1.

INTRODUCTION

Today and future space missions are expected to transmit huge volumes of scientific data, including high-definition images and video especially from Space to Earth.^1,2 Moreover, communications and network providers have to face the ever-increasing demands for high capacity from business customers, which are driven by many different applications (e.g., e-commerce, industrial automation, Internet of things (IoT) etc.). This leads to a significant increase of the required bandwidth and the need for new technologies capable of transmitting at much higher data rates than present. However, the licensed Radio Frequency (RF) spectrum is not enough,¹ so that the unique valid alternative is the Free Space Optical Communication (FSOC).

Indeed, both space agencies and private companies invest today hugely in several missions and programs to establish Inter-Satellite-Links (ISLs) and the most challenging Feeder-Links (FLs), passing through the terrestrial atmosphere. Among them, we can cite National Aeronautics and Space Administration (NASA),³ European Space Agency (ESA),^4,5 Japan Aerospace eXploration Agency (JAXA),^?,6 and SpaceX company.^7,8 To this aim, future FSOCs will take advantage of 50-years R&D in the optical fiber industry, including developments of high-speed photonic components, wavelength-multiplexing, modulation and coding schemes.

In the near future, FSOCs can offer very high data rates, while satisfying the Size Weight and Power consumption (SWaP) constraints. Indeed, at a given transmitter power, the received intensity can be much greater thanks to the much lower beam divergence (proportional to λ/D, where λ is the carrier wavelength and D is the aperture diameter) than in RF communications. On the other hand, this leads to stricter requirements for beam pointing (e.g. around few micro radians), so much that, usually, an accurate Pointing, Acquisition and Tracking (PAT) technique is adopted to satisfy them.^{9, 10}

However, the performance of a very high data rate FL can be quite limited by the propagation through the atmosphere (e.g., absorption, scattering and, most important, turbulence). As in uplink the channel impairments take place close to the transmitter, while in downlink they take place close to the receiver, this leads to asymmetric impairments of the communication in the two cases.¹¹

Today the demand for a very high-capacity downlink is very urgent¹ and this is particularly challenging for the Geostationary (GEO) satellites (typical distance from Earth surface around 36000 km). Thus, we present here the theoretical assessment of the GEO-Earth downlink power budget of a Wavelength Division Multiplexing (WDM) optical communication systems, with a realistic approach. The FSOC system is possible thanks to a new solution based on transparent terminals described in 2, and leverages upon existing optical fiber communication technology subsystems. We choose 40 wavelength channels in the C-band (1530 nm to 1565 nm), each carrying a 40 Gbit/s On-Off Keying (OOK) signal with a Reed-Solomon (RS) Forward Error Correction (FEC) code. This allows to reach a total aggregated throughput of 1.6 Tbit/s and we will see that the system design can be successfully carried out, provided that a careful design is carried out. We will also determine the conditions under which the system can provide the expected performance.

2.

SCENARIO

Despite the great potential of FSOCs, feeder links can suffer from significant drawbacks due to the atmosphere, including attenuation (e.g., different types of clouds), scattering and refractive index fluctuations, which can resulting into fades at the expense of lowers link availability compared with the RF band.

In Fig. 1, we show the typical scenario for feeder links, where an OGS is receiving optical signals from a satellite. For the sake of clarity, uplink and downlink are indicated with two different colors. In downlink, the broadening of the optical beam is mostly due to diffraction, and a minor spread is due to the atmospheric effect or variation in the beam steering. In fact, close to the receiver, the wave can be modelled as a plan wave. The scintillation penalty is in general quite small with respect to the uplink. In the downlink signal, the main contribution for degradation of the signal quality is due to atmospheric losses (related to absorption and scattering), beam spreading due to diffraction effect, and loss of spatial coherence. To mitigate all those impairments affecting the propagating signal some countermeasures might be adopted.¹² The overall scenario parameters are summarized in Table 1.

Figure 1:

Satellite-ground communication scenario.

Table 1:

Summary of scenario parameters for this study.

