Receiver synchronisation based on a single dummy frame for DVB-S2/S2X beam hopping systems Xavier Giraud

Guy Lesthievent

Hugo M´eric

Cabinet NOVACOM Email: [email protected]

CNES Email: [email protected]

CNES Email: [email protected]

Abstract—We study the receiver synchronisation in a DVBS2/S2X beam hopping system. In such a scenario, each beam is illuminated during defined temporal windows. As a consequence, the receivers do not receive a continuous signal which strongly disturbs their synchronisation. To tackle this problem without using a specific format, we introduce a dedicated dummy frame with a synchronisation field that enables the receiver to lock with one frame even with poor channel conditions (a signal-tonoise ratio down to -10 dB). This solution is fully compliant with the existing standards. Through simulations, we present the performance of our scheme and demonstrate its effectiveness for DVB-S2/S2X systems.

switching and system uncertainty, hardware complexity, etc. Before going any further, we shortly discuss the impact of the super-frame (or any possible framing) length in a beam hopping system. Even at 500 MBauds, the super-frame lasts more than 1 ms as shown in Fig. 1c. Thus the gap between two consecutive illuminations may be quite long. A long period 270 symb. 450 symb.

611 820 symb.

SOSF SFFI

Ressource allocation (format specific)

I. I NTRODUCTION 612 540 symb.

(a) Super-frame (optionnal) format composed of: start of super-frame, super-frame format indicator and the resource (several PLFRAMES as depicted in Fig. 1b and pilots) 80 bits

BB header

BBFRAME

Bits

Data field

FEC encoding

FECFRAME

BCH FEC

BBFRAME

Bit mapping to symbols

K BCH bits

XFECFRAME

Slot 1 Slot 2

LDPC FEC

N LDPC −KBCH bits Symbols

Slot S

PL framing

PL

PLFRAME

header

Slot 1

Slot 16

90 symb. 90 symb.

Pilot block

Slot S

36 symb.

(b) DVB-S2/S2X framing (mandatory) 10−1

Dummy Super

−2

Frame duration [s]

Very high throughput satellites combined with multi-beams systems provide a significant increase of the available capacity. However on-board resources (especially high power amplifiers) are limited due to consumption, mass and dissipation. Moreover, the traffic demand is time varying and differs between beams. One solution to properly allocate the resources and solve previous constraints is beam hopping. To that end, beam hopping proposes a temporal allocation of the resources by illuminating the beams in various temporal windows (thus avoiding a complicated frequency allocation). First studies regarding beam hopping dated back to 2006 [1]. Since then, the literature has been growing [2]–[9]. Previous works cover the payload architecture [1]–[4], [6], the system performance combined with resource allocation algorithms [1], [3]–[7] and the receiver synchronisation [8], [9]. In this work, we study the receiver synchronisation. This means the frame, frequency, phase and timing synchronisations as the receiver needs to lock after a period without signal. All previous works studying synchronisation for beam hopping systems rely on the super-frame format introduced in DVB-S2X [10, Annex E] and depicted in Fig. 1a (we also remind the DVB-S2/S2X framing in Fig. 1b). In [8], [9], the authors use this new framing as a solution to synchronisation issues in beam hopping and very-low signal-to-noise ratio (VL-SNR) modes. They present synchronisation performance such as false alarm and missed detection probabilities. Even if the super-frame presents some benefits, its adoption by the satellite community is not that simple as it has many impacts on the transmitter and the receiver: special header (not backward-compatible with DVB-S2), long and fixed framing size (the super-frame lasts more than 6 ms for Rs < 100 MBauds), limited granularity, insufficient guard symbols for

10

QPSK 8PSK

16APSK 32APSK

10−3 10−4 10−5 10−6

50

100

150

200

250 300 Rs [MBaud]

350

400

450

500

(c) Duration of normal frame (64 800 encoded bits) with pilots

Fig. 1: DVB-S2/S2X framing

BHSync

BHSync

BHSync

BHSync

BHSync

BHSync

DU

MM Y

BHSync BHSync

BHSync

BHSync

BHSync BHSync

BHSync BHSync BHSync

BHSync BHSync

BHSync BHSync

BHSync BHSync

BHSync

BHSync

BHSync BHSync

BHSync BHSync

BHSync BHSync BHSYnc

GROUND − TRANSMITTERS

BHSync

GROUND − RECEIVERS

HPA Gateway

BHSync

BHSync

Beam antennas

BHSync

BHSync

Gateway antennas

BHSync

BHSync

MM Y

BHSync

DU

Dummy frame for padding

BHSync

Dummy frames for switching and synchronisation

BHSync

Gateway

BHSync

BHSync

HPA

BHSync

BHSync

SATELLITE

BHSync

A timing synchronisation exists between the gateways and the satellite.

