2.1

QKD protocol and key-rate

The vast and growing set of QKD protocols differ mainly on the degree of freedom encoding information and how this information is transmitted and processed. In this paper, we consider BB84 as the baseline protocol [16]. It is the quintessential QKD protocol, i.e., it is not overly exotic and forms a reference case to existing literature.

In particular, we consider a ‘vacuum plus weak decoy-state’ BB84 implementation using polarization for encoding [17]. In this implementation, Alice randomly chooses to transmit vacuum, a signal or a decoy-state, with a mean photon number per pulse μ and v photons^†, respectively. For each signal pulse, Alice chooses a bit (‘0’ or ‘1’) and one of two bases to encode this bit (e.g., Horizontal / Vertical or 45 / -45 polarization bases). She then sends the encoded bit to Bob, who randomly chooses a basis, measures, and records the results. Once quantum communication has finished, Bob announces the position of the bits he detected and the basis used. Alice then announces her basis choices. Alice and Bob retain the bits for which the basis employed in the encoding and the decoding processes coincide and discard the rest. This subset is called the sifted key.

Using a segment of the shifted key, Alice and Bob estimate the quantum bit or QBER. When the estimated QBER is below a predefined threshold (e.g., 11%), the protocol continues with error correction. If the QBER is above the threshold, they abort that run of the protocol. The QBER is defined as the ratio of erroneous bits over the total number of transmitted bits. For the protocol that we are considering, the QBER of the signal states (E_μ) can be estimated using [18]:

in which Y₀ is the background detection probability consisting of noise, e_det is the probability of an erroneous detection due to cross talk (e.g., detecting as ‘vertical’ a ‘horizontal’ polarized photon), η = η_Channel + η_detector is the channel and detector loss. Using the QBER it is also possible to estimate a lower bound on the secure key rate:

With the gain of the signal states Q_μ (i.e., the ratio of signal detections by Bob to signal pulses sent by Alice) given by Q_μ = Y₀ + 1 – e^−ημ, f(e) is the efficiency of error correction, here we consider f(e) = 1.22, the binary entropy function is H₂(x) = −x log₂ x − (1 − x) log₂(1 − x), Q_l is a lower bound for the gain of single-photon states given by , Q_v is the gain of the weak decoy state; e₁ is the error rate of single-photon states, which has an upper bound given by with . For additional details, we refer the reader to Reference [18].

The following sections describe our estimations for the channel loss η_Channel and noise required to estimate the secret key.

2.2

Channel loss

To estimate the LEO-to-ground channel’s losses η_channel, we define the link budget, as presented in Source, with both generic items and specific items. The transmit antenna gain G_Tx, the receive antenna gain G_Rx and the free space losses L_fs together are also known as geometric losses and are defined as:

In which θ_div is the divergence angle of the transmit beam. D_R the diameter of the receiver, λ the used wavelength, and L the propagation length. The transmittance of the satellite and optical ground station results from the analysis of the optical systems, e.g., the summation of the optics’ transmission losses and the clipping losses.

The link budget has different items. The transmit channel consists of the transmit antenna gain, the transmittance, and additional divergence losses. These are caused by residual wave front errors of the optical system.

The propagation losses consist of free space loss and atmospheric loss. These losses are modelled, according to [3]:

In which χ_ext is the atmospheric extinction, β_ext is the extinction coefficient, and the extinction coefficient on the ground level, H_OGS the altitude of the ground station, L_Path the path length, ℋ₀ the scale parameter and ζ the zenith angle.

The atmospheric turbulence related losses relevant for QKD are the systematic pointing loss, the average pointing jitter loss, and the wave front error loss. Optical power fluctuations (scintillation) do not play a role for QKD since it is about the average power received [19]. In addition, the beam spread loss is negligible for a downlink since most turbulence is located at the ground layer, which is at the end of the downlink path. Static pointing loss is quantified by the deviation from the Gaussian center, i.e. . Pointing jitter is characterized by the beta distribution [20], and its average can be quantified by: with . This formula only holds for a system without systematic error, and the model is adopted to take the dependence on the systematic error into account. The wave front error loss is due to atmospheric distortion. In this term, the correction is already included and is described in Section 4.

Source focusses on what is in technical context required to establish an QKD channel during day-time conditions. As can be observed the apertures on both satellite and ground terminal side have reasonably large sizes. For a commercial application the technical complexity should be traded-off towards operation range. The operational properties to compromise on could be for instance day-time support or elevation angle range. Night-time conditions would allow for significantly higher link losses.

