Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [HSSG] Higher speed trade offs



Andrew,

 

There were two companies that developed products similar to the optical FM techniques that you are proposing.

 

In late 1990s, one of those companies, Kestrel Solutions of Mt. View, California, developed a technology they called Optical Frequency Division Multiplexing (OFDM,) for 10G and 40G transmission applications. Their 10G product modulated 64 sub-carriers with 64 OC-3 data streams. To do this transceiver function required seventeen 10”x 8” line cards to produce one 10G FM optical stream. The 40G product modulated 16 sub-carriers with 16 OC-48 data streams. To do this transceiver function required three 10”x8” line cards to produce one 40G FM optical stream. One of the major performance claims was the immunity of the modulation scheme to dispersion effects, which is a related problem to the skew problem that you are trying to solve. The products were not purchased by any service provider, so the company closed in early 2000s, after spending over $300M. This information is second hand so anyone with first hand knowledge should correct any errors in the above. I am not familiar with the details of the FM implementation used by the second company.

 

While it could be argued that it is straightforward to integrate all those line cards populated with discrete components onto a PIC (Photonic Integrated Circuit,) the above example should give caution about the difficulty of commercialization this type of technology.

 

Chris

 


From: Ellis, Andrew [mailto:Andrew.Ellis@xxxxxx]
Sent: Thursday, September 14, 2006 11:07 AM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [HSSG] Higher speed trade offs

 

Roger,

 

Some brief comments which I hope address some of your points below:

 

The term Coherent is intentional, but does not refer to coherent detection at the receiver, rather coherent addition of the WDM channels at the transmitter.

 

Thank you for your note regarding IP bylaws.

 

Coherent WDM does indeed require multiplexing and demultiplexing optics, in common with any WDM solution. However, within the transmitter, the WDM component is used to process continuous wave signals, and so dispersion is less of an issue. At the receiver, standard WDM components are used, with additional complexity compared to a traditional WDM system. The implications of a tighter channel spacing are compensated for the fact that one wavelength tuning operation (if necessary) gives the correct filter position for all 10 (in a 10 x 10 Gb system) wavelengths. So, to first order, I do not believe that the optical components for Coherent WDM would be particularly more expensive than for a WDM solution.

 

We have a solution for the MWS which provides arbitrary channel uniformity with good power efficiency, based on a single standard continuous wave laser source. An optical amplifier is indeed currently used between the MWS and the modulator array however, this may not always be necessary, firstly depending on reach and receiver sensitivities, and secondly based on refinements currently under investigation which eliminate the problem.

 

For the multiplexing, I think you already have the bones of the idea. You split the MWS into separate channels, each individually modulated. When recombining them however, you ensure that the optical phases and data delays are controlled to within a reasonable accuracy. This is the key aspect.  At the receiver, spectral overlap indeed occurs, however, considering that an interfering channel is effectively high pass filtered by the WDM demultiplexer (only the high frequency components are passed) the resulting interference is in the form of return to zero like pulses. Given that the signals originated from a single transmitter, with a common clock frequency, these RZ interference pulses may be aligned to the bit crossing of the demultiplexed channel, where they have negligible impact. Whilst this surely reduces the phase margin of the receiver slightly, it has enabled transmission of WDM signals with an information spectral density of 1 b/s/Hz with a data rate per wavelength of 42 Gbit/s. 10 Gb/s at 10 GHz should therefore be possible.

 

A more polished explanation of Coherent WDM multiplexing and demultiplexing may be found in Photonics Technology Letters, Vol 18, No 12, pp1338-

 

In summary, I agree, cost and power budget will be key factors, but Coherent WDM does offer the possibility of simultaneously achieving all of the technical targets.

 

Huge,

 

With regard to coding, multi-wavelength codes have indeed been investigated for use in WDM systems, including Coherent WDM. However, the ability to minimize the crosstalk within Coherent WDM itself was found to be so successful that minimal additional benefit from FEC was available.

 

Best wishes

 

Andrew Ellis

 

 

 

Senior Research Fellow

Photonic Systems Group

Tyndall National Institute and Department of Physics

University College Cork

Ireland

 

Phone: +353 21 490 4858

Fax: +353 21 490 4880

e-mail: andrew.ellis@xxxxxxxxxx

web site: www.tyndall.ie/research/photonics-systems-group/index.htm


From: Roger Merel [mailto:roger@xxxxxxxxxxx]
Sent: 14 September 2006 15:27
To: Ellis, Andrew; STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: RE: [HSSG] Higher speed trade offs

 

Andrew,

 

Welcome to the conversation!!!  You make some excellent points about the benefits of tighter wavelength packing which can be highly beneficial.  If I understand what you are proposing correctly, it is not truly Coherent but rather simply a technique to generate a comb of tightly spaced wavelengths.  If there is indeed a different form of modulation proposed, please clarify that.

