NL2018496B1 - Flexible modulation in pon networks - Google Patents

Abstract

An optical line terminal for use in an optical communication system is provided. The optical line terminal comprises a light source for generating optical signals through optical fibers of the system and a processor arranged to modulate the optical signals of the light source using a modulation format and a clock rate, and to receive information about the quality of one or more channels between the optical line terminal and one or more end users, and depending on the received quality for a specific channel, adjust the modulation of the optical signals for that specific channel by way of changing the clock rate.

Description

FIELD OF THE INVENTION

The invention relates to optical data communication, and specifically in the context of high-speed access networks such as passive optical networks.

BACKGROUND ART

Higher capacity passive optical networks (PONs) continue to be a topic of interest with the upcoming standardizations of 100G PONs by IEEE and ITU-T. Various modulation formats have been proposed to increase the peak data rate of the access network. Among others electrical duobinary (DB-OOK), pulse amplitude modulation (PAM), discrete multi-tone (DMT), and duobinary-PAM4 (DB-PAM4) formats have been discussed [1,2], In PONs in particular, the power budget is important, and puts stringent requirements on the sensitivity degradation of the receiver due to the modulation format. Therefore, tradeoffs have to be made between capacity, reach, power budget, and complexity of implementation. Of the above-mentioned modulation formats, DMT requires the most complex implementation due to sophisticated signal processing. Simpler DB-PAM4 offers high capacity, but is limited in both reach and sensitivity [3, 4],

In publication [6] it is suggested to measure the signal quality at the receiver side, and depending on that quality change the order of the modulation format at the optical line terminal (OLT) side between 2, 4 or 8 levels (OOK/PAM4/PAM8). Depending on the measured signal quality the number of output levels can be changed so as to exploit unused budget.

However, changing the number of output levels has some limitations. For the highest modulation orders, there will be diminishing returns: the increase in bitrate becomes smaller, while the signal quality requirements keep rising. Second, the steps that can be made are quite coarse (i.e. it’s not possible to have fractional levels).

So there is still a need to find different methods to better fit the transmitted signal to the quality as experienced by an individual user.

SUMMARY OF THE INVENTION

One of the objects of the invention is to provide an optical line terminal which is able to better fit the transmitted signal to the quality as experienced by an individual user of a passive optical network connected to the line terminal.

A first aspect of the invention provides an optical line terminal for use in an optical communication system, the optical line terminal comprising a light source for generating optical signals through optical fibers of the system, and a processor arranged to modulate the optical signals of the light source using a modulation format and a clock rate, and to receive information about the quality of one or more channels between the optical line terminal and one or more end users, and depending on the received quality for a specific channel, adjust the modulation of the optical signals for that specific channel by way of changing the clock rate.

The optical line terminal (OLT) is a transmitting and receiving device that can be used in an optical access networks, more specifically in a passive optical network (PON). A PON network provides communication between one central point (OLT) and multiple users (Optical Network Units - ONUs). The OLT and ONUs can be connected through a fiber and optical splitter. Communication to different users can be divided in different time slots (Time Division Multiplexing). The method of transmission might be changed every timeslot. The quality of the channel between the OLT and a certain ONU might be different to the quality of the channel between the OLT and another ONU. The quality of the different channels can be quantified by e.g. received optical power (ROP), signal to noise ratio (SNR) and/or bit error rate (BER). By using the quality of the different channels, an optimal modulation format can be selected for every channel. For example, if the quality of a certain channel is low, the OLT may use a non-return-to-zero on-off-keying (NZR-OOK) modulation format at a low clock rate. If the quality of the channel is high, the OLT may use the same modulation format at a higher clock rate. Or, the OLT may use a multilevel pulse amplitude modulation format in combination with a higher clock rate.

Optionally, the clock rate is increased to such an extent that, while the bandwidth of the system is kept constant, the type of transmission changes from a binary to duobinary type of modulation. Changing from a binary type of modulation to a duobinary type of modulation enables communication at a higher data rate in the same available bandwidth, hence increasing the spectral efficiency.

Optionally, the processor is arranged to receive information on supported modulation formats for all the end users and depending on the supported modulation formats vary the length of a timeslot for a user, so that the length of the timeslot for a user with a higher supported modulation format can be shortened to increase the available capacity of the other users that do not support higher modulation formats and faster clock rates. So, if the OLT takes the supported method of transmission with a particular user into account, the total available network capacity can be shared with increased fairness_across all the users.

Optionally, the optical line terminal comprises an equalizer arranged to change the bandwidth of a channel, and therefore can change the transition point where the clock rate leads to a binary or duobinary type of modulation. This may be advantageous since both binary and duobinary type of modulation have a range of desired clock rate to bandwidth ratios. Having control over the bandwidth of the channel (for example increase or decrease of the bandwidth) can enable multiple clock rates to operate in the desired bandwidth ratio. Additionally, different users can experience a different bandwidth, the equalizer can change the bandwidth for each individual user. The equalizer can be, but is not limited to, a feed forward equalizer.

Optionally, the modulation format is one of the list including, but not limited to, PAM-M and the DB-PAM-M with M>2. This includes PAM2/3/4/8 and DB-PAM2/3/4.

Optionally, the optical line terminal comprises a number of light sources each operating on a different wavelength, wherein the processor is arranged to use the information from the quality of the channel of the various users to allocate them to a specific wavelength. In this way the processor can assign the users to a specific wavelength. Therefore the users can be optimally distributed across the wavelengths. The processor can use communication on each wavelength that uses the same set of modulation formats and clock rates for each wavelength. Alternatively, the processor can use communication on each wavelength that uses a different modulation formats and clock rates for different wavelengths.

Optionally, the processor is arranged to change the number of levels per transmitted symbol. This is advantageous because the processor will not only be able to switch between clock rates but additionally change the order of the modulation format, so for example from PAM2 with a clock rate of 10 GBaud to PAM3 with a different clock rate, such as 12 Gbaud or with the same clock rate.