Fig. 2: Use of a single dummy frame, noted BHSync, to manage switching and receiver synchronisation in a beam hopping system. If several time slots are dedicated to the same beam, the BHSync frame is skipped.

without signal impacts the synchronisation at the physical layer, the delay at the transport layer and even the application in case of video transmission. To avoid long period without signal, a shorter framing may be considered. Let consider a framing of around 200 000 symbols. Using the MODCOD distribution presented in [11], the average PLFRAME length is roughly 25 000 symbols and the framing can carry only 8 frames, which may not be enough. In the current work, we aim at providing an alternative to the super-frame by proposing a dedicated dummy frame, called BHSync, with a synchronisation field (thus compatible with the DVB-S2/S2X standard). The idea to rely on dummy frame was introduced in [12] to address synchronisation issues with standard modems. However the previous scheme requires around 20 dummy frames; in our current work a single dummy frame is sufficient thanks to its synchronisation field. We illustrate our idea in Fig. 2: the synchronisation between the gateways and the payload (necessary in a beam hopping scenario, but out of the scope of this paper) enables the gateways to insert a dedicated dummy frame, the BHSync, in the transmitted stream to facilitate the switching and the receiver synchronisation. At 500 MBauds, one dummy frame lasts 6.7 µs (with a 90 symbols PLHEADER). Assuming that the ferrite switches commutation time is less than 2 µs [13], [14] and that the system uncertainty is roughly 0.5 µs, a synchronisation field can be included in a dummy frame to help the receiver. Thus a single dummy frame deals with the switching and the receiver synchronisation. As depicted in Fig. 2, classical dummy frames may still be used for padding as in classical DVB-S2/S2X systems (as suggested in [10, Section 5.5.1]) and the BHSync frame is necessary only when a commutation occurs. If several consecutive time slots are allocated to the same beam, only one BHSync is relevant. The main contribution of our work, presented in Section II, is the design of a synchronisation field in the dummy frame and the corresponding receiver architecture. This enables the receiver to synchronise with a unique dummy frame even in a beam hopping scenario with poor channel condition (down to -

10 dB). This constitutes an alternative to the super-frame introduced in [10, Annex E]. Moreover, very few modifications are to be done at transmitter side as our proposal is fully compliant with DVB-S2/S2X. In Section III, we evaluate the performance of our scheme through simulations. We conclude this paper by concluding remarks and perspectives in Section IV. II. P ROPOSED SCHEME BHSync frame. Fig. 3a presents the BHSync frame structure. The frame is divided in three parts. The first part is the PLHEADER, as defined in the DVB-S2/S2X standard [16], with 90 or 180 symbols. Then there are Nc symbols that will manage the switching duration and the system uncertainty as discussed earlier. Finally, a synchronisation field of 924 symbols is inserted at the end of the frame to help the receiver after a period without signal. We depict the structure of the synchronisation field in Fig. 3b. This field has 33 blocks of 28 symbols each. We use a spreading factor of 28 to obtain a processing gain of 14.5 dB. Among the 33 blocks, 17 blocks (Pi with 0 6 i 6 16) serve to estimate the carrier frequency offset. The remaining 16 blocks (CWi with 0 6 i 6 15) are encoded with a ReedMuller code of parameter [4, 16] where the least significant bit 90 or 180 symb.

Nc symb.

924 symb.