Table 1.

Link budget for a LEO to ground QKD link for an elevation angle of 20 degrees and a wavelength of 1550 nm and daytime conditions.

2.3

Noise Analysis

For day-time operation of QKD, the two dominant noise sources for quantum key distribution are detector noise ξ_det and background noise ξ_BG [4]. These noise contributions relate to the background detection probability Y₀ = 4ξ_det + ξ_BG. The error sources a assumed to be randomly polarized, so the detection error rate e_det = 0.5.

Detector noise can be modelled as units of electrons:

with f_n the noise count in Hz and f_d the effective detector rate. Background radiation N_BG is quantified by the power per solid angle Ω, collecting area A and the wavelength band Δλ. For 1550 nm, we assume 10 W/m2/sr/μm. The amount of background radiation can be computed by:

With E_p = 𝒽f is the photon energy, in which 𝒽 is Planck’s constant, f = c/λ the frequency of the optical beam and τ is the gate time. The solid angle is dependent on the field of view of the telescope θ_FOV = 2π(1 − cos θ_FOV). The field of view can be reduced to the size of a single-mode fiber when using adaptive optics, without significantly losing link margin. By using a focal length of the telescope of f_T = d_T/2θ_NA, with a single-mode fiber diameter d_SMF = 10 μm and numeric aperture of θ_NA = 0.15, the field of view is θ_Р0Ѵ = d/2f_T. In addition, we assume a spectral filter characteristic of 1 pm bandwidth to do significant background supression.

2.4

Total channel performance

The turbulence and link budget modelling have been integrated into a software tool developed at TNO, which can assess a free space optical link’s feasibility. Figure 1 presents the results of integrating the case, as described, into this tool. Figure 1 (a) depicts the relation between the elevation angle and the link loss. As expected, the distance is longer at low elevation angles, leading to increasing free space losses. The effective path through the atmosphere is longer, and turbulence related losses increase. Figure 1 (b) presents the loss versus key-rate curve for a source rate of 10 GHz, background radiation of 10 W/(m²sr μm) and 100 Hz detector noise. From this figure, we conclude that a key can still be distilled with a loss of 28 dB. Using a detector efficiency of 85%, which accounts for 0.7 dB, a maximum channel loss of 27.3 dB is required. This holds for an elevation angle of 20 degrees and in day-time conditions.

Figure 1.

FSO channel performance: (a) link loss as a function of elevation angle and (b) the key rate as a function of channel loss.

3.1

Optical ground terminal Architecture

The OGT’s primary objective is to capture as much of the QKD signal into a single-mode fiber as possible while limiting the noise contribution. This is apart from its secondary objectives. For QUARTZ these objectives also involve classical downlink and uplink communication. For the OGT the process starts at the telescope aperture. This is a key aspect, as the aperture determines the light collection capabilities of the telescope. However, it also defines a lot in terms of its dimensions, mass, and system complexity. Consider QUARTZ is a LEO based service. The larger the telescope, the more challenging or complex the tracking of the satellite shall be. In addition, the higher the number of modes needs to be corrected for, by the adaptive optics system. For the QUARTZ design, an 800 mm classical Cassegrain telescope was selected with the optical bench on the telescope’s back. The classical Cassegrain configuration was chosen as this has, per definition, the least amount of optical elements and thus the highest transmission. The aperture diameter is on the edge, in terms of diameter, what is commercially available as a semi-standard product and thus maximizes the light collection area. The telescope is based on an a-focal design that reduces central obscuration and minimizes optical elements in the overall design. Due to the a-focal design, the FSM can be the first optical bench element, folding the optical beams immediately to the correct plane. The telescope’s transmission is optimized with customized coatings, achieving a 94% transmission in the applicable wavelength band. Figure 2 presents a mechanical view of the assembled telescope, including the beacons and the optical bench.

Figure 2.

CAD rendering of the telescope, including mounted beacon and optical bench. The left and right images show the telescope from different orientations.