Optionally, the processor is arranged to allocate a certain group of users to decode certain bits out of the multilevel signal. In this way the user only needs to decode part of the bits. When a user only needs to decode a part of the bits of the multilevel signal, the bit rate at the user can be lower, optionally simplifying the receiver hardware.

Optionally, the processor is arranged to change the relative position of levels of a multilevel symbol and create unequal relative position of levels of the multilevel symbol. This may be advantageous because this makes decoding of certain bits in the multilevel signal easier than decoding certain other bits in the multilevel signal.

Optionally, the processor is arranged to allocate users with a lower quality of the channel to those bits in the multilevel symbol that require a lower quality of the channel, and allocate users that have a higher quality channel to those bits that require a higher quality of the channel.

Due to the unequal relative position of the levels of a multilevel symbol, successful decoding of certain bits encoded in this multilevel symbol can be possible with less quality of the channel (it’s easier) than the quality of the channel that is required for certain other bits encoded in this multilevel symbol (it’s harder). By allocating the users according to their channel quality, more users can use multilevel signaling, thereby increasing the total capacity of the network.

Optionally, the optical line terminal comprises multiple OOK modules, and the processor is arranged to generate multilevel signals by changing the output of the OOK modules.

Optionally, the optical line terminal comprises a number of variable attenuators, each of which being arranged after one of the OOK modules.

In this way, the output of the terminal can easily be changed by changing the attenuation. The typical way to generate multiple modulation formats is by using a digital to analog convertor. A transmitter structure with a number of variable attenuators can simplify the transmitter structure, and therefore can reduce the cost.

Optionally, the output is changed by changing a variable current source in the OOK modules.

A second aspect of the invention provides an optical communication system comprising an optical line terminal as described above, and a number of optical network units connected to the optical line terminal via a passive optical network.

A third aspect of the invention provides an optical network unit for use in an optical communication system, the optical network unit comprising a light source for generating optical signals through optical fibers of the system, and a processor arranged to modulate the optical signals of the light source using a modulation format and a clock rate, and to receive information on the modulation type to be used on a channel between the optical network unit and an optical line terminal of the communication system, and depending on the received information, adjust the modulation of the optical signals by way of changing the clock rate.

Optionally, the optical network unit receives information on the modulation type that is used when the optical line terminal sends signals to the optical network unit.

Optionally, the optical network unit has an optical light source for generating optical signals that is tunable to a specific wavelength, and wherein the receiver is tunable to a specific wavelength.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects of the invention are apparent from and will be elucidated with reference to the embodiments described hereinafter. In the drawings,

Figure 1 shows a graph of the relative increase in bitrate as a function of PAM order;

Figure 2 shows a graph of the relative increase in bitrate as a function of PAM order and as a function of relative symbol rate;

Figure 3 schematically shows a passive optical network (PON) 1 according to an embodiment of the invention;

Figure 4A schematically shows an embodiment of the OLT;

Figure 4B schematically shows the processor according to an embodiment;

Figure 5A schematically shows an embodiment of an ONU;

Figure 5B schematically shows the processor in the ONU according to an embodiment;

Figure 6 schematic shows four representations of the shift of the required Received optical power (ROP) for a certain modulation level;

Figure 7 shows schematic representations of the shift of the required Received optical power for a certain clock rate and modulation level;

Figure 8 shows an example of an eye diagram for Standard PAM4;

Figure 9 shows an example of an eye diagram for Compressed PAM4 wherein the inner two levels are moved closer to the outer two levels;

Figure 10 shows eye diagrams for the reception of multiple bits (MSB and LSB) by two user groups with a PAM4 input signal;

Figure 11 shows two graphs of histograms of Standard PAM4 with high and low channel quality;

Figure 12 shows two graphs of histograms of Compressed PAM4 with high and low channel quality;

Figure 13 shows a scheme of an experimental setup, together with eye diagrams as generated by up sampling of filtered data after the receiver in offline processing;

Figures 14A, 14 B and 14C show graphs of BER measurements as a function of received power at the photodiode (PD) shown in Figure 13;

Figure 15 schematically shows a signal generator of an OLT according to an embodiment;

Figure 16 shows an alternative embodiment of the signal generator;

Figure 17 shows a further alternative of the signal generator;

Figure 18 shows yet a further alternative of the signal generator;

Figure 19 is a scheme to explain the principle of a Directly Modulated Laser;

Figure 20 a scheme to explain the principle of an Electro-Absorption Modulator, and

Figure 21 a scheme to explain the principle of a Mach-Zehnder Modulator.

It should be noted that items which have the same reference numbers in different Figures, have the same structural features and the same functions, or are the same signals. Where the function and/or structure of such an item has been explained, there is no necessity for repeated explanation thereof in the detailed description.

DETAILED DESCRIPTION OF EMBODIMENTS

Fiber provides different properties for the communication channel than wireless or fixed copper networks. In contrast to these networks, crosstalk is much less of an issue, and the channel is much more static. Therefore, adaptive techniques are much less required to achieve an acceptable performance. This might be one of the reasons why historically adaptive modulation techniques have never been implemented in fiber based systems. The lack of adaptive techniques in fiber systems has a negative impact on the efficiency of these systems. As one single method of communication has to be used for all possible situations, this method cannot be tailored to the exact needs of the specific user and channel. This is especially true for PON fiber systems. In these systems multiple ONU transceivers are connected to a single transceiver on the central office (CO) side. Because every user needs to be guaranteed of a working system, the worst case scenario effectively determines the properties of the system. This means, that in one way or the other, resources are wasted i.e., either too much power is used to transmit the data, the data rate is lower than it could be, too complex signal processing is put to use, or in general a transceiver with too high performance, and thus being more expensive, is used. The broader the range of channel properties inside a network, the larger the waste.

To better use the available resources in a PON access networks, an optical line terminal (OLT) is provided to use a flexible modulation technique. Prior to transmission, for each individual user, the channel quality can be measured in the upand downstream direction. Channel quality can be expressed for example in received optical power, signal to noise ratio, or ultimately the bit error ratio. Depending on the actual channel quality between the OLT and a certain user, subsequent data transmission is adapted to be tailored to this specific channel.

