Patent Application
20110020006
An optical link for communicating a payload data stream between a near end transceiver and a far end transceiver via an optical communication channel.
The demands for ever-increasing bandwidths in digital data communication equipment at reduced power consumption levels are constantly growing. These demands not only require more efficient integrated-circuit components, but also higher performances, interconnect structures and devices. Indeed, as one example, the International Technology Roadmap for Semiconductors (ITRS) projects that high performance chips in the very near future will have operating frequencies, both on-chip and off-chip, rising above 50 GHz. Conventional metal-wire based interconnects have played a central role in the microelectronics revolution. It is apparent that wire-based interconnect devices will be challenged to enabling even higher operating frequencies.
However, besides challenges with regard to bandwidth, the wire-based interconnect of the future may struggle significantly with a high power consumption. The power requirement of electronic components typically increases with increased bandwidth, which in some cases results in increased cooling requirements which further increases the power consumption of the electronic system as a whole. The power and cooling requirement may be particularly challenging to meet in data centers where larger quantities of servers are pooled often closely spaced. Such pooling inherently requires large quantities of interconnects which therefore may add significantly to the power and cooling requirements of the datacenter.
One approach to solve this problem includes utilizing optical interconnects as an alternative to wire-based interconnections, as optical fibers have a significantly higher bandwidth relative to an electrical wire. It is therefore an object of the present invention to provide means for reducing the power requirement of an optical interconnect.
An optical interconnect is typically composed by a transceiver module in each end adapted to transmit optical information along one or two optical fibers. The transceiver may be formed by a single chip or two or more chips. The transmitter of each transceiver typically comprises a driver circuit arranged to drive a light emitter coupled to the fiber typically with a binary signal, and a receiver circuit coupled to a photo diode coupled to the fiber arranged to receive the signal. In such a setup the light emitter typically consumes a significant part of the power requirement of the optical interconnect.
In a typical optical interconnect Vertical Cavity Surface Emitting Laser (VCSEL) diodes are utilized as light emitter to transmit binary data over optical fibers. However, the light source may in principle be any suitable light source and the transmitted waveform may be any suitable waveform for transmitting information. Most light emitters have a threshold current above which they substantially begin to emit light. Increasing the current driven through the emitter from zero to above said threshold may be time consuming, and therefore a bias current is typically driven through the light source. Often the bias current is set just below, at the threshold or above the threshold but it may also be set to be well above threshold. This bias current is often programmable so as the same circuit design may be utilized to drive different light emitters and/or used for different applications. Additional time varying current which modulates the emission from the light emitter is referred to as the modulation current.
For a VCSEL the suitable modulation and bias current values change with temperature so the driver is programmed with VCSEL characteristics that describe the performance as a function of temperature. The actual operation of the driver is then determined by the temperature and the programmed values.
To ensure stable operation it is often necessary to characterize each VCSEL individually, or if the uniformity from VCSEL to VCSEL is sufficiently good it is necessary to characterize the VCSEL on a batch level to determine suitable bias and modulation currents. One of the main problems with the manufacturing of VCSEL is that the uniformity from batch to batch may be poor and therefore it may not be possible to use the same programming for all VCSEL's.
In addition, the bias and modulation currents change with the age of the VCSEL, but most systems do not have a method for compensating for aging of the VCSEL. Some systems overcome this by a build in timer adding further complexity to the system. Again aging may not be well characterized for VCSEL's and may have large batch to batch variation. This also means that it may not be possible to have a reliable end of life warning at least not without a large safety margin.
An object of the present invention is to overcome one or more of the drawbacks of the prior art described above.
In one embodiment the invention relates to an optical link for communicating a payload data stream between a near end transceiver and a far end transceiver via an optical communication channel, said near end transceiver comprising a near end receiver (near-Rx) and a near end transmitter (near-Tx) and said far end transceiver comprising a far end receiver (far-Rx) and a far end transmitter (far-Tx), wherein said far-TX is adapted to transmit a link data stream to the near-RX beside the payload data stream from the far end to the near end. In one embodiment the link is arranged so that the link data may be relayed from said near-RX to the near-TX so that adjustment may be made to the near TX according to the link data. In one embodiment this means that the far-RX, which is connected to the near-TX, functions as a monitor of the link quality and provides feedback, via the optical link, to the near-TX. In one embodiment the near-RX performs the same function in relation to the far-TX. Any system variations due to temperature and/or aging may be removed automatically. As a consequence the transmitter, in particular the light emitter such as a VCSEL, may not have to be characterized and adjusted in the manufacturing process. In one embodiment a simple link test is performed to ensure that the link is operating correctly. By way of the invention the optical link may be allowed to configure the bias, modulation currents as well as other performance parameters such as e.g. equalization, (e.g. pre-emphasis), and receiver settings.
