optical signal at the wavelength λ = 1550 nm. In order to obtain the necessary optical RF power, the signal is first amplified using an erbium doped fiber amplifier (EDFA) and then the average optical power is controlled using an optical attenuator. The electrical RF output of the DUT (port #1) is measured using the VNA at the operating frequency. The bias voltage is applied using a power supply while the photocurrent is measured by the bias voltage drop across a 100 Ω resistor in the biasing circuit. In order to guarantee a fixed, nominal bias voltage at the DUT (port #3), the voltage drop across the 100 Ω resistor is compensated by increasing the voltage of the power supply as a function of the average photocurrent at every optical input power.

Figure 7.

1 dB compression point measurement setup.

The linearity measurement at a fixed frequency of f = 40 GHz is shown in Figure 8. Increasing the average optical input power results in a linear increase of the average photocurrent flowing through the photodiode. Above the optical saturation current I_s, the optical power density on the photodiode active diameter is high enough to create very high density of electron-hole pairs which screen the applied electric field, known as space-charge effect¹⁷. As a result, carrier transport is slowed down, the photodiode bandwidth decreases, and non-linearities in the response of the detector are introduced. This detrimental effect depends first and foremost on the optical power density on the detector¹⁸. In order to increase the linear range, the optical power density should be as uniform as possible over the detecting area. Considering that the main design target of this photodiode is to optimize optical coupling and responsivity, overfilling of the active element with the optical spot size to create a more uniform optical power density profile is not an option¹⁹ and optical beam shaping20 is difficult to implement for small area detectors in such small package. Therefore, we have focussed our effort on increasing the spot size by tightly controlling the fiber position and the lens shape of the monolithically integrated backside lens, reducing the non-spherical aberrations. This avoids distortion to the Gaussian optical power density profile created by the single mode fiber illumination and hence makes the optical power distribution reproduceable.

Figure 8.

Left axis Change of the RF output power of PQS40A-L at f = 40 GHz as a function of measured average photocurrent and applied DC bias voltage V_B. Right axis shows the compression of the relative RF output power with respect to a linear relationship.

A further effect which reduces the optical saturation current is a photocurrent dependent voltage drop caused by the 80 Ω resistor as part of the bias-T (see Figure 3) and the 50 Ω load resistance in parallel with the 50 Ω matching termination. The series resistance of the photodiode itself also causes an electric field drop in the junction²¹. However, this effect is minor considering the photocurrents is in orders of tens of mA and the series resistance of the photodiode is only around 7 Ω, extracted from S₁₁ measurement of the photodiode chip. The bias voltage drop lowers the available bias across the photodiode and hence the field inside the junction²². Therefore, less charge or optical input power is required to create the space-charge effect. Furthermore, the on-set of optical saturation effect is temperature dependent. At higher temperature, the mobility of the charge carriers, in particular for holes, decreases. It inhibits transport of the carrier from the junction which feeds back to enhanced charge accumulation which typically starts on the n-side of the depletion region¹⁷. These dependencies can be observed in the measured temperature and bias voltage dependence of the optical saturation current shown in Figure 9. At V_B = 5 V and T = 90 °C, linear operation at f = 40 GHz is maintained up to I_s = 8 mA corresponding to an average optical input power of approximately P_in ~ +10.2 dBm considering the measured responsivity shown in Figure 4. The linear operation regime increases for decreasing temperature to I_s = 10 mA (Pin ~ +11.2 dBm) and Is = 12 mA (Pin = +12.5 dBm), at T = 25 °C and T = -40 °C, respectively. These photocurrents result only in moderate dissipated power P = I_sV_B ~ 50 mW in the junction of the photodiode, and therefore thermal runaway¹⁸ and catastrophic failure are not a concern. The optical saturation current is only slightly frequency-dependent as seen in Figure 9, which indicates that the optical saturation mechanism mostly depends on the average photocurrent.

Figure 9.

Left: Bias voltage and temperature dependence of optical saturation current at f = 40 GHz. Right: Temperature dependence of optical saturation current at three different frequencies.