PLH*

Dummy

Sync. field

switching area (a) BHSync frame

28 P0

CW0

P1

CW1

P15 CW15

P16

56 (b) Structure of the synchronisation field

Fig. 3: BHSync frame and synchronisation field

Detection confirmation Frequency offset estimation

Decoded CW

MAX

MAX

FFT (128)

FWHT (16)

Zero padding

Step 2

Step 3

Frequency offset correction

P

0

CW

0

P

1

DL1 (33)

CW

1

CW

15

Receiver

P

16

state machine 28xOSF 28xOSF 28xOSF 28xOSF SS0

SS1

SS2

DL0 (32x28xOSF)

SS3

28xOSF SS3x

Samples Mean power: P

2

16 multipliers

Adder (16) ABS

Step 1

Score Primary detection alarm

Fig. 4: Receiver architecture to detect and decode the BHSync frame

is forced to zero to ensure that the XOR of two consecutive bits is 1. These blocks carry 3 information bits. π We assume that the pilot symbols are equal to ei 4 and the iπ coded symbols are equal to ±e 4 . We apply a scrambling sequence SSi (0 6 i 6 32) to the entire synchronisation field. Receiver architecture. The synchronisation after a period without signal concerns both timing and frequency offsets and relies a unique word detection. Even more, clock and frequency reference drifts in the user terminal must be considered for a cold start. The receiver must work with an oversampling factor (OSF) with respect to the channel symbol rate Rs . In practice, OSF is greater or equal to 2 (we use 2 in the simulations). The general receiver architecture is depicted in Fig. 4. The detection and estimation of the various parameters work in three steps. In Step 1, the objective is to detect the BHSync frame. To that end, the samples are stored in a delay line (DL0 in Fig. 4) and we compute the correlations with the sequences SSi (0 6 i 6 31). The correlations are stored in the delay line 1. At the same time, we add the correlation products for a differential detection. This operation requires 16 complex multipliers on 28 symbols. This choice has been driven to reuse the VL-SNR correlator (16 × 56) hardware. If the score

at the bottom of Fig. 4 is greater than a threshold, a primary alarm is declared and we go into Step 2. In Step 2, the values in DL1 does not change (the switches between DL0 and DL1 are openend). A last correlation is done to obtain P16 . We can see in Fig. 4 that the receiver state machine controls the last switch). From these 33 correlations, we use the correlations corresponding to the Pi (0 6 i 6 16) to obtain a coarse frequency offset estimation. This requires a FFT of size 128 where zero padding has been used. The FFT also confirms the synchronisation (in order to decrease the false alarm probability). The receiver is now aware of the frequency offset and can enter Step 3. In Step 3, we start by correcting the correlation results CWi (0 6 i 6 15). The corrected values are sent to a Fast Walsh Hadamard Transform (FWHT) to decode the three information bits carried in the BHSync frame. The maximum value of the FWHT is also helpful to confirm the synchronisation and exclude false alarms. The proposed architecture has a low complexity in order to facilitate its hardware implementation. Indeed, the complexity drove the synchronisation field design. We will now study its performance.

10−1

CRB (L=924) CRB (L=476) Timing offset = T/4 Timing offset = T/8

10−4

Timing offset = 0

10−5

10−6 −15

−10

−5

0

5

CRB (L=896, roll−off=0.2) Timing offset = T/4

Standard deviation ∆T/T

Standard deviation CFO

10−3

Timing offset = T/8 Timing offset = 0

10−2 −14

10

−12

−10

SNR [dB]

(a) Carrier frequency offset estimation accuracy 100 Probability (FA threshold=T/4)

Timing offset = T/4

SNR estimation [dB]

Timing offset = T/8 Timing offset = 0

0 −5 −10 −15 −15

−10

−4

−5

0

−2

0

(b) Timing offset estimation accuracy

10 5

−8 −6 SNR [dB]

5

SNR [dB]

Missed detection

10−2 10−3 10−4 10−5 10−6 −13

10

False alarm

10−1

−12

−11

−10

SNR [dB]

(c) SNR estimation accuracy

(d) False alarm and missed detection probabilities

Fig. 5: Synchronisation performance of the proposed scheme (T is the symbol duration)