As mentioned before, the QKD detector is coupled to a single-mode fiber. To efficiently couple info, a single-mode fiber unperturbed wave fronts are required. In the use case, the wave front is being distorted by atmospheric turbulence, a potential mitigation technology for this is adaptive optics. Figure 3 presents a high-level schematically overview of the optical bench design, including adaptive optics. The fast steering mirror (FSM) is defined as the interface plane between the Telescope and the optical bench and corrects tip/tilt errors based on the tip/tilt detector’s measurements. After a relay, the two Tx beams are coupled into the common path using a polarization beam splitter. The beam combining or isolation is based on a combination of polarization and wavelength splitting to maximize the suppression of the Tx light straylight which is back reflected toward the Rx channel. Consider that the possibility of using polarization isolation is dependent of the QKD protocol encoding principle used. The Point Ahead Mirror allows to apply an offset with respect to the Rx pointing and accommodate the point ahead angle induced by the satellite’s relative velocity. The deformable mirror corrects the wave front phase errors of the Rx beam. The correction of the 192 actuator deformable mirror is based on the measurements performed by the 17 x 17 sub-aperture wave front sensor and calculated by a real-time control platform. After a secondary relay, part of the beam is split off for the OGT Optical Bench sensors using a dichroic beam splitter, only affecting the signal level in the communication wavelength. The light (both QKD and communication wavelengths) passed by the beam splitter is now coupled into a fiber.

Figure 3.

(a) Adaptive Optics System consisting of a fast steering mirror (FSM), a deformable mirror (DM) a Tip/Tilt Detector (TTD), a wave front sensor (WFS), a point ahead mirror (PAM) and a single-mode fiber (SMF).

3.2

Beam shaper

The limited aperture of the OGS with the central obscuration causes the Gaussian beam transmitted by the satellite to become a top-hat, donut shaped beam at the receiver. Coupling a top-hat profile would result in an ideal coupling efficiency of 81% as an result of the overlap integral between airy disk on gaussian mode field of single mode fiber. Coupling a donut shaped top-hat intensity profile into a single-mode induces additional losses due to the larger mode mismatch and drops down to 64% fiber coupling efficiency. A beam shaper is introduced that fits the intensity profile into a Gaussian profile. It consists of two plano-convex aspheric lenses. The first lens maps the homogenous top-head intensity profile of the collimated beam into a diverging Gaussian intensity distribution at the second lens location. The second aspheric lens collimates this diverging beam while maintaining a near flat wave front with minimal errors. The resulting Gaussian beam profile is then focused into a polarization maintaining single-mode fiber. The beam shaper and corresponding ray mapping is shown in Figure 4 (a), and the intensity before and after shaping the beam profile is illustrated in Figure 4 (b, c). The introduction of beam shaper provides an increase of fiber coupling efficiency from about 64% to 85%, assuming an unperturbed wave front at the telescope entrance.

Figure 4.

(a) Layout of the Rx fiber coupling, WFS and PAT beam paths including beam shaper. (b) Intensity beam profiles before and (b) after the beam shaper. The top-hat profile (with obscuration) shown on the left. The 1/e² diameter of the Gaussian fit of the resulting beam shown on the right.

3.3

Thermal mechanical design

The OGS will be located on various sites globally and have to operate over the different seasons of the year. Posing some challenging environmental conditions on the system to maximize operational availability. To address this, the system was subdivided into zones. Each zone has a set of environmental conditions to cope with and implement the measures to uphold the zonal conditions. e.g., the most unfavourable environmental survival conditions are for the zone outside for which the telescope dome has to cope with. The dome slit reduces to the operational range applicable to the telescope. The most sensitive to environmental disturbances is the optical bench. Temperature variations can introduce misalignments between the different optical paths. Tracking is based on the PAT sensor, but performance is determined at the Rx fiber-coupling path, As such mechanical stability directly affects fiber coupling efficiency. On top of this condensation shall degrade the components inside. For this reason the optical bench is fully enclosed and thermally controlled at a temperature of 20°C ± 1°C. Apart from an optical-mechanical stable design this is achieved in three ways:

Figure 5.

Double optical windows as interface between the telescope subsystem (left). Mounted into the optical bench enclosure (right)

4.1

Fiber coupling efficiency

Due to the atmospheric turbulence-induced wave front aberrations, the efficiency of coupling a free space optical beam into a single mode fiber can be reduced severely. The essence of coupling an optical beam into a focal-plane single mode fiber is given by the spatial match between the PSF and the fundamental mode shape of the fiber (often approximated by a Gaussian). In the nominal case of a flat wave front, the coupling efficiency depends on the match of the Airy disk and the fiber mode. Typical values for the nominal Fiber Coupling Efficiency (FCE) η_nom range from 0.7 to 0.8, with 0.81 being the maximum for an unobstructed pupil; see for instance [26].

