3.1

Evaluation plan methodology

With the lot of 57 laser modules considered as a representative manufacturing population, the evaluation started with an initial characterisation encompassing L(I), V(I), spectrum, low level I(V), relative intensity noise (RIN) and linewidth. The laser diodes were accordingly found to have nominal ex-fibre powers between 70 and 100mW, and threshold currents between 35mA and 65mA. Then, under evaluation testing, measurements were repeated at each test step, allowing any slight variation in the laser parameters to be determined. To allow assessment of the reliability of the technology, the Figure 3 test groups encompass step-stressing, environmental testing, irradiation, electro-optic capability and endurance.

3.2

Step-stress sub-group

This concerned an evaluation of the effects of step-stressing of laser temperature between -25°C and 50°C, and output power above 100mW (with the laser current ranging from 450mA to 525mA). No major defect was noted and no significant drift was observed on the laser performance. Figure 4 illustrates the monitoring of output power drift with time.

Figure 4.

Power monitoring during the test at 100mW.

3.3

Environmental sub-group

Mechanical testing concerned 30g rms random vibration followed by half-sine shocks from 500g/1ms up to 1500g/0.5ms; Figure 5a shows samples subjected to random vibration. Temperature cycling concerned 500 rapid thermal cycles between -55°C and 70°C with an average ramp of 15 deg/min and dwell time of 15 minutes. Depressurisation involved rapid pressure cycling down to 10mbar over 5 seconds at ambient temperature, and with the laser diodes driven at 375mA and 30°C. Low temperature testing concerned 2000h of non-biased storage at -55°C. Thermal vacuum testing comprised 1000h of secondary vacuum at <10^-5 mbar with the laser diodes driven at 375mA and 30°C, and the module case temperature held at 65°C. Humidity testing concerned 240h of 85°C/85% damp heat, followed by -10°C to +50°C temperature cycling around the dew point.

Figure 5.

(a) Laser modules on mechanical shaker, and (b) devices inside the gamma radiation chamber.

For all the environmental tests, the modules showed no significant drifts except for the shock test where the limit is 1000g/0.5ms, and rapid thermal cycling wherein one of three laser diodes failed, indicating sensitivity of the module to fast temperature variation. The latter sensitivity is to be screened as discussed in Section 4.

3.4

Irradiation sub-group

Under radiation testing, both ionising dose and non-ionising dose were considered. The ionising dose was provided by gamma irradiation (total dose of 100 krads), whereas the non-ionising dose was provided by proton irradiation (total fluence of 10¹² p+/cm² for an energy of 30 MeV). After both gamma and proton irradiation, no significant drift was observed on the laser diode. As expected, only the monitoring photodiode drifted after proton exposure. Figure 5b shows the samples inside the gamma radiation chamber.

3.5

Electro-optic sub-group

In this group, 4 types of tests were concerned. First, characterisation was conducted at extreme temperatures of -20°C and +65°C. No major issues were encountered. Second, characterisation was conducted under secondary vacuum (around 10^-5mbar) under various temperatures of laser-diode and module-case. Relative to similar measurements in air, no differences were noted. The spectra given in Figure 6 illustrate the effect of different temperatures of laser-diode and module-case.

Figure 6.

Optical spectrum at various module-case and laser-diode temperatures, Tc and Tw.

Third, electro-static discharge sensitivity was evaluated following the Human Body Model (HBM) of MIL-STD-883 standard, method 3015 [3]. As per Figure 7, the tested laser modules showed no degradation up to the 8kV step.

Figure 7.

Pulse monitoring during the test at+/-6kV and +/-8kV on laser modules.

Fourth, catastrophic optical damage (COD) was investigated wherein the laser current is increased until failure and the forward voltage, optical power and laser temperature were continuously monitored. The resultant L(I) curves are shown in Figure 8. We note the typical L(I) bell shape in which the optical signal extinguishes after a laser current of 1A and is reversible thereafter; this shows that the laser diode can be driven up to 1.8A without degradation and verifies its robustness to high electrical current.

Figure 8.

L(I) curves obtained over 3 COD test runs.

3.6

Endurance sub-group

Life-testing was performed on 15 laser modules at a case temperature of 65°C for 5000h. The laser diode was driven at a current of 375mA and a temperature of 30°C. The in-situ monitoring of the optical power is represented by Figure 9.

