A.

Development of large area single element HgCdTe APDs for LIDAR and FSO

The required temporal resolution in most LIDAR and FSO applications ranges between 50 ns to 250 ps, corresponding to bandwidths between 100 MHz and 2 GHz. The active optical diameter dependence on the receiver optics and is typically in range of 80 to 200 μm. These specifications define the parameter space in terms of Cd composition as a function of system requirements such as operating temperature, sensitivity and active area. The sensitivity is ultimately limited by the dark noise of the APD, induced by the dark current and residual thermal flux, which should be lower than the input referred noise of the pre-amplifier. In the present development, we have chosen to work with detectors cooled down to temperatures around 200 K. This allows using thermo-electric cooling (TEC), in ground based applications, or passive cooling, in space applications. In order to obtain the lowest possible temperature and largest optical active area, we have chosen to deport the amplifier off the cold-down surface rather than to hybridize the APD on the amplifier. This choice implies a higher amplifier noise which needs to be overcome by a higher APD gain. In addition to guarantee a lower detector noise, due to lower APD dark noise, this choice is also more versatile as it allows changing the APD λ_c and surface with a given amplifier and/or the amplifier with a given APD as a function of application requirements. Higher sensitivities, down to single photon sensitivity, can be obtained at lower temperatures using application hybridized read-out circuits. The development and perspectives for such detectors are described in sub-section C.

A model have been developed to predict the performance of such HgCdTe APDs coupled with a pre-amplifier with a given noise and bandwidth. The gain of the APD as a function of x_Cd and operating temperature based on an analytical gain model we presented in reference [22]. The dark current noise and noise is calculated supposing a diffusion current generation at all temperatures, which is multiplied with the total gain of the APD. The sensitivity of the detector is measured as the equivalent photon number noise, NEPh, integrated during the characteristic time of the amplifier TC = 1/2 BW.

where I_n is the input referred rms noise current:

The Fig. 2 shows the calculated gain and NEPh of a single element detector as a function of the reverse bias for a 100 μm diameter APD with x_Cd = 0.37 operated at T = 190 K (λ_c = 3.1 μm), coupled with a preamplifier with a bandwidth of 300 MHz and an input referred noise i_TIA = 2 pA/√Hz. The sensitivity approaches single photon resolution with NEPh = 2.87 photons per TC, corresponding to a noise equivalent power (NEP) of less than 300 pW. This is 10 better than the highest performance reported for InGaAs APDs receiver with equivalent size and lower bandwidth, NEP = 3 nW at BW = 100 MHz [23]. Photon resolution can however not be achieved at this operating temperature due to the dark current noise of the APD which is the origin of the saturation of the NEPh reduction observed at high reverse bias.

Fig. 2.

Estimation of the gain and the sensitivity (NEPh per TC) of a 100 μm diameter HgCdTe APD with λ_c=3.1 μm operated at T=190 K in coupling with a TIA with an input noise of i_TIA =2 pA/Hz and a BW of 300 MHz.

Two large area TEC cooled detector modules with deported TIA have been developed using TIAs with bandwidth and noise adapted for FSO (BW=400 MHz, i_TIA=2 pA/) and LIDAR (BW=20 MHz, i_TIA=1 pA/Hz). The detector assembly with the high BW TIA is illustrated in Fig. 3. The performance of the two detector modules are compared in Tab. 3. The measurements on the detector APD3-TIA-1 were made by imaging the output of a 200 μm diameter multi-mode fibre on to the APD. Hence, the characterisation of the total detector gain (3 MV/W) and estimated quantum efficiency (58 %) include losses due to the optical coupling from the fibre to the detector. The measured variation of the signal gain and NEPh is reported in Fig. 4. The output noise at the nominal gain was slightly higher than the amplifier noise, mainly due to the detection of residual thermal photon flux which needs to be limited by, for example, adding cold optical filters. A record low NEP=0.73 nW for this level of bandwidth was still observed, which is a factor 4 lower than the values observed in InGaAs APDs for detectors with a factor 4 lower TIA bandwidth. The corresponding rms input noise is about 9 photons per TC, which is comparable to the predictions in Fig 2. The bandwidth of the detector module were limited the diffusion collection of photo-generated carriers before avalanche multiplication, BW=80 MHz. The impulse response time of detector is illustrated in Fig. 5. This detector module has been used to demonstrate error free data transmission from the moon to ESAs OGS telescope at Tenerife at rate of 80 Mbit/s. The results of these measurements are presented in the next section.