Accordingly, in one embodiment the invention relates to a method of transmitting and receiving data via an optical link according to the invention comprising
As the transceivers are optical transceivers the transmitter of either end will inherently comprises a light emitter, throughout this text exemplified by a VCSEL, and a corresponding driver circuit. Therefore, in one embodiment adjusting the near-TX comprises adjusting driving conditions for said light emitter, such as bias and/or modulation currents. The link inherently comprises a near-end input from which the link may receive payload data stream from a system in which the link operates. Similarly the link comprises a far-end output where the payload data stream may be delivered by the link.
For convenience the two ends of the optical link has been referred to as near and far ends where the far-TX transmits link data to the near-RX. However, in one embodiment the ends are reversed so that the near end performs the functions of the far end described above and all other features are adapted accordingly. In one embodiment both ends transmit and receive link data from the other end so that both the far-TX and the near-TX may be adjusted according to the invention. Link data sent in the link data stream may in principle be any data suitable for indicating performance or quality of the link. As will be realized by the skilled person several figures of merit are possible. When the far-RX receives payload data it may measure one or more parameters related to the quality of the received data. These parameters may include one or more of the average optical power (AOP), the optical modulation amplitude (OMA), and extinction ratio (ER) as well as a power in the signal in one or more frequency intervals. The latter may in one embodiment be applied to determine the shape of the signal or some indicator of the shape of the signal. In one embodiment the shape of the signal may be applied to determine whether sufficient bias is supplied to the VCSEL as insufficient bias current may affect the shape of e.g. binary 1's. Similarly in one embodiment the link may be designed to find and operate the VCSEL with minimum bias. In one embodiment the near-TX may reduce bias current until the shape of the signal is affected to determine a lower threshold. In one embodiment a lower threshold is determined at least partly by determining the extinction ratio. In one embodiment the link may be designed to operate the VCSEL with a bias current above threshold. In one embodiment with a minimum bias current satisfying this requirement the ER may be determined as the bias current is decreased. When the ER rises significantly the bias current may be below threshold. In one embodiment measurement of the shape of the signal is applied to determine a setting of pre-emphasis of e.g. on the front or back of a pulse corresponding to a binary 1. As will be clear to the skilled person, the link data may comprise the measured parameters themselves and/or calculated derivates thereof. For example, the ER may be a calculated measure at least partly based in the OMA and the AOP. Depending on the design the calculation of the ER may be located at the far-end transceiver (e.g. in far-RX), the near-end transceiver, or in a controller device. The same consideration may be valid for other calculated parameters. However, whether the measured parameters themselves, are transmitted via the link data stream or data processing is performed prior to this transmission may affect the required bandwidth of the link data stream. In one embodiment one or more of said far-RX, far-TX, near-RX, near-TX and a controller circuit perform computation on the measured parameters. In one embodiment such processing is performed in the far-end to reduce the amount of link data to be transmitted. In one embodiment processing comprises filtering and/or averaging either weighted or un-weighted. The link data may also comprise adjustment commands for the near-TX, such as increase/decrease bias etc. In one embodiment such adjustment commands will be derived from measured parameters e.g. in one embodiment derived from an initialization or adjustment algorithm arranged to adjust the parameters for the near-TX.
In one embodiment the link comprises one or more controllers adapted to control the adjustment such as setting the current source in the driver circuit of the transmitter based on the link data stream. The controller may be external to the transceiver, internal to the transceiver, form part of the receiver, the transmitter or a combination thereof.
The quality of data is in this context taken to relate to characteristics of the transmitted payload data relating to the comprehensibility of the payload data rather than the contents of the payload data.
In use the optical link is used to transport, generally referred to as transmit, payload data between two parts of a system. This could be internally in a computer device, for example relaying data from a CPU to a harddrive, between two servers, between memory and processor, to and from a graphics adapter in a computer, or in a link between components in a network. In order to provide easy integration it may in one embodiment be advantageous that the link does not impose protocol requirements upon the payload data. In that way that optical link and the system may in one embodiment be designed separately with fewer or no consideration of functionality of the other. However, in one embodiment such considerations do comprise rudimentary requirements such as maximum bandwidth of the link and current supply to the links etc.
In one embodiment the optical link is an optical active cable where the transceivers and the optical communication channels (commonly one or more optical fibers) are integrated in such as way that the cable interfaces with the system in a manner similar to that of an electrical cable. From a system perspective the only differences is in one embodiment that the active cable requires a supply of electrical power and that it has a larger bandwidth than the electrical cable.