Figure 9.

L(I) Optical power drift ΔP_out during life-testing.

The stability of the laser diodes illustrated by Figure 9 indicates minimal drift over the 5000h life-test.

3.7

Evaluation test conclusion

With optical output power exceeding 70mW, no catastrophic failures were encountered, and the samples were shown to be robust to thermo-mechanical as well as to radiation tests. With potential space compliance summarised as in Table 2, two significant points concern the mechanical shock level which is limited to 1000g/0.5ms, and the coupling sensitivity to rapid temperature change. Near-term addressing of the latter point is by screening, as discussed in Section 4.

Table 2.

A summary of ESCC-23201 [2] based evaluation.

4.1

Introduction

For validation-tested devices to be assured of fulfilling mission profiles, their random failure rate must be below the maximum accepted rate, and their Mean-Time-To-Failure (MTTF) for parametric wear-out must be greater than the mission lifetime. Figure 6 is a schematic of the bathtub curve and the conditions to be fulfilled.

Figure 6.

Bathtub curve and Satellite mission profile limits.

According to the Telcordia quality standard [4], the time to failure τ for laser diodes follows two types of models:

where τ_o is a constant, E_{a_random} (the activation energy for random failure) =0.35eV, k (Boltzmann’s constant) =8.617×10^-5eV/K, T_wave is the temperature of the laser diode chip (as set by the TEC), R_th (the thermal resistance of the laser diode) is evaluated by a non-destructive method to be in the range [8.8 K/W, 10.4 K/W], I_nom and V_nom are the nominal laser current and voltage biasing conditions, P_out is the nominal optical output power, I_burnout is the laser current that leads to instantaneous catastrophic failure, and E_{a_wearout} (the activation energy for wear-out failure) =0.4eV. In particular, we can note that by removing the operational bias factor (I_nom/I_burnout)^-2 from Equation 1b, τ_wearout becomes consistent with an Arrhenius (pure thermal) law.

However, our module’s components are carefully selected, characterised, and assembled according to space-quality standards (in particular ECSS-Q-ST-60-05 [5]) and do not exhibit random failure. Hence only Equation-1b wear-out applies. Also, with spectral stability ensured through operating the laser diode at constant current and temperature, and with change in dynamic line-width (or frequency chirp) having already demonstrated strong technological stability, our work highlights ex-fibre power Pout is the only affected parameter. This is seen to seen, to a very limited extent, during life-testing of the laser diode. Such degradation differs from that in other electronic devices due to the recombination of electron-hole pairs and the presence of high optical densities within the active region and at the output facet [6]. The applicable degradation mechanisms are (i) defect formation in the active region, and (ii) degradation of current-confining junctions [7]. As first observed in AIGaAs lasers and LEDs, the defect structures generally comprise dark spot defects (DSDs) and the dark line defects (DLDs) of the (100) crystal direction. Operation at high injected current densities creates high-energy electrons and holes, thermal gradients, potential for strain fields, and a high non-radiative recombination rate inside the active region. This, in turn, can promote the motion, multiplication, and growth of isolated defects into clusters. Although InGaAsP/InP buried double heterostructure laser diodes are less vulnerable to some of the dominant degradations mechanisms of other technologies (e.g. GaAs, InGaAs, InP, etc) [8], [9], the fact remains that reliability can still be critical according to the specific epitaxial growth, blocking layers, and (in DFB lasers) the selectivity of the engraving of the Bragg grating. In particular, in empiric studies on the technology by Fukuda et al. [10], it was found that activation energies for generating DSDs and DLDs are respectively 0.16eV and 0.2eV, that the generation time of the first DSD depends strongly on the operating current density, and that the time to failure follows the factor exp(A · J²), where A is a constant and J is the current density.

Since, in addition to temperature, degradation is enhanced by increasing the laser current and output power, aging studies are conducted in one of the following two modes of operation. The first is constant current aging – often referred to as ACC mode (automatic current control) – where the laser current is held constant for the duration of the test. The second is constant power aging, which is often referred to as APC mode (automatic power control). Under APC, the optical output power is measured either with an external photodetector or by using an internal monitor photodiode if one is available within the laser package, and this power is then held constant by continuously adjusting the laser current as required to maintain constant output power. While constant power aging is frequently used as a mimic of typical operation, our work uses constant current because this mode is essential to achieving the specified operating-wavelength of within ±0.1nm.