Introduction of this flexible modulation scheme in PONs allows each ONU to communicate at the highest data rate possible in a specific situation. Taking advantage of TDM timeslot redistribution, it also enables those users with a low received optical power, to take advantage of this technique.

Probing of the channel and selection of the transmission method can be automated to provide ease of use for the operator. If over time the channel quality changes, the transmission method can be changed automatically. Depending on the channel quality the order of the modulation format may be changed, for example between 2, 3, 4 or 8 levels (OOK/PAM3/PAM4/PAM8). So the number of levels per symbol may be changed. Higher number of levels mean more bits (information) that can be transmitted per symbol, at the cost of a higher required signal quality, e.g. smaller amounts of noise can be tolerated (as the signal levels are closer to each other).

It is noted that changing the number of output levels has limitations. For the highest modulation orders, we get diminishing returns: the increase in bitrate becomes smaller, while the signal quality requirements keep rising. Second, the steps that can be made are quite coarse (i.e. it’s not possible to have fractional levels).

To better fit the signal to the exact quality as experienced by an individual user, a method is proposed wherein also the clock frequency of the signal can be changed depending on the channel quality. Current optical access networks only use a single clock frequency for all their users in the same network.

Figure 1 shows a graph of the relative increase in bitrate as a function of PAM order. As can be seen from the curved line 11 in Figure 1, an Increasing of the PAM-order has diminishing returns in terms of increasing the bit rate.

Figure 2 shows a graph of the relative increase in bitrate as a function of PAM order and as a function of relative symbol rate. As can be seen from the curved graph of Figure 2, increasing both clock rate and modulation order enables finer control over the increase in bitrate. So by taking advantage of a different dimension, more freedom in tuning the bitrate and incurred penalties is possible.

Figure 3 schematically shows a passive optical network (PON) 1 according to an embodiment of the invention. The PON 1 comprises an OLT 2, a number of ONUs 3 and an optical splitter 4.

Figure 4A schematically shows an embodiment of the OLT 2. The OLT 2 comprises a processor 21, see CPU 21, a light source 22, a Mach-Zehnder Modulator (MZM) 23, and receiver 26. The light source 22 may be a laser producing light having a specific wavelength. The processor 21 is arranged to receive quality information 10 on channels between the OLT and each of the ONUs 3, see Figure 3. The quality information 10 can be received through the receiver 26. The processor 21 is also arranged to generate modulated signals for modulating the MZM 23 depending on the received quality information. As a result, the OLT produces modulated light signal 15, which are forwarded via optical fibers of the PON 1. Other embodiments for modulating the light signals are possible as will be explained below. The processor 21 is also arranged to generate control signals for the receiver 26.

It is noted that the communication in the PON 1 is bidirectional, thus it will be clear to the skilled reader that the OLT 2 does not only comprise a transmitter, but also a receiver, and that the receiver will receive also data apart from the quality information 10 (not further detailed in Figure 4A). Similarly, the ONUs 3 won’t only have a receiver but also a transmitter.

Figure 4A shows an embodiment of the OLT 2. The depicted setup shows a MZM 23. Although this is one possible setup that can be used, alternatives exist. Alternatively, the modulated light signals are generated by the use of a directly modulated laser instead of an MZM. An example of such a directly modulated laser is shown in Figure 19.

Figure 4B schematically shows the processor 21 according to an embodiment. The processor 21 comprises a clock controller 51, a modulation controller 52 and a signal generator 53. The signal generator 53 receives input from both the clock controller 51 and the modulation controller 52. The clock controller 51 and the modulation controller 52 receive quality information 10. Depending on the quality information, the clock controller 51 and the modulation controller 52 produce output to the signal generator 53. Quality information 10 can be received through the same optical fibers of PON 1. One option for the clock controller 51 and the modulation controller 52 is to command the signal generator 53 to use a modulation format and clock rate as high as possible without crossing the forward error correction bit error rate limit. Another option for the clock controller 51 and the modulation controller 52 is to command the signal generator 53 to use a lower clock rate or modulation format to reduce the required applied forward error correction. The clock controller 51 and the modulation controller 52 can favor the increase of modulation order over increase of clock rate or vice versa. The clock controller 51 and modulation controller 52 can control the receiver 26 through communication path 27.

Optionally, the OLT 2 comprises an equalizer 54 arranged to change the bandwidth of one or more of the channels.

Figure 5A schematically shows an embodiment of an ONU 3. The ONU in this example comprise a processor 31, a light source 32, and a Mach-Zehnder Modulator (MZM) 33.

The ONU 3 is arranged to support multilevel modulation with multiple clock rates, and the option to change these. The decision on when to adjust the modulation format and which modulation format is to be used by which ONU 3 is made by the OLT 2, see Figure 4A. The decision is communicated to the ONUs 3 through information 40. Information 40 can be included in the received data stream 18. The processor 31 in ONU 3 does not have to make a decision on which modulation format and clock rate to be used. The ONUs 3 will follow the directions of the OLT 2.

Figure 5B schematically shows the processor 31 according to an embodiment. The processor 31 comprises a clock controller 41, a modulation controller 42 and a signal generator 43. The signal generator 43 receives input from both the clock controller 41 and the modulation controller 42. The clock controller 41 and the modulation controller 42 receive decision information 40 from the OLT 2. Depending on the decision information, the clock controller 41 and the modulation controller 42 produce output to the signal generator 43. The clock controller 41 and modulation controller 42 can control the receiver 36 through communication path 37.

It is noted that the quality of the channel in the two directions (OLT to ONU and ONU to OLT) could be different. In such situations, the chosen modulation type can be different in the two directions. In an embodiment, the OLT 2 is arranged to decide on which modulation type to use in both directions. The ONU 3 reports to the OLT 2 and follows its decisions.