III. S YNCHRONISATION PERFORMANCE Frequency synchronisation. The first result evaluates the frequency estimation accuracy. To that end, we compare the standard deviation of the normalized carrier frequency offset (defined as ∆f × T where ∆f is the carrier frequency offset and T is the symbol period) with the Cramer-Rao bound for a unique word of length L symbols [15] given by E[(fˆ − f )2 ] > CRB(f ) =

3 1 1 × × Es , 2 2 2π L(L − 1) N

(1)

0

where E is the expected value and fˆ is the frequency estimation. Our simulator work with two samples per symbols, thus a processing gain of 2 is necessary to compute the standard deviation. Moreover, we assume that the normalized CFO is lower than 0.003. Fig. 5a presents the results for various timing offsets. We remind that the frequency estimation uses 17 fields of 28 symbols each (P X with 0 6 X 6 16 in Fig. 3b), resulting in 476 symbols. However, we also depict the Cramer-Rao bound for a unique word of 924 symbols (the entire length of the synchronisation field). The results point out that the detection is almost optimal when the receiver is synchronised in time. Thus we verify that the FFT is an optimal estimator. When the timing offset increases, the frequency error also increases and reaches its

maximum at T /4, where T is the symbol duration. For very low SNR, typically below -11 dB, the results have a bias as the missed detections are not taken into account (and the missed detection probability becomes significant). Timing synchronisation. As for the frequency estimation, we compute the standard deviation of the timing estimation error and compare it to the Cramer-Rao bound [15] given by E[(ˆ τ − τ )2 ] > CRB(τ ) =

−1 1 1 × × Es , 2gα00 (0) L N

(2)

0

where τˆ is the timing estimation and gα is the raised cosine function with roll-off α (α = 0.2 in the simulations). Fig. 5b presents the results. The estimation is excellent when the timing offset is low. We obtain a loss of 2.5 dB compared to the Cramer-Rao bound when the timing offset is equal to T /4 (worst case). SNR estimation. Fig. 5c shows the SNR estimation accuracy. The results are excellent when the timing offset is low. Note that the estimation variance increases when the SNR decreases, even for small timing offsets. Finally, we see that the SNR estimation degrades significantly at large SNR if the timing offset is large. The estimation error may impact the phase tracking loop (if the loop parameters depends on the SNR) and the LLR computation. If the receiver performance is degraded, a second SNR

estimator may be installed after the timing synchronisation. This is part of future work. False alarm/Missed detection. Finally, we present the false alarm and missed detection probabilities in Fig. 5d. We compute the different detection thresholds empirically. A false alarm is returned when the estimated sampling instant is greater than T /4 which corresponds to a worst case. The results exhibit a false alarm and a missed detection probabilities below 10−4 at -10 dB. The performance is similar to the results presented in [9]. Setting different detection thesholds will change the false alarm and missed detection rates. However, the important performance is the frame error rate at the receiver. By reaching a false alarm and missed detection rates of about 10−5 at -9.8 dB, we comply with the DVB-S2X objective in terms of frame error rate (the targeted performance is 10−5 ). IV. C ONCLUSION In this work, we introduce a dedicated synchronisation field in dummy frames for helping the receiver synchronisation in a beam hopping systems. By carefully designing the receiver algorithms, we demonstrate the effectiveness of our scheme through simulations. The main advantage is that a unique dummy frame enables to deal with the switching and the receiver synchronisation, while being compatible with the DVB-S2/S2X standard. Future work will extend the current results in three directions. First, we will complete the performance evaluation by obtaining frame error rate performance for various MODCODS from very-low to large SNR. Then we are currently investigating the impact of beam interference, in case of frequency reuse, with our solution. We rely on different scrambling sequences in neighboring beams to fight againt these interferences. Finally, we plan to test our proposal on a testbed. R EFERENCES [1] P. Angeletti, D. Prim, and R. Rinaldo, “Beam hopping in multibeam broadband satellite systems: System performance and payload architecture analysis,” in AIAA International Communications Satellite Systems Conference, 2006. [2] T. Pecorella, R. Fantacci, C. Lasagni, L. Rosati, and P. Todorova, “Study and implementation of switching and beam-hopping techniques in satellites with on board processing,” in International Workshop on Satellite and Space Communications, 2007. [3] J. Anzalchi, A. Couchman, C. Topping, P. Gabellini, G. Galinaro, L. D’Agristina, P. Angeletti, N. Alagha, and A. Vernucci, “Beam hopping in multi-beam broadband satellite systems,” in AIAA International Communications Satellite Systems Conference, 2009. [4] J. Anzalchi, A. Couchman, P. Gabellini, G. Gallinaro, L. D’Agristina, N. Alagha, and P. Angeletti, “Beam hopping in multi-beam broadband satellite systems: System simulation and performance comparison with non-hopped systems,” in Advanced Satellite Multimedia Systems Conference, 2010. [5] J. Lei, “Multi-beam satellite resource allocation optimisation for beam hopping transmission,” Ph.D. dissertation, Universitat Autnoma de Barcelona, 2011. [6] R. Alegre, N. Alagha, and M. V´azquez-Castro, “Heuristic algorithms for flexible resource allocation in beam hopping multi-beam satellite systems,” in AIAA International Communications Satellite Systems Conference, 2011.