In one embodiment the optical link is a short reach optical link having a length of the optical communication channel of less than 5 km, such as less than 1 km, such as less than 100 m, such as less than 50 m, such as less than 20 m, such as less than 10 m, such as less than 1 m, such as less than 50 cm, such as less than 25 cm, such as less than 5 cm.
In the context of the present invention the phrase “beside the payload data stream” is in one embodiment taken to mean that the link data stream may be transmitted while payload data is transmitted as opposed to between packets of payload data. In one such embodiment the link advantageously does not impose requirements on the temporal structure of the payload data stream provided to the input of the link. Such a requirement imposed on the temporal structure could for example be that the payload data stream is in the form of a burst mode data stream where a header may be transmitted before each pulse of data. In one embodiment the “beside the payload data” means that the temporal structure of the payload data stream is not altered in the link. In one embodiment “beside the payload data” the link data stream is modulated on top of the payload data stream rather than encoded into the data stream. In one such embodiment transmitting beside the payload data also allows the link to be implemented without circuitry for re-clocking the transmitted data streams (such as via a CDR) such as to remove an encoding of the link data stream into the payload data stream. Such re-clocking could for example otherwise be applied e.g. to receive a payload from the system in which the optical link is used, recode the payload data to allow for transmission of the link data stream, and remove this recoding at the other end. Accordingly, in one embodiment the invention provides the benefits discussed above without requiring adaptation by the system, in which the link is to function and/or without requiring substantial complexity in the transceivers such as by requiring re-clocking and/or buffering of payload data. In one embodiment the link may require the payload data stream to be encoded for example by 8B/10B encoding, to ensure a certain minimum bandwidth and/or a DC balance in the signal. Such encodings may in one embodiment be inherent to the system, in which the link is inserted, such as network system based on the Ethernet protocol. In one embodiment in which the payload data stream is such encoded the encoded payload data stream provided to the input of the link is in the context of the present text considered the payload data stream in relation to the link. In one such embodiment the link is arranged so that an encoded payload data stream is input to the link and the same encoding is found in the transmitted payload data stream on the output of the link.
In one embodiment the link data stream is transmitted beside the payload data stream along said optical communication channel. In one embodiment said link data stream is multiplexed with said payload data steam. In one embodiment the link data stream is modulated on the payload data stream. In one embodiment said multiplexing comprising one or more of DC level modulation, modulation of the modulation current and phase modulation. In one embodiment the near end receiver comprises a circuit for determining threshold of the signal in case of binary signal. This value is in one embodiment related to the DC level of the signal. In one embodiment the near end receiver comprises a circuit for determining the peak value of the signal, for example in order to set the threshold for a binary signal. In one embodiment this peak value is related to the modulation current. In one embodiment a link data stream may be received by monitoring one or more of such determined variables in the near end receiver within a frequency band. In one embodiment two or more of such variables are modulated with the link data stream to improve robustness of the detection of the link data stream. In one embodiment two or more of such variables are modulated in phase. In one embodiment two or more of such variables are modulated out of phase, such as 90 degrees or 180 degrees out of phase.
Compared to the bandwidth of the payload data stream Bdata, e.g. 10 GBit, the bandwidth of the link data stream Blink is in one embodiment considerably less, which in one embodiment allows the link data stream to be multiplexed with the payload data e.g. by modulation of the modulation current. In one embodiment Bdata/Blink≧10, such as Bdata/Blink≧102, such as Bdata/Blink≧103, such as Bdata/Blink≧104, such as Bdata/Blink≧105, such as Bdata/Blink≧106. The link data stream may utilize an error correcting coding to ensure that the link data is correctly transmitted.
In one embodiment the link data stream is transmitted along a parallel communication channel, such as an electrical wire. For an optical link this embodiment typically requires additional wiring. In some applications such complication may be tolerable whereas other applications will prohibit such an implementation. An additional channel such as an electrical wire may also require additional circuitry in the transceiver. However, in one embodiment a parallel communication may provide a better sensitivity in the transmission of the payload data stream because this has little or no interference from the link data stream. In one embodiment the parallel communication channel is an optical channel. In one embodiment such a channel is applied to test the link quality of two or more channels in the array sequentially.
In one embodiment the optical link is part of an array of optical links, such as e.g. 4, 8 or 12 channels. In one embodiment such an array comprises a channel dedicated to transmission of link data. In one embodiment such a channel or another channel is applied to transmission of a redundant payload data stream. In that way the quality of the transmission of the payload data of one of the other channels in the array may be investigated. In one embodiment such a channel is applied to test the link quality and subsequently adjust the drive parameters of one of the other channels.