In an embodiment, the proposed flexible modulation scheme uses the following modulation formats: non-return-to-zero on-off-keying (NRZ-OOK), 3 level pulse amplitude modulation (PAM3), 4 level pulse amplitude modulation (PAM4),

Standard electrical duobinary based on NRZ-OOK (DB), duobinary based on PAM3 (DB-PAM3), and duobinary based on PAM4 (DB-PAM4).

PAM (pulse amplitude modulation) based modulation formats are based on multilevel transmission, thereby increasing the number of bits per symbol, while keeping the clock rate the same. Duobinary based modulation formats are based on shaping the signal based on bandwidth limitations. Standard electrical duobinary is transmitted as normal 2 level NRZ-OOK. Due to appropriate combined bandwidth limitations of the transmitter (i.e. OLT 2), channel, and receiver (i.e. ONU 3) combined this signal can be received as a 3 level signal. With appropriate, yet simple, pre- and decoding, the 3 level signal can be decoded back to the original transmitted signal. The bits per symbol stay identical, but the clock rate of the symbol increases. If PAM3 is combined with duobinary, DB-PAM3 is transmitted with 3 levels, and received with 5 levels. If PAM4 is combined with duobinary, DB-PAM4 is transmitted with 4 levels, and received with 7 levels. As an example, the relative data rates of the various modulation formats are, based on a 10 Gbps system can be:

OOK: 10 Gbps

PAM3: 15 Gbps PAM4: 20 Gbps DB: 25 Gbps DB-PAM3: 37.5 Gbps DB-PAM4: 50 Gbps It could be an option to reduce the symbol rate of the duobinary based modulation formats from 25 GBaud to 20 GBaud, to reduce the requirements on the bandwidth of the system. Additionally this has the advantage that the clock rates existing in the system are multiples of each other.

In prior art networks the total available pool of capacity is shared among the users by allocating timeslots (in which data is transmitted to a single user) of different duration to the various users, depending on the users capacity demand. If the capacity demands exceed the total capacity limit it is possible to assign certain users to a different optical wavelength. This brings a different configuration of the network as more optics are involved which can be problematic in cost for the network owners.

According to an embodiment, the OLT 1 is arranged to only use a single wavelength, but use optimization techniques to get more capacity from this single wavelength. Optimization is achieved by allocating different clock rates for different users in addition to using flexible modulation formats.

The basis system works using only one optical wavelength. The system measures the quality of the channel for all the users and picks the optimal modulation level and clock rate. A select number of clock rates may be used by the OLT 2, and preferably at integer multiplies of each other. For example, 10 Gsymbols/s, and 20 Gsymbols/s. This is due to the practical way that the receivers work as will be appreciated by the skilled person.

Alternatively to operating on a single wavelength, it is possible to further improve the performance by using multiple wavelengths. In a first embodiment this allows a straightforward increase in capacity, proportionally to the increase in wavelengths. In another embodiment this enables a more efficient use of clock rates, and thus a more than proportional increase in capacity. An example of such a system is where 4 transmitters at the central location operate at 4 distinct wavelength channels. At the users location a tunable transceiver tunes to one of these 4 wavelength channels. By this the system becomes a stack of 4 communication channels.

In an embodiment, the OLT 2 is arranged to use identical clocks for all wavelengths. For example, the used modulation format and clock rate can be applied independent to all these 4 independent channels. This will work just as in the single wavelength case. For example, starting from a system with a base rate of 10 Gb/s. If in the single wavelength case the proposed flexible modulation scheme would increase the capacity of the network with a factor X, then the capacity of the system with 4 independent channels would be increased by a factor 10*4*X. This is graphically shown in Figure 6.

Figure 6 schematic shows four representations of the shift of the required Received optical power (ROP) for a certain modulation level based on an example network. Low ROP typically means low SNR. λ-ι, λ₂, λ₃, denote the wavelength. Example limits of ROP are (line 61: 2 levels, 1 bit/symbol); (line 62: 4 levels, 2 bits/symbol); (line 63: 8 levels, 3 bits/symbol). A user with a certain ROP can use the modulation format if it is on the right-hand side of the line. Thus each wavelength uses the same clock rates and modulation levels.

Additionally, in another embodiment more benefits can be obtained if the chosen clock rates are chosen intelligently over the wavelengths. As explained before, each combination of modulation level and clock rate requires a certain channel quality to operate. In principle this clock rate can be tuned continuously, in practice it might be beneficial for the complexity of the receiver to limit the clock rates to a certain set. This limitation only applies to all communication on a certain wavelength. Thus, a different set of clock rates can be selected on each wavelength. This allows even more optimal selection of modulation levels and clock rates on the available channel quality distribution of the users. This is expressed graphically in Figure 7 showing schematic representations of the shift of the required Received optical power for a certain clock rate and modulation level. Low received optical power typically means low SNR. λ denotes the wavelength. Example limits of SNR are 2 levels (line 71), 4 levels (line 72), 8 levels (line 73). Going from one number of levels to another has a certain step size, whereas the clock rate can be adjusted continuously.

For example, a user receiving -19 dBm power would at wavelength 1 be put in the group belonging to line 72, having a data rate of 20 Gbit/s (10GBaud * 2 bits/symbol). If instead this user is moved to wavelength 3, he will have a higher data rate of 24 Gbit/s (12 GBaud * 2bit/symbol). In a single wavelength scenario it wouldn’t have been possible to operate the entire network at the clock rate of 12 GBaud, as some users won’t have a working connection (all the users to the left of the line 71). In this multiple wavelength scenario those users not having a working connection at 12 GBaud can be dealt with at wavelength 1.

Explained in a different way: the straight lines 71, 72, 73 in the Figure 7 represent a staircase. If a user has enough channel quality to reach the next step, its supported data rate increases. By using multiple clock rates over multiple wavelengths, the locations of the steps shift, thereby effectively allowing us to use a smaller step size. Therefore more users are able to use a higher data rate.