Coupling a distorted optical beam into a single mode fiber again depends on the match of the distorted PSF and the SMF mode. Ignoring amplitude aberrations, Ruilier [27] has derived an approximation which relates the effective FCE η_FCE to the variance of the pupil-plane phase ϕ matched to the fundamental SMF mode:

Generally this approximation leads to an under-estimation of the FCE. Furthermore, when decomposing the wavefront phase into Zernike’s modes for instance, the true FCE exhibits a clear dependency on the specific Zernike mode shape. This property has gone lost in eq. (1).

4.2

Laboratory experiments on fiber coupling

In order to find a practical guideline for the design of the AO system, we have conducted a set of laboratory experiments on fiber coupling efficiency. For this purpose TNO’s optical feeder link adaptive optics (OFELIA) breadboard has been used. This experimental set-up has been built to test and verify the performance of Adaptive Optics pre-correction for optical feeder links [13, 14]; see also Figure 6. It consists of an AO system, a Tx and Rx channel, and a focus camera to verify beam quality. For the experiments the breadboard was modified to assess the fiber coupling efficiency (FCE) of optical beams under partial AO correction of atmospheric-turbulence-induced aberrations.

Figure 6.

OFELIA breadboard (a) fully integrated breadboard (b) schematic overview of the breadboard with several beam splitters (BS), a focus camera (FC), a tip-tilt sensor (TTS), a wave front sensor (WFS), a real time computer (RTC), the transmitter (Tx), the point ahead mirror (PAM), the fast steering mirror (FSM) and deformable mirror (DM).

A collimator is used to generate a receiving laser beam. This laser beam runs over a deformable mirror (DM) and is split towards a wave front sensor and a single mode fiber. The single mode fiber has a nominal coupling efficiency of 0.74, which is mainly caused by alignment error. Residual wave front errors, that cannot be eliminated by the AO system, further limit the coupling efficiency.

The control scheme of the AO can be represented by the block diagram in Figure 7.

Figure 7.

Control architecture with C the controller, H the modal decoupling matrix, P the DM and electronics, S the WFS and G^† the decoupling matrix. In addition the signals are r the reference, e the error signal u_m the modal actuation signal, u_a the actuator signals and ϕ_DM the output, which is the deformation of the DM. The distorted wave front is represented by ϕ_tur and the residual wave front error by ϕ_res.

The deformable mirror and its electronics (P) actively correct for the wave front aberrations. In order to do so, the wave front is measured by the wave front sensor S. For the adaptive optics system architecture, we have chosen a modal correction scheme. Decoupling the modes from the wave front sensor measurement is done via the decoupling matrix G^†. The matrix H is the modal coupling matrix, which couples the modal control signals u_m into the actuator control signals u_a. Hence, a reference signal r₂ corresponds to a disturbance signal of the modal shapes (e.g. Zernike shapes) at the DM, i.e. G^†SPH ≈ I for low spatial frequencies.

In order to simulate turbulence-induced wave front errors in the experiments, the DM has been used. Individual Zernike modes with various amplitudes have been imposed on the DM, both as static and dynamic aberrations. Although using reference, r₁ would yield the modal shapes with higher accuracy, reference r₂ yields more flexibility. Adding a signal to r₂ is equivalent to a disturbing signal ϕ_tur, so the effect of the AO feedback loop can be verified.

The coupling results for the static Zernike modes are shown in Figure 8 (a). The individual Zernike modes have been imposed on the DM through reference r₂, in open loop (i.e. without the use of the feedback loop over the controller C). The graphs show that the minimum WFE equals 90 nm RMS and consequently the FCE in these tests has a maximum value of 0.68.

The measured FCE values are fitted on the exponential function in Eq. 1. To further reduce the complexity of the approach, the SMF mode matching operation has been left out and the plain wave front phase variance is used instead. In this way, effectively the Strehl ratio (under the extended Marechal approximation) now determines the reduction of the nominal FCE. From Figure 8 (a), it can be clearly observed that the analytic model and the least-squares fit correspond well. Also, the analytical model has a slightly lower FCE than the least-squares fit. Still, the variance of the measured FCE values around the least-squares fit is relatively high. A possible explanation of this phenomenon can be found in the specific behaviour of the FCE as a function of the Zernike modal shapes. This corresponds to the analysis in [27].