In one embodiment a redundant payload data stream may be utilized to compare the opening in an eye diagram or other eye diagram parameters. The channel transmitting redundant payload data may in one embodiment use settings with relatively high power consumption but where good link quality is more likely. This link may then be applied for reference. As several channels may in one embodiment be optimized via a single channel transmitting redundant data the array may be run more effectively in total. In one embodiment the channel sending redundant payload data is used infrequently, such as to increase the life time of this channel and/or to save power. In one embodiment the link further comprises means for testing the channel applied to sending redundant payload data so a reduced life time of this channel is less likely to disrupt the performance of the system. In one embodiment the channel applied to send redundant data may change between the channels of the array.
To ensure a good link performance and/or to minimize the power consumption the skilled person may implement a large variety of control schemes. In one embodiment such a control schemes comprises
The critical link quality parameters provide a measure of the acceptable link quality. Such critical parameters may in principle be any parameter which may be determined from the received payload data stream. Examples comprise the OMA, ER, AOP, features of the eye diagram and bit error rates. Features of the eye diagram and/or bit error rates may be determinable via a redundant channel or via protocol implementation, such as via a checksum or a similar measure. Such an encoding of the payload data stream may be provided to the link in the payload data stream, i.e. from the system in which the link is inserted. This may be the case where the system applies such an encoding for other purposes.
The control scheme may further comprise:
The safe throughput parameters are drive parameters and perhaps receiver parameters such as parameter for which a safe transmission of payload data is guaranteed or at least very likely. These are in one embodiment used during start-up of the link, or at least one of the transceivers, so as to provide a starting-point for the link after which the transmitters, and perhaps receivers, may be adjusted based on the transmitted link data stream, which in turn is derived from the transmitted payload data stream. In one embodiment the link uses a known data set, such as a stored dataset or a dotting sequence, as the initial payload data in order to initialize the link. The safe throughput parameters may also be applied in the event of a data loss, so that the transmission of payload data may be restored. Depending on the design and implementation of the invention the safe throughput parameters may be provide throughput without exact knowledge of one or more parameters which would have been applied in the transmission of a prior art link. As discussed above in one embodiment the VCSEL will not have been subjected to calibration and therefore the safe throughput parameters may in such an embodiment comprise a bias current and modulation current which is likely to provide throughput regardless of, at least expected, production variations. In one embodiment the bias current of the safe throughput parameters is sufficiently high that the VCSEL is run above threshold to ensure that the 1's of the payload data stream are not distorted and/or that the due cycle is far from 50% due to a long turn on time for the VCSEL. In one embodiment the safe throughput parameters have different settings depending on the temperature of the transceivers. In one embodiment the safe throughput parameters are set independent of temperature. In one embodiment the operations of the transceivers of the link is independent of a measured temperature. In one embodiment the link is arranged to store operational parameters (dependent or independent of temperature) in order to utilize these parameters at another start-up and/or in a data loss event. In one embodiment the link may utilize two or more sets of safe throughput parameters, so that if e.g. the first set of safe throughput parameters is sufficient for providing a safe transmission of payload data the link may start-up operating closer to the desired operation. The second set of safe throughput parameters may in one embodiment provide higher likelihood of safe transmission but start the link up further from desired operation.
Once in operation the link may in one embodiment
In one embodiment said adjustment may be to achieve operation near said critical link parameter. In one embodiment the critical link parameter is an interval. Commonly, increased link quality will come at the cost of increased power consumption. In one such embodiment, a link quality better than said one or more critical link quality parameters will prompt the near-TX and/or far-RX to reduce power consumption. The critical link parameters are in one embodiment related to the requirement for providing a sufficiently stable link. Whether an adjustment is required or not may in one embodiment be determined via a specified tolerance.
In one embodiment the link comprises a control scheme comprising an iterative procedure for achieving safe throughput. In one embodiment the link initializes after having been turned on by transmitting real or stored data along the optical communication channel while indicating to the system in which the link is operating that the link is not operational. First after at least safe throughput is achieved does the link indicate that it is operational. In one embodiment this initialization procedure involves the utilizing safe throughput parameters such as described above.
The invention will be explained more fully below in connection with a preferred embodiment and with reference to the drawings in which:
The figures are schematic and are simplified for clarity.
In one embodiment the link is subjected to the following tests prior to being put into operation:
In one embodiment an initialization process is utilized instead of having a stored minimum bias current as a function of temperature. Here a safe throughput value well above threshold is utilized after which the transmitter is allowed to decrease the bias current until the ER is above max after which bias current is increased again.
As mentioned above, the light emitter of a transmitter may in principle be any suitable light transmitter. Accordingly, when reference is made to VCSEL's this should be viewed as exemplary rather than limiting to this particular light emitter. It should also noted that any features provided as part of the examples may be combined with any other features provided in the description unless the features are mutually exclusive.
| Number | Date | Country | |
|---|---|---|---|
| 61228848 | Jul 2009 | US |