In an embodiment, the processor 21 is arranged to change the relative position of levels of a multilevel symbol and create unequal relative position of levels of the multilevel symbol. In this way the processor 21 generates a modulation format that is referred to as Compressed PAM. In the embodiments described above the quality of the channel is measured and the modulation level and clock rate is adjusted depending on the channel quality, using for example the SNR. When using compressed PAM, additionally (or instead of) also the relative location of the PAM levels is changed. This will be explained with reference to Figure 8 and 9. Figure 8 shows an example of an eye diagram for Standard PAM4. Figure 9 shows an example of an eye diagram for Compressed PAM4 wherein the inner two levels are moved closer to the outer two levels, thereby increasing the distance between the inner two levels.

One of the purposes of optical access networks is to transmit as much data from the central location to the users, and back. Transmission through PAM transmits multiple bits per symbol. In prior art [8] it is shown that by using an appropriate receiver, multi-level multi-user interleaving is possible, that is it is not necessary that all the bits are intended for one user. For example, PAM4 has 4 levels per symbol, which represent 2 bits of information per symbol. It is possible to transmit one bit to user A, while user B selects the other bit in the same symbol to decode. For PAM4, one of these bits is called the Most Significant Bit (MSB), the other is called the Least Significant Bit (LSB). This principle is shown by Figure 10 which shows an eye diagram for the reception of multiple bits (MSB and LSB) by two user groups with a PAM4 input signal. The top drawing of Figure 10 shows an eye diagram for a first user group wherein MSB detection is done by using only a zero voltage threshold 301. The bottom drawing of Figure 10 shows an eye diagram for a second user group wherein LSB detection is done by using a lower threshold 300 and an upper threshold 302 and appropriate logic circuitry 303. Therefore, both users effectively decode the PAM4 signal to two NRZ signals at half the bit rate of the PAM4 signal.

It is noted that communication with 4 levels requires a better channel quality than communication through 2 levels, as the levels are closer to each other, therefore the decoding process is more prone to noise. For standard PAM4 this extra required channel quality is the same for the MSB and LSB user group. With Compressed PAM it is possible to take advantage of the varying channel quality between the users. By positioning the inner two levels closer to the outer two levels instead of equidistant, it becomes easier for the MSB group of users to decode the signal at the cost of a more difficult decoding for the LSB group.

As some users can have a better channel quality available than required for Standard PAM4 decoding, the OLT can allocate them to the LSB group. Thereby these users with a good channel quality help the users with a not good enough channel quality to be allocated to the MSB group. Without this solution these users with a not good enough channel quality would need to be communicated to with a two level modulation format only, thereby decreasing the transmission capacity of the entire system. Thus the use of asymmetrical compressed PAM allows more users to use multilevel PAM. Also users with a channel quality lower than required for Standard PAM can use Compressed PAM as long as there is another user that has a higher channel quality than required for Standard PAM. The OLT can inform the ONU about the relative levels of the Compressed PAM. The ONU receiver therefore can change its threshold levels.

Although explained here with PAM4 for simplicity, Compressed PAM also operates on higher orders of PAM.

Alternative graphical representations using a histogram of received symbols are shown in Figure 11 and 12. Figure 11 shows two graphs of histograms of received

Standard PAM4 symbols. The left graph shows the histogram of received values for a user experiencing a high channel quality. The right graph shows the histogram of received values for a user experiencing a low channel quality. Due to the larger noise contribution, the histogram for the low channel quality is broader than the histogram for the high quality channel. A user experiencing a high channel quality can decode both the MSB successfully with the MSB slicer 301, and the LSB with the combination of the lower LSB slicer 300 and upper LSB slicer 302. A user experiencing a low channel quality cannot accurately decode the MSB and cannot accurately decode the LSB because the noise contribution is too large.

Figure 12 shows two graphs of histograms of received Compressed PAM4 symbols. The left graph shows the histogram of received values for a user experiencing a high channel quality. The right graph shows the histogram of received values for a user experiencing a low channel quality. A user experiencing a high channel quality can decode both the MSB successfully with the MSB slicer 301, and the LSB with the combination of the lower LSB slicer 300 and upper LSB slicer 302. A user experiencing a low signal channel can successfully decode the MSB with the MSB slicer 301, the LSB cannot successfully be decoded by this user. Therefore, the processor 21 can assign the high channel quality user to decode the LSBs, whereas the low channel quality users decode the MSBs.

One of the modulation formats that could be used in the flexible modulation method as described above is duobinary-PAM3 (DB-PAM3). DB-PAM3 can offer a good tradeoff in this respect for certain applications. Duobinary-PAM3 is based on transmission of standard PAM3 signals at a higher symbol rate. Due to low-pass filtering in the channel and at the receiver, the 3 transmitted levels are received as 5 electrical levels after the receiver frontend. At an identical symbol rate it offers 1.5 times the capacity of normal duobinary, while having better sensitivity than DB-PAM4.

Furthermore, the increase in spectral efficiency enables a reduction in the symbol rate, thereby making DB-PAM3 more resilient to dispersion than standard duobinary at the same net data rate. Finally, the linearity requirements of DB-PAM3 for the transmitter are significantly reduced compared to PAM4 and other higher-order modulation formats, as only 3 levels of regular PAM3 are transmitted.

Figure 13 shows a scheme of an experimental setup, together with eye diagrams as generated by up sampling of filtered data after the receiver in offline processing. Figures 14A, 14 B and 14C show graphs of BER measurements as a function of received power at the photodiode (PD) shown in Figure 13. In Figure 14 results are shown for 20 GBaud DB-PAM3 and DB-PAM4, in Figure 14B results are shown for 25 GBaud DB-PAM3 and DB-PAM4, and Figure 14C shows results for 10 GBaud PAM4 and PAM8. Figures 14A, 14B and 14C also show 10 Gb/s OOK and 25 Gb/s DB-OOK.

Duobinary-PAM3 is part of the partial response family of modulation formats. Similar to standard duobinary, low-pass filtering a multi-level PAM format introduces a controlled amount of intersymbol interference (ISI). This allows transmission of a higher data rate through a smaller bandwidth, and forms additional signal levels in the received signal. As the amount of ISI added is known, this can easily be removed at the receiver side. Below, some possible embodiments are described relating to 37.5 Gb/s DB-PAM3 (25 GBaud) and 30 Gb/s DB-PAM3 (20 GBaud).

