2.1.1

Fabrication, functional verification and environmental test plan

VCSEL and PD arrays have been fabricated in a circular configuration to match the geometry of the multi-core fiber (MCF) that would be used for the optical transceiver pigtailing. The circular, multi-core solution was chosen in order to enable a 150 Gb/s parallel optical data link through a single optical fiber, thus minimizing the module fiber count and improving the integration density of the interconnections. As an upside benefit, the capability to mate the MCF with multi-core MiniAVIM connectors, makes the solution attractive in terms of reliability compared to COTS ribbon fiber connectors. In addition, the multi-core cable favors the manufacturability of robust hermetic fiber feedthroughs required for the module hermetic packaging/pigtailing. In a previous publication [2] we have presented robust performance of 40 Gb/s VCSEL prototypes at temperatures up to 85°C and at low bias/modulation currents by Chalmers University of Technology. Here we present the fabrication and performance of devices that have been fabricated within PHILIPS qualified manufacturing line. The fabrication transfer to PHILIPS was done in an effort to increase the device TRL level and progress the research prototype to an industrial product. The figure below (left) shows the microscope image of a 3 um (Ø22) and a 5 um (Ø24) aperture size circular VCSEL array. The chip size is 720 x 420 um (LxW). The figure below (mid and right) shows the L-I-V performance recorded at 25°C. At a bias current of 5 mA - which is within the current range provided by the BICMOS VCSEL driver IC - both VCSELs deliver >1.5 mW output power. As expected the larger diameter VCSEL exhibits a better scaling of the output power with no roll-off of the output power at currents up to 10 mA. The output power at 10 mA is 3.2 mW.

Figure 1.

(left) Microscope image of the 6-channel VCSEL arrays and L-I-V curve of 3 um VCSEL array L-I-V curve of 3 um VCSEL array (middle) and 5 um VCSEL (right)

Figure 2.

Small signal modulation response at different currents recorded at 25°C: (left) 3 um aperture VCSEL, (right) 5 um aperture VCSEL

The small signal modulation response (from S21 measurements) at different bias currents and at 25°C is shown in the figure above. The 3 um VCSEL demonstrates a -3dB modulation bandwidth of >17 GHz at 4 mA bias current. For the 5 um aperture VCSEL, at a current of 4.7 mA mA the -3dB modulation bandwidth is >15 GHz. Increasing the current to 11 mA increases the bandwidth closer to 20 GHz. Both performances are sufficient for 25 Gb/s modulation and transmission. To verify this assumption high speed modulation tests at 25 Gb/s and at various currents have been performed. The figures below illustrate the eye diagram analysis for both VCSEL arrays and at driving currents ranging from 4 mA up to 8 mA. In all cases clear eye diagrams have been obtained with extinction ratios (ER) as high as 5 dB. For the 5 um VCSEL the eye diagram improvement is evident as the bias current increases due to the improvement of the device modulation bandwidth.

Figure 3.

Eye diagrams of VCSEL arrays: (top) 3 um aperture VCSEL, (bottom) 5 um aperture VCSEL. Modulation speed is 25 Gb/s.

Multimode 850 nm PDs were also fabricated within PHILIPS manufacturing line. The figure below (inset) shows the microscope image of the six element circular array. Chip size is similar to the VCSEL array. The aperture of each PD is 24 um and was chosen as an optimum trade-off between bandwidth and coupling to the MCF cores.

Figure 4.