[7] J. Lei and M. V´azquez-Castro, “Multibeam satellite frequency/time duality study and capacity optimization,” Journal of Communications and Networks, vol. 13, no. 5, pp. 472–480, 2011. [8] C. Rohde, H. Stadali, J. Perez-Trufero, S. Watts, N. Alagha, and R. De Gaudenzi, “Implementation of DVB-S2X Super-Frame Format 4 for Wideband Transmission,” in International Conference on Wireless and Satellite Systems, 2015. [9] C. Rohde, N. Alagha, R. De Gaudenzi, H. Stadali, and G. Mocker, “Super-framing: a powerful physical layer frame structure for next generation satellite broadband systems,” International Journal of Satellite Communications and Networking, vol. 34, no. 3, 2016. [10] ETSI EN 302 307-2, “Digital Video Broadcasting (DVB); Second generation framing structure, channel coding and modulation systems for Broadcasting, Interactive Services, News Gathering and other broadband satellite applications; Part 2: DVB-S2 Extensions (DVB-S2X),” 2014. [11] G. Colavolpe, A. Modenini, A. Piemontese, and A. Ugolini, “Multiuser detection in multibeam satellite systems: Theoretical analysis and practical schemes,” in IEEE International Workshop on Signal Processing Advances in Wireless Communications, 2015. [12] H. M´eric and G. Lesthievent, “On the use of dummy frames for receiver synchronisation in a DVB-S2(X) beam hopping system,” in AIAA International Communications Satellite Systems Conference, 2017. [13] T. MESL, “Ka-band waveguide ferrite switches and switch matrices,” http://www.microwavejournal.com/articles/1341-ka-band-waveguideferrite-switches-and-switch-matrices. [14] R. Billings and T. Edridge, “Ferrite circulator switches and their applications,” Microwave Journal, vol. 46, no. 11, pp. 124–126, 2003. [15] A. D’Andrea, U. Mengali, and R. Reggiannini, “The modified CramerRao bound and its application to synchronization problems,” IEEE Transactions on Communications, vol. 42, no. 234, pp. 1391–1399, 1994. [16] ETSI EN 302 307-1, “Digital Video Broadcasting (DVB); Second generation framing structure, channel coding and modulation systems for Broadcasting, Interactive Services, News Gathering and other broadband satellite applications; Part 1: DVB-S2,” 2014. [17] F. Sun, Y. Jiang, and L. Lee, “Frame synchronization and pilot structure for second generation DVB via satellites,” International Journal of Satellite Communications and Networking, vol. 22, no. 3, pp. 319–339, 2004. [18] G. Colavolpe, A. Barbieri, and G. Caire, “Algorithms for iterative decoding in the presence of strong phase noise,” IEEE Journal on Selected Areas in Communications, vol. 23, no. 9, 2005. [19] D. Howe and T. Tasset, “Clock jitter based on PM noise measurements,” in IEEE International Frequency Control Symposium and PDA exhibition, 2003. [20] N. Semiconductors, “AN11144 - Universal single LNB with TFF101x FIMOD IC,” 2012. [21] M. Miller, “Flexible capacity satellite communications system with dynamic distribution and coverage areas,” Patent US 8218476 B2, 2012.