Fig. 3.

Photography of a) the inside of a HgCdTe APD detector module and b) of a closed detector assembly.

Fig. 4.

Variation of the detector gain and input equivalent photon noise of the detector module used during the lunar laser communication demonstration at ESAs OGS at Tenerife.

Fig 5.

Impulse response of a 160 μm diameter APD coupled with TIA-1, operated at T=190 K with a gain of M=95 to an impulse of 100 photons. The vertical and horizontal scales are 5 ns and 10 mV per division, respectively.

Tab 3.

Performances of the high and low bandwidth detectors.

The slower detector module was build using TIA-2. This TIA is DC coupled and the total gain can be measured using a CW laser light (λ = 1.55 μm). The total detector gain of S = 294 kV/W was measured at unity APD gain, which corresponds to an external quantum efficiency of 67 %. At 16 V bias, the apparent APD gain is 80 and the total detector gain reaches a value of 23 MV/W. The linearity of the detector was tested at an APD gain of 23. The detector was found to have a good linearity of more than 3 orders of magnitudes, as illustrated in Fig. 6. The output noise of the TIA was higher than expected even at low reverse bias. The increase in noise is probably induced by parasitic impedances at the input of the amplifier. At higher bias we observed a slight increase in noise, again due to residual thermal photons. The corresponding NEP at a gain of 80 of this slower detector module is 0.13 nW (30 fW/√Hz). High detector gain and lower noise can be expected in detectors with lower TIA noise and higher APD gain with NEP values lower than 20 fW/√Hz. An optimized version of this detector will be transferred to Laboratoire de Metrologie Dynamique (LMD) to be used for Differential absorption LIDAR (DIAL) detection of atmospheric CO₂, in the scope of a CNES R& T project. Similar detector elements will also be subject to first space reliability tests on APDs made at CEA/Leti.

Fig 6.

Pulse amplitude linearity measured on a 200 μm diameter HgCdTe APD coupled with TIA-2 at T=180 K with and APD gain M=23.

B.

Demonstration of lunar optical communication link (LOCL) with HgCdTe APD during the LLCD-LADEE mission

The APD3-TIA-1 detector module have been used during the demonstration of a laser optical communication down link from the moon, as a part of ESAs contribution to the Lunar Laser Communication Demonstration (LLCD) with the LADEE satellite orbiting around the moon. In this demonstration, the data was encoded using 16ary pulse position modulation (PPM) with pulse durations from 3.2 ns (Mode 5, decoded transmission rate of 39 Mbits/s) down to 0.2 ns (Mode 1, decoded transmission rate of 625 Mbits/s). The expected average power on the detector has been predicted to be around 300 pW, corresponding to about 120 photons per pulse during Mode 5 modulation, i.e. close to the average photon number of the impulses in Fig. 5.

The performance of the APD detector module was first tested as a function of input average power using laboratory generated Mode 5 and Mode 4 modulated light. The detector output signal was decoded using an electron unit developed at RUAG as described in [24]. Fig. 7 a) and b) report the variation of the symbol and decoded frame error rate as a function of the average power at the output of the optical power. The error rate was estimated over a test pattern with 10⁵ bits, which enable the estimation of frame error rates down to 10^-5. To this limit error free detection was achieved for average optical power down to 20 photons per decoded bit, corresponding to 40 photons per impulse and 100 pW average power, in mode 5 (Fig. 7 a)). Due to the bandwidth limitations of the APD, the mode 4 required higher optical power, at least 200 pW to achieve error rates below 10^-5.

Fig. 7.

Symbole error rate (SER) and fram error rate (FER) after error correction measured in) mode 5 and b) mode 4.

This level of performance yielded a margin in the power compared to the expected power levels and both mode 4 and mode 5 were tested to detect and decode the light from the moon. A stable error free Mode 5 (40 Mbits/s) down link was first demonstrated on 3/4/2014 for an operation lasting over 9 minutes. A Mode 4 down link (80 Mbits/s) was demonstrated later the same day. The Fig. 8 shows the average power level at the detector and the decoded frame error rate as a function of time. As can be seen, a stable error free operation is obtained for average power levels above 200 pW, in agreement with the characterisation using the stable laboratory source. The loss of connection in this mode for low power levels indicates that the current detector could not be used in a faster mode, i.e. Mode 3 and higher.

Fig. 8.