In this type of FSOC system, a key role is played by the Adaptive Optics (AO) at the ground station. In fact, in the transparent terminal represented in Figure 2, the received WDM signals must be efficiently coupled into a single mode fiber (SMF). To this aim, the phase distortions due to turbulence have to be removed by the AO, which corrects the distorted wavefront to increase the coupling efficiency. Then, the signals are amplified by a low noise Erbium-doped Fiber Amplifier (EDFA): due to high losses, the power at the input of this Low Noise Amplifier (LNA) is quite low, which has a direct impact on the Optical Signal to Noise Ratio (OSNR). The system performance is then basically determined by the OSNR value, a regime that we can call OSNR-limited. Then the signals are WDM-demultiplexed and filtered by a common component (Arrayed Waveguide Grating (AWG)). Finally, each wavelength signal is translated to the electrical domain by a common opto-electronic board.

Figure 2:

Structure of the WDM FSOC system, including (de-)multiplexers, amplifiers (EDFA) and telescopes.

3.

TRANSMISSION IMPAIRMENTS

3.1

Absorption and scattering

Along the atmospheric path, the optical signal experiences an attenuation caused by absorption by suspended molecules, Mie scattering by molecules or aerosol particles, and Rayleigh scattering. In C band, Mie scatter (especially due to fog) is dominant since the scatter site is on the order of the wavelength of light.^{16, 17} As the optical beam propagates, its attenuation is expressed by the Beer-Lambert law as function of the atmospheric transmittance¹⁷

where I(0) is the transmitted laser power at the source and I(L) is the laser power at a distance L, α is the total attenuation factor in km⁻¹, which includes both absorption and scattering. Although small,¹⁷ we also include the absorption contribution. We then express α as:

where α_abs and α_scatt are the absorption and the scattering contribution, respectively. For the given atmospheric conditions, at the 86° of elevation angle we computed the first input using MODTRAN®. We set the Mid-Latitude Summer model and a rural aerosol composition with 50 km visibility (clear conditions) and 15 km visibility (slightly hazy conditions). In our band of interest, the α_abs can be approximated to a constant value resulting respectively around 0.95 and 0.91 (see Fig. 3).

Figure 3:

Atmospheric transmittance in C-band in a Space-to-OGS path at 86° of elevation angle. A rural aerosol composition with a surface meteorological range, or visibility, of 50 km (clear conditions, on the left) and 15 km (slightly hazy conditions, on the right). The data refers to the case of an observer located at 2 km above sea level.

In slightly hazy conditions, the aerosol extinction coefficient is almost constant up to 1 km. In clear conditions, the vertical distribution of aerosols is exponential. The α_scatt coefficient is given by Kruse-Kim model^{18, 19} as

where α_scatt is given in dB/km, V is the visibility in km, λ is the wavelength expressed in nm, and q(V) is the particle size distribution,¹⁶ which can be expressed by the Kim coefficient.¹⁸ The total atmospheric loss due to absorption and scattering is estimated by

where L_scatt(θ) in km describes the slant range varying the elevation angle θ, α_abs is expressed in dB, α_scatt is expressed in dB/km

3.2

Scintillation

The effect of the atmospheric turbulence was characterized by the refractive index structure parameter . To describe this fluctuation strength along the vertical propagation path, we used the Hufnagel-Valley model, emended to include an OGS not located at sea level.²⁰

where w_rms is the root mean squared wind speed in [m/s] (pseudo-wind described in¹² following the Bufton wind profile), h is the satellite altitude [m], h₀ the OGS altitude [m] and is the refractive index structure parameter at ground level in [m^−2/3].

From a communication point of view, the effect of the turbulence is a statistical fluctuation of the irradiance seen by the receiver, which can be modelled mathematically with a probability density function (pdf). In the last decades, different pdfs were proposed. The widely used ones are the log-normal distribution, which was demonstrated a reliably model in case of weak turbulence, and the Gamma-Gamma distribution, which could be used with a wider degree of generality.¹² In our work, thanks to the aperture averaging, which strongly reduces the value of Scintillation Index (SI), we used only the log-normal pdf.