 

Also note that if there is intellectual property involved regarding this technique, then you should familiarize yourself with the intellectual property bylaws of the IEEE.

 

Further, in your 2nd paragraph, you suggest the importance of making choices which reduce electronics cost (which is also desirable); however, then you suggest an optical solution which would superficially seem to consume (and possibly then some) the savings from this electronics with complex optics.

 

Such multi-wavelength sources (MWS) usually seem like an excellent idea (I’ve fallen for them too), but usually they do not reliably provide sufficient power in each of the small number of desired wavelengths (and often lack uniform power per channel).  The result is that to make them useful in a communication system, they require significant optical amplification.  For a longer distance telecom system (which tend to have greater tolerance to higher cost), this may be a viable option, but it is not clear how it would be viable in the relatively lower-cost application space that we’d be defining here.  Even if viable, it would have to be superior relative to the other alternatives available to us.

 

I’d also appreciate if you would suggest how to modulate and receive the individual carriers.  Presumeably, some form of demux is required before modulation, and multiplex them back together.  Multiplexing and demultiplexing with such tight spacing is certainly non trivial; however once the carriers have a modulated data signal (which for 10Gbps) is certainly on the order of 10GHz ; it suggests that, at a minimum, the demultiplexing at the receiver is likely to have data energy overlapping into adjacent channels even with perfect demultiplexing (but in reality optical filters will not have infinite slope roll-off and require some guard band).  As such, it seems challenging to carry 10Gbps of data in channels spaced ~10Ghz apart.  Historically, there have been extremely fine WDM proposals; however, these are usually limited to ~1Gbps in a channel (and they have not been successful to my knowledge).

 

I look forward to your further thoughts on this subject.

 

Best Regards,

 

Roger Merel

                                                                                                 

 

 


From: Ellis, Andrew [mailto:Andrew.Ellis@xxxxxx]
Sent: Thursday, September 14, 2006 6:41 AM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: [HSSG] Higher speed trade offs

 

Dear all,

 

I have recently joined this reflector, and I would like to make one observation which may be of benefit to the group. However, please forgive me if I speak out of turn, am confused about some of the acronyms or cover material already agreed.

 

There appears to be some considerable debate relating to the trade-off between reach, data rate, buffer requirements and the capabilities of electronics with manufacturability at a reasonable cost. This has lead to the M lanes at N Gbit/s per lane discussion, with M varying between 1 and 10 and N taking values between 100 and 10 Gbit/s respectively. Higher numbers of lanes appears good for electronics cost and reach, whilst a higher serial data rate appears good for minimizing buffer requirements and maintains a narrow occupancy of the optical spectrum (to allow operation in a WDM environment, in turn allowing multiple 100 Gbps Ethernet links to operate over the same fibre).

 

 I would like to propose the HSSG to consider the use of a new modulation format, currently known as “Coherent WDM”, in order to simultaneously meet all of these requirements. In Coherent WDM, we use a single laser source, minimizing inventory. This source is either a mode locked source, producing multiple carriers, or a standard cw source followed by at least one sine wave driven modulator (10 GHz in this example) in order to generate an optical carrier for each lane. These carriers are then modulated using an array of modulators (one for each lane, and each driven at 10 Gbit/s in this example), with, for example, a PIC similar to the one proposed by Drew Perkins. This produces a single 100 Gbit/s (in this example) signal, occupying a small spectral width of very close to 110 GHz which is transmitted as a single entity over a link (either point-to-point or a WDM network). The compact spectrum and careful design of the PIC and drive circuits combine to give negligible skew between the lanes, minimizing buffer requirements. You thus obtain the key features of the high serial data rates. It has been demonstrated that the reach of a Coherent WDM system is dominated by effects proportional to the data rate of each lane rather than the total data rate, and 10 Gbit/s electronics may be used. You thus also obtain the key features of a high lane count, low serial data rate link.

 

 I would be very happy to provide further details of Coherent WDM should anybody reading this contribution feel that it is appropriate.

 

 Thank you for your attention

 

 

 

Andrew Ellis

 

 

Senior Research Fellow

Photonic Systems Group

Tyndall National Institute and Department of Physics

University College Cork

Ireland

 

Phone: +353 21 490 4858

Fax: +353 21 490 4880

e-mail: andrew.ellis@xxxxxxxxxx

web site: www.tyndall.ie/research/photonics-systems-group/index.htm