In practice, difficulties in laser diode life testing arise from temperature instability, equipment measurement and control instability, equipment reliability, and power failures. A challenge associated with temperature control is the self-heating of the laser during operation. The temperature elevation at junction level is in the range [8K, 10K] at the maximum rated bias current of 375mA (and bias voltage of 2.4V). Additionally, heat-sink temperature fluctuations as small as 0.1°C manifest as noise in the measurement of optical power due to the temperature sensitivity of laser output power at a given current. Finally, if an external photodetector is used to measure optical output power, its temperature must also be controlled to ensure stable measurements. Laser diode life test studies require an accurate measurement of changes in laser operating parameters as small as a few percent over thousands of hours. Consequently, the stability of the measurement equipment must be on the order of 0.1% per 1000 hours. In most locations, occasional electrical power failures are inevitable during the course of a multi-thousand-hour life-test study. In many cases, it is impractical to provide battery backup systems due to the high power required for heating in life test systems. As a result, the life test system must handle power failures without damage to the lasers, and must be able to resume a test precisely after power is restored.

To model the resultant changes in ex-fibre power P_out, we follow [11] wherein the Eyring law of Equation 1b is extended via the so-called Boltzmann-Arrhenius-Zhurkov (BAZ) model. Hence, the time to failure τ for the combined action of an elevated temperature T and external stress represents S is written as in Equation 1c.

where S is can be any stimulus or group of stimuli (such a voltage, current, signal input, irradiation, etc), τ _A is a pre-factor, γ is a factor of loading characterising the role of the level of stress, and the product γ · S is the stress per unit volume and is measured in the same units as the activation energy E_a. By considering multiple stresses including temperature, the approach offers the establishment of a unified reliability model for electronic devices operating under multiple-stress harsh environments; whether submarine, nuclear, on the ground, in transport, or in aerospace.

The changes in ex-fibre power P_out, which are also seen under temperature cycling, can be mitigated by screening. According to the Telcordia standard [12], laser modules should normally be 100% screened according to:

In terms of production burn-in screening, the challenge is to achieve high manufacturing throughput and accurate measurements at very low cost. The two most common strategies are the following:

4.2

Application and results

With regard to Equation 1b, we consider that the expression suffers a lack of precision at the limit when the current flow is zero since the implied infinite time to failure at I_nom=0 is inconsistent with the Arrhenius law limit. To overcome this contradiction, we propose to update the Eyring model as in Equation 1d.

where, by using the ratio (I_burnout-I_nom)/I_burnout instead of I_nom/I_burnout, a value of 1 is obtained at I_nom=0 which makes τ_wearout consistent with the Arrhenius (pure thermal) law of Equation 1a. The power factor n is then calculated by equivalence to Equation 2b, with the result obtained given in Equation 1e.

A life-test sequence was performed on 15 DFB laser modules at case temperature, T_case = 65°C, for 5000 hours. Also, the laser current I_nom and laser temperature T_wave were set at constant respective values of 375mA and 30°C. We measured the devices at interim times during this ACC stress. A linear drift of output power from 1000 to 5000h was observed with a maximum value close to 4%. Based on a median ranking parameter and the time to failure t_i for a part i, linearly extrapolated to reach the 30% drift, a Weibull plot was obtained as shown in Figure 7.

Figure 7.

Weibull plot considering ΔP_out=-30%, I_nom=375mA, T_wave=30°C and T_case=65°C; the dashed line shows 95% CL.

The analysis shows the lot of devices to be homogeneous, and extracts a Weibull β parameter of β=2.61. Extracted data for a range of ΔP_out values are in Table 3.

Table 3.

Extracted reliability data using Inom=375mA, Twave=30°C and Tcase=65°C for various values of ΔPout.

Other values of predicted MTTF and τ_wearout may be obtained through changed values of current (I_nom) and temperature (T_wave) applied to the laser diode. First, Equation 1c may be simplified by noting that the temperature of the junction T_j of the laser diode is given by Equation 2.

Then, considering user and test-stress levels of laser current and junction temperature (respectively I_user, I_stress, T_{j_user} and T_{j_stress}), we may deduce an accelerating factor AF from Equations 1c, 1d and 2 as given in Equation 3:

With E_a=0.4eV as manufacturer data, and user values reduced to I_user=275mA and T_{wave_user}=20°C, we may then deduce AF as 3.27 and update the Table-3 data as in Table 4.