Since atmospheric conditions can widely change, in our study, we considered two atmospheric and meteorological conditions, i.e. a favourable case, where the conditions pose very limited impairments, and a stressing case, where the propagation is highly affected. These two conditions are characterized by the parameters reported in Table 2. In these two cases, varying the elevation angle from the zenith to 30 deg we calculate the most relevant atmospheric parameters¹² (i.e., atmospheric Fried parameter r₀, Greenwood frequency f_G, Isoplanatic angle (IPA) θ_IPA, and Rytov scintillation index ) in the two conditions. The results are reported in Table 3. As we can see, the most challenging link is for the “stressing atmospheric and meteorological conditions” at the lowest 30 degree elevation angle with an OGS at the sea level. In this scenario, the Rytov SI is quite high, causing a coherence diameter few centimeters.

Table 3:

Calculated atmospheric parameters based on atmospheric conditions defined in Table 2.

ATMOSPHERIC PARAMETERS
	Favourable conditions	Stressing conditions
	At 86° elevation	At 30° elevation	At 86° elevation	At 30° elevation
fG at the sea level [Hz]	14.4	21.8	31.7	48
fG at 2370 m [Hz]	11.4	17.2	14	21.3
r0 at the sea level [cm]	19.3	12.7	5.6	3.7
r0 at 2370 m [cm]	67.4	44.4	35.7	23.5
θIPA at the sea level [μrad]	26.2	8.7	25.6	8.5
θIPA at 2370 m [μrad]	35.0	11.5	35.3	11.6
at the sea level	0.06	0.23	0.14	0.50
at 2370 m	0.03	0.12	0.04	0.13

A key parameter of the system is the size of the OGS telescope. Its immediate role is to determine the collected signal power, thus affecting the final OSNR at the pre-amplifier output. Furthermore, it can also help reducing the impact of scintillation: as known, the irradiance at the receiver is a random variable, whose fluctuations are characterized by the SI. In our case, the receiver aperture of the OGS telescope is usually much larger than the atmospheric coherence diameter r₀.¹² Thus, the telescope collects several correlation patches partly compensating the effect of the fluctuations.

3.3

Pointing loss

As said, we did not consider in our model the PAT subsystem, which works before the link is established and during communication. However, once the two transceivers have locked onto each other, mechanical vibrations and noise in the tracking system can cause beam jitter or uncontrolled movements resulting in a residual pointing error θ_point.¹⁴ This miss-pointing error is computed assuming a maximized transmitting antenna gain G_T with respect to pointing losses having a diffraction limited Gaussian beam in the paraxial case.^{21, 22} The residual transmitter pointing losses are modeled by²³

where a Gaussian beam at the transmitter is assumed.

4.

APERTURE AVERAGING AND AO SYSTEM

From the design point of view, it is known that we must equip the OGS with a sizable telescope and a performing system of AO, correcting a high number of Zernike modes.²⁴ Moreover, to further mitigate the atmospheric effects, the OGS should be sited at high altitude. Therefore, we introduce in our model these countermeasures, which are the key to detect the signal on the ground.

The aperture averaging effect becomes significant when the OGS telescope diameter is much greater than the minimum coherence diameter r₀, thus mitigating the intensity scintillation induced by the atmospheric turbulence.²⁵ In Fig. 4, we report the computed SI in the strong turbulence conditions, for a point receiver (black dots) and for 1.2 m telescope, at OGS sited at 2370 m. As can be noted, the aperture averaging significantly reduces the value of the SI, strongly reducing the fluctuations at the receiver side.

Figure 4:

Effect of aperture averaging vs. elevation angle in the stressing condition, with a 250 mm GEO telescope: we report the SI for a point receiver (black dots) and for a 1.2 m wide telescope at 2370 m altitude (red dots).

When the ratio D/r₀ is greater than 1, the AO technique (partially) compensates for wavefront phase errors due to atmospheric turbulence, this is needed to increase the efficiency of the coupling of optical power from free-space to a single-mode fiber, which is the input to the pre-amplified WDM receiver. The modeled AO system is based on a Shack-Hartmann wavefront sensor (WFS), which measures the distortion on the downlink, a deformable mirror (DM) and a tip tilt mirror (TTM).^{24, 26–29} The performance are estimated according to.^{24, 26–29}

The reduction of scintillation and coupling efficiency are limited by the actual performance of the AO correction system. As example, in Fig. 5 we report the coupling loss values calculated as a function of the compensated Zernike modes, in the two cases of favourable and stressing conditions at 30° of elevation, in the case of a 1.2 m-wide telescope OGS at 2370 m altitude. As we see, the number of modes that we must be corrected is quite higher in the second case. In both cases, a good performance is obtained for at least 100 corrected modes. The two curves saturate at around -2.5 and -2.9 dB values, respectively, which are fully acceptable values.