Similar to 25 Gb/s duobinary, DB-PAM3 requires approximately the same analog bandwidth as 10G NRZ receivers provide. For reference purposes, Figure 14A, 14B and 14C also include measurements with DB-PAM4 at 50 Gb/s (25 GBaud) and 40 Gb/s (20 GBaud), and PAM8 at 30 Gb/s (10 GBaud).

The experimental setup shown in Figure 13 was used for measurements of 10 Gb/s OOK, 25 Gb/s duobinary, 30 Gb/s PAM-8, 30 Gb/s and 37.5 Gb/s DB-PAM3, and finally 40 Gb/s and 50 Gb/s DB-PAM4. Data signals were generated by a 65 GSa/s arbitrary waveform generator (AWG) and drive a Mach-Zehnder modulator (MZM), thereby modulating the light from a 1550 nm CW DFB laser. The generation of the data varied by the modulation format used. For duobinary OOK, the original data was precoded to prevent error propagation. For PAM8, three binary sequences were combined to form the multilevel signal. In the case of DB-PAM3, a binary data stream was first mapped to the three-level PAM3 signal, and subsequently precoded in the same manner as normal duobinary. To keep complexity for mapping bits to the PAM3 symbols low, a simple 3B2T coding is used to convert the binary input to PAM3 symbols, resulting in an efficiency of 1.5 bit/symbol. At the expense of a longer and more complex coding, this could be increased up to the limit of 1.59 bit/symbol. The 4level signal required for transmission of DB-PAM4 is constructed by combining two binary sequences.

At the receiver side, the receiver consists of a 10 GHz PIN+TIA, subsequently low-pass filtered to a -3 dB bandwidth of 7.5 GHz during offline processing. The same bandwidth is used for all modulation formats examined. Depending on the received modulation format various steps are performed. Duobinary is taken modulo 2 to convert it back to binary data. PAM8 is sliced by 7 slicers and converted to three individual binary streams. The received 5(7) level DB-PAM3(DBPAM4) signal is first sliced by 4(6) slicers, taken modulo 3(4), and subsequently mapped back to a binary data stream.

Figure 14A shows the BER results for DB-PAM3 and DB-PAM4 at 20 GBaud, with data rates of 30 and 40 Gb/s respectively. Figure 14B shows the results of DB-PAM3 and DB-PAM4 at 25 GBaud, corresponding data rates are 37.5 and 50 Gb/s respectively. Figure 14C shows the BER for 20 Gb/s PAM4 and 30 Gb/s PAM8 for reference purposes. The figures 14A-14C show also 10 Gb/s OOK and 25 Gb/s duobinary. The measured sensitivity of 25 Gb/s duobinary at BER = 10-3 is -17.7 dBm, showing a penalty of 3.9 dB relative to 10 Gb/s OOK. 30 Gb/s DB-PAM3 has a sensitivity of -14.7 dBm, thus a penalty of 3.0 dB relative to 25 Gb/s duobinary. 37.5 Gb/s DB-PAM3 has a penalty of 3.8 dB relative to 25 Gb/s duobinary. For comparison BER results of DB-PAM4 are also shown. At data rates of 40 and 50 Gb/s, it offers higher capacity than DB-PAM3, but increases the power penalty relative to 25 Gb/s duobinary even more up to 4.8 and 6.2 dB.

Two observations are highlighted. First, 30 Gb/s DB-PAM3 shows 2.0 dB better receiver sensitivity than PAM8 at the same data rate. 37 Gb/s DB-PAM3 even offers 7.5Gb/s higher capacity while still having 1.2 dB margin to PAM8. Second, even though signals were transmitted with a MZM, 25 Gb/s duobinary experiences a severe dispersion penalty of 6.2 dB after 40 km of fiber at 1550 nm. Due to the lower symbol rate of DB-PAM3 at 30 Gb/s, its 40 km fiber penalty is only 1.6 dB. Even with the lower B2B base sensitivity of DB-PAM3, it outperforms normal duobinary after 40 km of fiber.

Among the various possible application scenarios of DB-PAM3 we highlight the following three scenarios. First, current efforts for 100 Gb/s PONs focus on 4 wavelengths with 25 Gb/s each, therefore requiring 4 transceivers. With the use of 25 GBaud DB-PAM3, the available data rate per wavelength is 37.5 Gb/s. Thus, with the use of 3 transceivers a data rate of 112.5 Gb/s can be achieved. The additional 12.5 Gb/s can be utilized for the FEC overhead, thereby bringing the net data rate available to the users closer to 100 Gb/s. While it is unlikely that in the near-term the extended power budgets (e.g. PR30) can be met by DB-PAM3 due to the additional 3.8 dB receiver penalty relative to normal duobinary, the low and medium power budgets, PR10 and PR20, seem plausible, due to the at least 5 dB smaller channel insertion loss at these power budgets. A significant number of applications are actually in these lower power budget classes [5, 6], Thus a smaller size PON can be served closer to a net 100 Gb/s data rate in a cost-effective manner by a 25% reduction in number of transceivers. Second, if longer reach and receiver sensitivity is required, the symbol rate of DB-PAM3 can be reduced to 20 GBaud. Using 4 transceivers a total data rate of 120 Gb/s would be available, allowing room for a stronger FEC with 20% overhead. Depending on the particularly chosen FEC, this can raise the pre-FEC BER limit to 10-2. Under these conditions the difference between duobinary and DB-PAM3 at their respective FEC limits reduces to 1.7 dB, while DB-PAM3 still offers at least 10% higher net data rate than DB-OOK.

In an embodiment of the invention the modulation format DB-PAM3 is used in a flexible modulation scheme, as was described above. The tradeoff of DB-PAM3 in capacity and receiver sensitivity is well suited for the use inside such a scheme. Flexible modulation allocates higher-order modulation formats to those ONUs that receive enough optical power to support it, while at the same time retaining support for lower-order modulation formats for those ONUs that require it. Therefore the deployed network can be utilized more efficiently, resulting in an increase in aggregated data rate of the network and reduction in congestion probability.