Table 4.

Updated MTTF and failure rate data considering Inom=275mA and Twave=20°C.

Accordingly, we demonstrate an MTTF of >15 years under I_nom=275mA and T_wave=20°C operating conditions, taking ΔP_out=-30% as an acceptable failure criterion.

Nevertheless, these preliminary data are a first assessment of MTTF and FIT values, and need to be consolidated with additional steady-state life-testing as follows:

To quantify the degradation of the laser diode with junction temperature T_j, we empirically define the laser threshold current Ith and the external differential quantum efficiency η_D as in Equation 4.

where T₀=198K, T₁=178K, I₀=10.2mA and η₀=1.32W/A.

To permit quantification of all potential manufacturing, testing and operational failures of the module’s laser diode (whether arising from electrical biasing, atmospheric, radiation-based, thermo-mechanical, or other applied environmental stresses), we implement the BAZ-modelling approach by setting x%=I_nom/I_burnout within Equation 1d, and then using Equation 2 to give:

We now set

to obtain

It is thus seen that the effective activation energy depends on stress condition applied, and varies linearly with the current bias applied as a first approximation. As a numerical example, we can assess γ_I = 0.246 and E_{a_eff} = 0.347 eV based on the following typical data: T_wave=20°C, I_nom=375mA (=75%of the maximum 500mA rating), I_th=50mA, V_nom=2.4V, n=3.8, E_a=0.4eV, R_th=9.6K/W, P_out=80mW, and I_burnout=900mA. Figure 8 represents the effective activation energy E_{a_eff} as function of the stress condition (x%=I_nom/I_burnout) applied. Accordingly, E_{a_eff} decreases with x%=I_nom/I_burnout.

Figure 8.

Effective activation energy E_{a_eff} versus applied current stress condition x(%) calculated for T_wave=20°C.

4.3

Optimising the testing methodology for new laser-diode technology

Following Section 4.2, we recommend optimising the BAZ-model as per the following steps:

Step 1: Characterise the laser-diode lot in chip form, and at module level after screening, comprising:

Step 2: Perform step-stressing according to Table 5, where 15 test devices and 2 extra control devices (CD) are suggested.

Table 5.

Proposed modification to the step-stress testing of reference [2], including an update to the number of devices.

Step 3: As consolidated from step-stress data to achieve more than 50% failure in a short period of time (typically close to 1000h), define an accelerated endurance program. Each test sequence is to include at least 6 devices, and involve 3 T_wave, temperatures and 3 I_nom biasing conditions, that are proposed to be 50°C, 80°C, 100°C, 350mA, 375mA and 400mA. Perform testing and use control devices for every interim measurement. Use high temperature data to calculate E_a: this is the baseline of the BAZ model (use chip on carrier for simplification of measurement). Calculate values of n and γ.

Step 4: Repeat above methodology, where appropriate, for other stressing parameters – e.g. output power (may be done simultaneously as for current), and irradiation.

4.4

Key outcomes and proposals

Guaranteed reliability: Under, for example, at I_nom=275mA and T_wave=20°C, the analysis of Section 4.2 predicts <10% drift of lasing current threshold, <30% drift of output power, within ±0.1nm drift of wavelength, and a 15yr MTTF.

Space-quality lot validation: For a 15-year operating life and an acceptable quality limit of 0.65%, we propose steady-state life-testing as in Table 6, and acceptable parametric drifts as in Table 7.

Table 6.

Steady-state life-testing for lot validation.

Table 7.

Acceptable parametric drifts for lot validation.

Screening and burn-in: Based on the obtained evaluation test data (Section 3), we recommend the following order of conditioning to be completed on a 100% basis for space lot delivery:

Table 8.

Burn-in conditions.

Maximum ratings: We propose these, from known device heritage and Section-3 data, as in Table 9.

Table 9.

Absolute maximum ratings for some important parameters.

Operational ratings: Table 10 shows the main operational ratings for space applications, listed by key component.

Table 10.

Some key operational ratings, where start-of-life and Tcase ∈[0°C, 50°C] apply unless otherwise specified.

Reliability testing methodology for new laser-diode technology: For this, we make the recommended optimisations as per Section 4.3.