Figure 5:

Effect of AO on fiber coupling: we report the coupling loss vs. the number of compensated Zernike modes, for the favourable condition (black solid curve) and the stressing condition (red dashed curve) at 30° of elevation, for a 1.2 m-wide telescope (red dots) at 2370 m altitude.

5.

LINK BUDGET EVALUATION

Based on the previously defined atmospheric conditions, we investigated the system feasibility, by computing the link budget.

First, we started our analysis by considering a system configuration that is based on existing technologies. As an example, this is the case of telescope size, both on satellite and at OGS. Namely, in the following analysis, the system assumes the diameter of the GEO and OGS telescopes to be 135 mm and 1000 mm, respectively. Both telescopes have an obscuration factor of 0.3. We refer to this case as System-A: its detailed features are summarized in Table 4.

Table 4:

Summary of initial system parameters for this study (System A).

All channels are intensity-modulated at 40 Gbit/s. The system employs a FEC scheme for sensitivity improvement based on a RS(255,223). The AWG multiplexer has 100 GHz of channel spacing and 50 GHz 3 dB equivalent bandwidth on each optical port. The signals are then amplified by a single 50 dBm space qualified optical booster amplifying simultaneously all the optical WDM channels.¹³ After free-space propagation, the signals are collected by the telescope, amplified by a EDFA (30 dB optical gain, 4 dB noise figure). Then the channels are demultiplexed by a similar AWG and each of them is detected by a conventional NRZ-OOK receiver. A summary of these parameters is presented in Table 5.

Table 5:

Summary of WDM communication system parameters.

After careful evaluation, we derive that this system configuration can not meet the requirements neither for favourable nor for stressing conditions. As we see in Table 6, in order to make the link work, the expected transmitter power should indeed exceed the limit of 50 dBm, which is already challenging. We note that free-space propagation is by far the most significant effect. Thus this is a clear indication that wider optics is needed both at transmitter and receiver size to increase the antenna gains.

Table 6:

Summary of link budget for System A. The link would be considered feasible if PTX ≤ 50 dBm, which is not found in any condition.

We therefore studied the total link loss (including all propagation losses and telescope gains) as a function of the size of the transmitter (TX) and receiver (RX) telescopes. In all the following cases, we assume also that on the satellite we can have available telescopes with no obscuration factor.³⁰ However, it is known that an OGS telescope without obscuration would be hard to be realized. Hence, we assumed that at OGS the telescope still has an obscuration of 30%.

We report in Fig. 6 the corresponding total losses at 2370 m altitude versus the diameter values of the TX and RX telescopes, for the favourable conditions, at elevation angles of 30 deg (left) and 86 deg (right). We spanned the GEO diameter from a minimum of 50mm to 500mm, with a step of 50 mm, and the OGS diameter from 200 mm to 1000 mm, with a step of 100 mm. The figures indicate the overall link losses, taking into account all the impairments and gains described in the previous sections. In the following Fig. 7, we present the data obtained, for the same elevation angles, in the stressing conditions. As expected, the optimal parameters are practically the same, although with higher total losses.

Figure 6:

Estimated total losses (in dB) at 2370 m altitude as function of GEO and OGS telescope diameters under favourable condition; elevation angle is 30 (a) and 86 (b) degrees.

Figure 7:

Estimated total losses (in dB) at 2370 m altitude as function of GEO and OGS telescope diameters under stressing condition; elevation angle is 30 (a) and 86 (b) degrees.

From these figures, we see a minimum on the link losses (red colored) with a GEO diameter of 250 mm and a OGS diameter of 1200 mm: we define this configuration as System B. Further increasing the transmitting telescope diameter leads to a steep increase of the pointing error, since the diffraction angle decreases, while reducing it, the diffraction angle increases, reducing the received optical power on the optical axis.