Using the statistics of discussed and shown in [6], the aggregated downstream data rate of a 10G based PON can be increased by 192% to 29.2 Gb/s if 10 Gb/s OOK, 25 Gb/s DB-OOK, and 37.5 Gb/s DB-PAM3 are used together.

Due to the use of the flexible modulation scheme, this additional aggregated data rate comes from the more efficient use of the deployed network and would not require any sensitivity upgrade of currently deployed 10G receivers in the ONUs. It would be expected that the OLT transmitter 2 would require a higher bandwidth than the currently deployed 10G transmitters. As shown previously [6,7] the use of multiple modulation formats within a PON does not introduce any penalty, furthermore TDM time-slot redistribution allows nearly all users to experience an increase in network capacity.

DB-PAM3 offers an increase in capacity while reusing current 10 Gb/s based receivers. It offers an interesting mid-point between on the one hand duobinaryOOK that has a smaller receiver sensitivity degradation and on the other hand DBPAM4 with high capacity but poorer receiver sensitivity. Compared to the use of PAM8, DB-PAM3 is able to provide the same data rate, with 2 dB improved receiver sensitivity, and less strict linearity requirements. For a small-scale PON, a cost-attractive attribute is the possibility to provide 112.5 Gb/s data rate over only 3, instead of 4, transceivers. In a flexible modulation scheme, DB-PAM3 enables the combined use of 10 Gb/s OOK, 25 Gb/s DB-OOK, and 37.5 Gb/s DB-PAM3. This more efficient use of the deployed network can increase the downstream aggregated data rate of a 10G PON from 10 Gb/s to 29.2 Gb/s.

For the above mentioned flexible aspects of the networks, flexible transmitters are preferred, which enable switchable output levels according to the user demands. As costs are the important factor in the field of access networks, an OLT is provided which is arranged to generate the various modulation formats at various clock frequencies in a cost-efficient manner. The OLT described with reference to Figures 15-21 makes an expensive, digital to analog converter (DAC) superfluous. The OLT supports the basic system, and the extensions towards multiple wavelengths and Compressed PAM. The transmitter may require some adaptations, but these will be self-explanatory for experts.

The schematics of Figures 15-18 all serve the same purpose: to generate with the same components three possible modulation formats by changing the number of possible signal levels: OOK (On-Off Keying, sometimes also called NRZ (non return to zero)), PAM3 (3 level Pulse Amplitude Modulation), and PAM4. Furthermore the clock rate can be changed between e.g. a base rate, a rate 2 times higher, and rate 2.5 times higher. This higher clock rate, if sufficiently increased, changes the shape of the original modulation formats (OOK I PAM3 I PAM4 ) to a duobinary form of these modulation formats: DB-OOK, DB-PAM3, DB-PAM4.

Figure 15 schematically shows a signal generator 80 of an OLT according to an embodiment. The signal generator 80 comprises a first OOK 81 and a second OOK 82 each comprising electrical circuitry for converting a digital binary signal to a driving current with two possible output levels. The signal generator 80 further comprises a power combiner 83 for combing the two incoming signals into one output port. The signal generator 80 generates an output (OUT) towards an optical conversion circuit, as detailed in the next section with reference to Figures 19-21. Between the output of the first OOK 81 and the input of the power combiner 83 a first variable attenuator 84 is arranged. The first variable attenuator 84 can be switched between 0 and -2.5 dB depending on a modulation control signal. The modulation control signal could be generated by the processor itself using a channel quality signal received from the ONU 3. Alternatively, the modulation control signal can be generated by a modulation controller external from the signal generator 80. Between the output of the second OOK 82 and the input of the Power combiner 83 a second variable attenuator 85 is arranged. The second variable attenuator 85 can be switched between 0 and -2.5 dB depending on the modulation control signal.

Figure 16 shows an alternative embodiment of the signal generator 80. In Figure 16 the signal generator 80 comprises an attenuator 87 switchable between -6 and 0 dB, and a phase delay module 86 to ensure that the signals coming from both the first OOK 81 and the second OOK 82 take the same time to propagate towards the Power combiner 83. So the signals coming from the OOKs are in-phase, and can be suitably combined. In this embodiment a further switchable attenuator 88 is arranged between the output of the power combiner and the output of the processor 80. Bothe the attenuator 87, the further attenuator 88 and the phase delay module 86 are controlled by the modulation control signal.

Figure 17 shows a further alternative. This version requires one attenuator less as compared to Figure 15 and 16. In Figure 17 the total output swing at the output of the processor 80 will vary slightly for the different modulation formats, thereby possibly degrading the maximum signal quality.

Figure 18 shows yet a further alternative of the signal generator 80. The variable attenuators 84, 85 of Figure 15 are changed for a variable output stage 91, 92 of the OOK modules 81, 82. The variable output stages 91, 92 allow to change the output swing of the two OOK modules 81, 82 independently, thereby having the same effect as the variable attenuators 84, 85.

Figure 15-17 show examples of values of the attenuators (0/2.5/6) dB. The values are in no way intended as limiting, other values may be possible. It is noted that the difference between the attenuators is more important than the absolute values. So it would be possible to recreate more or less the same results with an absolute change in both of the attenuators.

Now various possible ways to modulate the electrical signal onto the light will be discussed. The schematics at Figures 15-18 all generate an electrical signal. This electrical signal in its turn has to be modulated onto an optical signal. The function of all schematics shown in Figures 19-21 is to convert a modulated electrical signal to a modulated optical signal. The schematics of Figure 19-21 are known as such, and are purely meant for explanation how these system commonly work. Three common ways to do so are by Directly Modulated Laser, see Figure 19, by Electro-Absorption Modulator (EAM), see Figure 20 and by Mach-Zehnder Modulator Figure 21.