Average optical power and FER detected measured with mode 4 signal from the moon on 4/4/2014.

The present data should be compared with the results obtained by NASA using super-conducting nanowire single photon detectors (SNSPD), developed by the Lincoln lab at MIT. The SNSPD detectors are cooled down to 2 K and had been specifically optimised for the LLCD experiment. Error free down-link connexion was demonstrated even in mode 1 (625 Mbits/s) for average signal levels down to 2 photons per pulse [25]. To concurrence this performance one need to improve the sensitive and the bandwidth of the HgCdTe APD detectors. Our measurements [12] have shown that 10 GHz BW should be achieved with HgCdTe APDs detectors and the sensitivity can be increased by the hybridization of the APD on a CMOS circuit. These perspectives show that HgCdTE APDs detectors are strong candidates for space and ground based Gbit/s FSO. One possible approach for this optimization is described in the next sub-section.

C.

Perspectives for time resolved high sensitivity detection using HgCdTe APDs for space application

The use of a deported preamplifier used for the large area single element detectors described above is a versatile solution that can be adapted to various applications requirements in terms of detector area, bandwidth and operating temperature. However, to achieve the highest sensitivity, it is mandatory to hybridize the detector onto a read-out circuit. In this case, the read-out noise can be limited to less than 100 electrons rms, without APD gain, and to values below 1 electron for high APD gains. Hence, single photon resolution can be achieved with HgCdTe APDs in this configuration, presenting a number of interesting properties for photon-counting applications such as high photon detection efficiency (PDE), absence of after-pulsing (yielding high count rates), photon number resolution (PNR) and single photon detection capability into the MWIR range.

These properties have to a large extension been confirmed by the first demonstrations of single photon detection with HgCdTe APD reported by DRS [26] and CEA/Leti [27], in particular concerning the high PDE and PNR capability. The Fig. 9 illustrates derived signals captured with the CEA/LETI HgCdTe single photon detector with a CMOS circuit limited bandwidth of 7 MHz associated with a low read-out noise of 20 electrons rms [27]. The latter allows counting photons already at low values of APD gain, which is in favor of low dark count rates. The detector used to capture the data in fig 9 was made with a MWIR APD and was consequently sensitive to residual thermal flux. The signal in Fig. 9 a) is due to the detection of such residual thermal photons which were detected even though a cold filter was used to filter photons with λ > 2.5 μm. The Fig. 9 b) reports the signal from the same detector at zero flux. The residual events are mainly due to dark count events generated vie thermal generation in the APD. The apparent difference in amplitude between the impulses in Fig. 9 a) and b) is due to a difference in gain between photon initiated events, which are completely amplified in the APD, and thermal events in generated in the multiplication layer, for which the gain depends on the position of generation. The difference in the amplitude distribution of photon event and dark event allows to some extent filter the dark counts while obtaining a good PDE. We found that an internal PDE of 90 % could be obtained with a good filtering of the dark counts. The dark count rate (DCR) in MWIR APD has been measured to be around 200 kHz. The DCR is expected to decrease exponentially at lower λ_c, but this measurement has not yet been performed. DCR rates in order of 1 kHz are expected at an operating temperature of 80 K in such devices.

Fig 9.

Output signal as a function of time with a MWIR HgCdTe APD h measured with a) a low thermal flux of photons and b) at zero flux.

The most recent development of the next generation of single photon MCT detectors has been dedicated to direct LIDAR detection of gazes such as CO2, for which single photon detection is required when the gaz concentration is measured over a large distance[28, 29]. The development of an IR photon counting LIDAR detector at LETI have been dedicated to the development of a read-out circuit which have not yet been hybridized with an APD array[28]. The detector is based on a small array of 5 × 5 pixels with a 30 μm pitch, in which each pixel is hybridized with an individual APD. The detector can be switched between an asynchronous photon counting mode, in which each pixel detect single photon events, and a linear mode, in which the output current of all APDs are amplified with a transimpedance amplifier. The photon counting to linear mode operation enables photon noise limited detection over more than 7 decades of input power, without modifying the APD gain.

The next priority of HgCdTe photon counting detectors will be the increase of the bandwidth of the CMOS amplifier circuit. The aim is to reach single photon sensitivity with a GHz bandwidth. A multi-pixel version of such a detector is a strong candidate for improving the sensitivity and bandwidth in deep-space FSO applications, with the potential to reach higher performance than in SNSPD combined with space compatibility in terms of operating temperature, weight and volume.