As can be seen, a small size of the receiving telescope clearly leads to a reduction of the collected light. Further increasing the size of the RX telescope, can still increase the collected light. This might be achieved, however, at expense of a worsening of AO performance, which strongly depend on the ratio D_r/r₀, in downlink. Therefore, an optimal TX diameter is found, at around 250 mm. On a satellite, this is a value that is not common, yet it seems quite feasible. A corresponding optimal value for the OGS could be expected from the above figures, but this value would exceed by far 1.2 m, which seems to be beyond the technology limits. Therefore, we did not consider having a telescope wider than 1.2 m. In conclusion, we included the two GEO and OGS telescope diameter values in the baseline for our further analyses. The results of this analysis are detailed in Table 7, which reports the technical features of System B. We note that, here, we are again assuming a maximum output power of 50 dBm. Actually, System B represents the best compromise solution to achieve the expected capacity in the considered link type.

Table 7:

Summary of system parameters for this study.

In Fig. 8, we report the minimum required transmitted optical power to effectively close the link as a function of the elevation angle for both favourable and stressing conditions, in System B. As can be noted, the dependency on the elevation angle is limited, since the aperture averaging strongly reduces the scintillation index and therefore the dependency on the elevation angle, which usually has a strong impact. We note that the system needs more power in the stressing case, as expected. However, we estimate that it could work fine in all cases, provided that the maximum required optical power in the worst case is still 50 dBm, which is compatible with the new generation space-graded booster amplifier under development. However, the 30 degree case is actually borderline if we assume the stressing atmospheric conditions: in that case, the required power is exactly matching the 50 dBm value. This leaves no space for additional margins.

Figure 8:

Minimum required transmitted optical power for System B.

In order to conclude our analyses, in Table 8 we report the detailed link budget considering the System B. We note that, in the favourable conditions, around 47 dBm would be enough in all cases. This is achieved thanks to the fact that both telescopes are much wider than in System A, which overtakes the fact that the pointing error loss is a bit higher.

Table 8:

Summary of link budget analysis: System B. The link is considered feasible if PTX ≤ 50 dBm.

6.

CONCLUSION

We have presented a complete numerical investigation of the feasibility of 1.6 Terabit/s feeder link (downlink). The assumed configuration is made of 40 Gbit/s OOK signals, with a RS FEC, in the C-band. The approach strongly relies upon the assumption that current photonic components can be used on a GEO-satellite and that a high power booster amplifier (≤ 50 dBm output power) and optimized optics can be deployed. We considered two scenarios of atmospheric conditions, i.e. a favourable and a stressing case.

Indeed, we found that the best (realistic) operating conditions are an un-obscurated 25 cm-wide GEO transmitter telescope and a 1.2 m-wide OGS receiver telescope sited at 2370 m altitude. These two values are achievable with existing technology, although quite challenging. They also require suitable AO system, able to correct the wavefront distortions due to turbulence effect, thus increasing the overall detected power. In the considered different atmospheric conditions, we found that, in both cases, the system can be designed to operate, although with higher TX power in the stressing condition.

Finally, we note that we did not include other modulation formats (e.g. DPSK) neither coherent detection. We expect, however, that, in the link budget, the benefits of these solutions may be on the order of few dB’s (considering that the system is OSNR-limited anyway). This would allow a reduction of the total output power, but the optics would likely stay untouched.

The present design indicates specifications of the system that are not far from those of existing elements. In particular, optical terminals with 135 mm diameter at GEO are already in place on satellites; they may be upgraded to a wider size. At OGS, we recall that a 1 m ESA telescope is already installed in the Teide Observatory in the Canary Islands (Spain) at 2370 m above the sea level. Not by chance, this altitude value is what we assumed in our calculations. Lower values would result in higher impairments.

On the other hand, photonic devices and subsystems compatible with our WDM design can benefit from a strong heritage from optical fiber communications, but they still have to be space-qualified. Probably, among the various building blocks, the most challenging technology is the high-power booster, which should be developed and space-qualified. Although space-qualified 40 dBm amplifiers exist, 50 dBm are very challenging. Finally, PAT and AO blocks should also be refined to meet the system requirements. These results can be used as first guidelines for FSOC system designers of future links.