In Figure 19 the arrow IN indicates a Modulated signal coming in from the previous section. The Bias Tee component shown is used for allowing to combine a modulated (high frequency) signal, and DC (low frequency) signal Laser: A laser capable of direct modulation may e.g. be a DFB (distributed feedback) or DBR (distributed Bragg reflector) laser.

In Figure 20 the principle of Electro-Absorption Modulator (EAM) is shown. Here the optical light generation and the modulation of the electrical signal onto to optical signal are separated. First the light is generated by a laser that is always on (see CW Laser 201), next the EAM 202 absorbs some of this light depending on the modulating electrical signal. This is a somewhat more expensive solution, but the resulting signal quality is better as compared to the solution of Figure 19.

In Figure 21 the principle of a Mach-Zehnder Modulator is shown. Similar to the setup of Figure 20, the optical signal generation and modulation are separated in two components. The signal quality may be even better as compared to EAM, but the price of the MZM component 211 is higher than that of the EAM component.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments.

In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. Use of the verb comprise and its conjugations does not exclude the presence of elements or steps other than those stated in a claim. The article a or an preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

REFERENCES

1. D. van Veen and V. Houtsma, “High speed TDM-PON beyond 10G,” in Proc. OFC 2016, paperTu3C.3

2. C. Ye et al., “Demonstration and Analysis on PAM-4/8, DB-PAM-2/4 and DMT Formatted TDM-PON with 25Gbps, 40Gbps, 50Gbps Capacity per Lane using Economical 10Gbps Transceivers,” in Proc. ECOC 2016, paper Tu3F.3

3. S. Walklin and J. Conradi, “Multilevel signaling for increasing the reach of 10 Gb/s lightwave systems,” JLT 17, 2235-2248 (1999).

4. L. Suhr et al., “112-Gbit/s x 4-Lane Duobinary-4-PAM for 400GBase,” in Proc. ECOC 2014, paper Tu.4.3.2

5. P. Miguelez, “Operator Consensus for NG-EPON,” IEEE802.3ca Task Force meeting Sept. 2016, miguelez 3ca 1a 0916

6. R. van der Linden et al., “Increasing Flexibility and Capacity in Real PON Deployments by Using 2/4/8-PAM Formats,” invited in J. Opt. Commun. Netw.

9, A1-A8 (2017)

7. V. Houtsma and D. van Veen, “Unified Evolution-Ready 25 Gbps NG-PON Architecture” in Proc. ECOC 2016, Th2P2.72

8. V. Houtsma et al., “Demonstration of Symmetrical 25 Gb/s TDM-PON With Multilevel Interleaving of Users” in JLT 34, 2005-2010 (2016)

Claims (17)

CONCLUSIONS An optical line terminal for use in an optical communication system, the optical line terminal comprising: - a light source for generating optical signals by optical fibers of the system; - a processor adapted to modulate the optical signals of the light source with a modulation format and a clock speed, and to receive information about the quality of one or more channels between the optical line terminal and one or more end users, and depending on the specific channel received quality, adjust the modulation of the optical signals for that particular channel by changing the clock speed. The optical line terminal of claim 1, wherein the clock speed is increased so that, while the system bandwidth remains constant, the transmission type changes from a binary to a dual binary modulation type. 3. Optical line terminal as claimed in any of the foregoing claims, wherein the processor is adapted to receive information about supported modulation formats for all end users and to vary the length of a time slot for a user, depending on the supported modulation formats, such that the length of the time slot for a user with a higher supported modulation format can be shortened to increase the available capacity of the other users who do not support higher modulation formats and faster clock speeds. The optical line terminal according to any of the preceding claims, wherein the optical line terminal comprises an equalizer adapted to change the bandwidth of a channel, and thus can change the transition point at which point the clock speed leads to a binary or duo binary modulation type. The optical line terminal according to any of the preceding claims, wherein the modulation format is a format from the non-limiting list of formats compliant with PAM-M and DB-PAM-M, with M> 2. An optical line terminal according to any one of the preceding claims, wherein the optical line terminal comprises a number of light sources each operating on a different wavelength, the processor being arranged to use the channel quality information of the various users for applying it indicate a specific wavelength. An optical line terminal according to any one of the preceding claims, wherein the processor is adapted to change the number of levels per symbol transmitted. An optical line terminal according to any one of the preceding claims, wherein the processor is arranged to allocate a particular group of users to decode certain bits from the multi-level signal. An optical line terminal according to claim 8, wherein the processor is adapted to change the relative position of levels of a multi-level symbol and to create uneven relative level positions of the multi-level symbol. An optical line terminal according to claim 8, wherein the processor is arranged to assign users with a lower quality of the channel to those bits in the multi-level symbol which require a lower quality of the channel, and to users who have a lower quality of the channel. higher channel quality to those bits that require a higher channel quality. An optical line terminal according to any of the preceding claims, wherein the terminal comprises a plurality of OOK modules and wherein the processor is adapted to generate multilevel signals by changing the output of the OOK modules. An optical line terminal according to claim 11, wherein the terminal comprises a number of variable attenuators, each arranged after one of the OOK modules. An optical line terminal according to claim 11, wherein the output is changed by a variable current source in the OOK modules. An optical communication system comprising an optical line terminal according to any one of the preceding claims and a plurality of optical network units connected to the optical line terminal via a passive optical network. An optical network unit for use in an optical communication system according to claim 14, the optical network unit comprising: - a light source for generating optical signals by optical fibers of the system; - a processor adapted to modulate the optical signals from the light source with a modulation format and a clock speed, and to receive information about the modulation type to be used on a channel between the optical network unit and an optical line terminal of the communication system, and depending on of the received information, adjust the modulation of the optical signals by 10 means of changing the clock speed. An optical network unit according to claim 15, wherein the optical network unit receives information about the modulation type used when the optical line terminal sends signals to the optical network unit. An optical network device according to claim 15, wherein the optical light source for generating optical signals is adjustable to a specific wavelength, and wherein the receiver is adjustable to a specific wavelength. 1/14