Figure 8.

Fiber coupling results in the lab (a) measurements with static, individual wave front aberrations (b) measurements with dynamic aberrations following a Kolmogorov spectrum.

Next to static Zernike mode aberrations, also dynamic tests, following the Kolmogorov spectrum have been done, which results are shown in Figure 8 (b). This approach is based on the Kolmogorov spectrum and the assumption of frozen flow (Taylor’s hypothesis). The spatial power spectrum per mode follows from the filter expressions presented in [28]. This can be converted to a temporal spectrum by inserting the transverse wind speed model, e.g. using the Bufton model. An example of this approach is depicted in Figure 9, where the spatial mode, the power spectrum and an amplitude time trace are shown. Time traces have been obtained for 40 Zernike modes, and have been superimposed using reference r₂.

Figure 9.

Dynamic response of astigmatism. (a) the spatial shape, (b) the power spectrum, the blue line is the computed spectrum, the red line is a filter approximation, (c) a time trace with the spectrum of (b).

The FCE measurements for the dynamic Zernike test show a similar behaviour as a for the static modes. It can be concluded that the basic approximation of Eq. (1) has a slight conservatism, hence it is a safe approach to use for analysis and design of the AO system. In conclusion, the basic approximation of FCE leads to the design of a regular AO system in which the wave front phase variance is to be minimized.

Besides this relation between WFE and FCE, also the ability of the AO system to follow the relation for spatial correction and temporal correction are verified. For the spatial error, turbulence signals for all Zernike modes have been injected at r₂ again. Now the controller has been switched on for a limited range of modes up to all available modes (see Figure 10 (a)). As is shown in the graph, the AO system is clearly able to follow the trend predicted by the model. For the temporal behaviour, the disturbance signal for all Zernike modes has been injected on r₂ again, and now the bandwidth has been varied from 16 to 125 Hz. Also here, this plot shows good resemblance with theory. To gain additional insight, in Figure 10 (b), the different colours represent the different bandwidths. For higher bandwidths, the resulting FCE is higher and more confined around the fitted curve. Hence, it is concluded that the AO system performs according to the known relation between wave front error and spatial and temporal correction.

Figure 10.

Experimental results for model verification. (a) relation between corrected Zernike modes and WFE. (b) relation between bandwidth and wave front error.

4.3

Adaptive Optics system design

The link budget in Table 1 aims at a wave front error loss of 4.3 dB, assuming a nominal FCE of 0.81. This implies that the AO system will have to reduce the variance of the wave front phase fluctuations down to a value of roughly 220 nm (RMS). This is similar to the requirements in QUARTZ requirements, in which the link budget allows 6 dB for the combination of optical transmission loss and fiber coupling loss. Based on the design about -1.4 dB is induced by transmission losses. Since the telescope aperture impacts the dimensioning of the AO system, it should be emphasizeded that the analysis below is performed in the context of the QUARTZ project.

To estimate the uncorrected beam distortion by atmospheric turbulence, a Hufnagel-Valley Cn2-profile has been assumed. A ground layer strength of 1.7 · 10⁻¹⁴ and a pseudo-wind of 21 m/s leads to the common HV5/7 model. Furthermore, the wind speed at ground level is set to 13.9 m/s using the Bufton wind profile. With a satellite Tx diameter of 0 60 mm and a Ground station Rx diameter of Ø 800 mm, the following beam distortion parameters can be derived as a function of elevation angle; based on expressions in [28].

The underlying expressions for computing turbulence parameters are given below. The Fried coherence length r₀, for plane wave can be evaluated as:

in which ζ is the zenith angle, is the wave number, h_OGS and h_Sat are the altitude of the OGS and the satellite respectively, indicates the refractive-index structure function as a function of h, the altitude along the path. The Greenwood frequency, relevant for the required bandwidth of the AO system is:

Here v_wind(h) is the (transverse) wind speed as a function of height. Specifically for the Z tip-tilt mode, the Greenwood frequency reads:

To estimate the amplitude of the tip-tilt mode the angle of arrival (AoA) is relevant. Its variance is:

with D the aperture diameter. Evaluation of the these distortion properties for several elevation angles leads to the numbers in Table 2.

Table 2.