Frequency response measurements of the PD wafer at 25°C and 100°C. Dashed lines present fits to the measured data. (inset) microscope image of the PD chip

The figure above shows measured frequency responses of the PD wafer at 25°C and 100°C along with the corresponding fits after subtracting out the VCSEL response from the total PD/VCSEL response. The 3dB cut-off frequency extracted from the fits are 25 GHz and 20 GHz at 25°C and 100°C respectively and at -5V reverse bias voltage. This measurement indicates that the PD is well capable for detection of 25 Gb/s modulated signals. Concerning the device testing both VCSEL and PD chips were subjected to a radiation and lifetest. Here we focus on the tests performed on the VCSEL parts. The figure below shows the test plan that was followed for the VCSEL environmental testing. A total of 32 samples were delivered for testing; 2 samples were used as reference, 4 samples were subjected to radiation tests (2x in Gamma and 2x in proton) and 16 samples were subjected to an accelerated life test. The radiation test conditions followed ESCC 22900 and the radiation test levels are tailored to comply with typical LEO and GEO telecommunication missions. The life test followed PHILIPS internal test protocol for VCSEL wafer qualification.

Figure 5.

(a) VCSEL array test plan, (b) packaging and configuration of the test samples for gamma radiation test, (c) proton radiation test set-up, front view: VCSEL and PD components aligned with proton beam, rear view: samples against the proton generator

2.1.2

Radiation and accelerated life testing

The two VCSEL designs (3 um and 5 um aperture size) have been subjected to gamma radiation testing. The radiation facility was CNA (Centro Nacional de Aceleradores) Gamma Facility in Seville, Spain. The chips were mounted and wire-bonded into DIL package modules to enable in-situ functional testing at the different steps during irradiation as shown in figure 5b). Current was supplied to the lasers through the package pins and the optical power emitted by the VCSELs was collected through free space by integrating sphere detector for L-I-V measurements. Packaging and testing was performed by ALTER. The TID was 100 krad at a dose rate of 210 rad/h. Modules were in OFF state (unbiased) during radiation. The following graphs present the results of the L-I-V characterization of the two VCSEL types. All six VCSELs in each array have been tested - here we present L-I-V characterization for one VCSEL in each array, since all the lasers in the array responded uniformly. The driving current was tuned from 0-10 mA at steps of 0.025 mA (for 0-2.5 mA) and 0.25 mA (for 2.5-10 mA). The results validate the efficiency saturation of the 3 um VCSEL variant. A reference VCSEL array sample for each array under test was used at each measurement step in order to ensure measurement repeatability.

Figure 6.

(left) VCSEL array gamma test results. L-IV curves at 0 and 100 krad for: (left) 3 um aperture VCSEL, (right) 5 um aperture VCSEL

The prototype VCSEL arrays exhibited stable performance under gamma irradiation. The results indicate no drift of the VCSEL array laser threshold current and as such no impact on the laser efficiency at the radiation levels of interest. Measurements are in agreement with literature which suggests that displacement damage effects induced by gamma radiation would appear at much higher TID levels. In addition two similar VCSEL arrays were subjected to proton radiation testing. Proton testing was done by ALTER using UCL CYCLONE110 (CYClotron de LOuvain-la-NEuve) facility. The test conditions are outlined below.

Table 1.

proton radiation test condition

The figure below shows the L-I-V test results recorded at the three steps. Similar to the gamma radiation tests, the L-I-V curves show no drift of the laser threshold current and as such no impact of the laser efficiency at a cumulative fluence up to 10¹² p/cm².

Figure 7.

(left) VCSEL array proton test results. L-IV curves at the three steps for: (left) 3 um aperture VCSEL, (right) 5 um aperture VCSEL

Figure 8.

VCSEL array ALT test result

Finally, 16 VCSEL samples were subjected to an accelerated life test 1500 h biased at 6 mA at a temperature of 170°C. The samples were selected randomly from the produced VCSEL batch and there was no screening applied following the part selection. The pass/fail criterion was a power drop of <2dB which corresponds to a MTTF>1000hrs. The figure above shows the evolution of the output power emitted by the VCSEL samples during the 1500h lifetest. There was an early failure observed on a channel whereas the other 15 channels operated normally and with no failure up to 1500h. The failure within the first hours of the test would have been captured by standard screening which foresees an 168 h burn-in to detect samples with infant mortality.