Beam distortion parameters.

Clearly, these beam parameters show a strong dependence on the elevation angle. A first conclusion from this analysis is that the wave front error is relatively large (1.4 μm at 20 degrees elevation) and hence feeding the uncorrected beam to the fiber is not an option at all. Adaptive Optics is a prerequisite to arrive at acceptable values of FCE.

The focus in designing the AO system has been put on the case of 20 degrees elevation angle. Specific design choices for the AO system are:

Table 3.

Main parameters of the AO system

In the selection of AO system parameters, the spatial error and the temporal error have been balanced. Next to the fitting error and the latency error, the vibrations-induced tip-tilt error (which are the non-turbulence induced errors, like alignment errors, vibration induced tip/tilt errors, drift and stability) is a major source of residual phase error. The typical use case pre-scribes that the OGT is able to be placed on high-rise buildings. These buildings come with significant vibrations induced by wind and ground vibrations, tilting the full OGT. These vibration not only affect the telescope pointing and tracking directly, but also indirectly by introducing vibration in the telescope structure itself. Analysis has shown significant tip/tilt errors, and although being suppressed by the AO system a considerable residual could remain. Given the AO system parameters and the description of the individual error sources in for instance [25], we arrive at the budget given in Table 4. All error sources have been verified using an end-to-end AO system simulation.

Table 4.

System errors determining the residual WFE at 20 degrees elevation angle

In conclusion, the residual of 224 nm (RMS) wave front error would lead to a reduced fiber coupling efficiency by a factor of 0.43 and hence an additional loss of 3.6 dB above the nominal fiber coupling loss in the case of a flat incident wave front. Having a nominal loss of 0.85 (as created by the beam shaper), the total fiber coupling loss using a dedicated AO system then arrives at 4.3 dB.

5.1

Perspective on QKD-node Earth Orbit

The presented OGT design is primarily driven by the BB84 protocol with a satellite service node orbiting at LEO. The BB84 approach is relativemature, for which subsystems have been demonstrated in laboratories or in relevant environments and even a comparable full end-to-end system has already been demonstrated [14, 15]. With an overpass time of (only) a few minutes, a LEO satellite based service could offer near global coverage. Altogether, this makes it a logical step for a first product line and QKD service.

Several alternatives for both satellite orbit and QKD protocol exist. For instance a QKD service from a GEO-satellite system would not be limited by the overpass time of a LEO-node and the link can be preserved for a long time. On the other hand, a GEO satellite link will encounter a much higher link loss, which inevitably leads to a lower key rate. Technologies – such as discussed in the previous sections - that will lead to a reduction of the link loss very much apply to the GEO-node case as well.

A relatively large region – say the size of a continent - could be serviced by a single GEO-satellite continuously. This makes it an attractive node for entanglement-based QKD, sharing keys between 2 (very) remote users; from a LEO orbit the distance between Alice and Bob would be restricted. Entanglement based QKD is a very attractive protocol from a security perspective [8]. This makes it suitable for users demanding ultimate security such as governments and MOD’s. For now however, trusted-node LEO systems based on BB84 for instance are regarded as a more mature and cost effective solution, which offer an acceptable security level for the majority of users.

5.2

Perspective on ground stations applications for QKD

The distribution of quantum keys over FSO channels offers several benefits. First of all, it overcomes the distance limit of fiber-based QKD, which is caused by the higher transmission loss. Second, naturally occurring obstructions, such as oceans or mountains can make it rather complex and expensive to deploy a fiber network. Then, an FSO link can be an attractive, alternative channel to distribute quantum keys. This FSO link could be a direct ground-to-ground coonection or via a satellite node.

Achieving a sufficiently high secure key rate for a satellite-based QKD protocol requires advanced technologies to restrict the quantum channel loss and noise levels. Reduction of the OGS cost will therefore remain to be a critical design driver for this generation of QKD systems. For the specific OGT design discussed in this paper, there are two sub-systems that play a significant role in the overall cost: the telescope and the AO system. In this QUARTZ design, a main driver is the need for a single-mode fiber based detector. Therefore, optical imaging quality of the receiver was required, leading to the demand of a very low wave front phase error and hence an AO system and a high optical quality telescope. Future developments in for instance multi-mode fiber coupled or free space detectors with large detector areas may be beneficial to effectively lead to a cost reduction for OGS’s in satellite-based QKD.