Design Article

Comment


BicycleBill

9/11/2010 4:12 PM EDT

Thanks to all for pointing out "when good links go bad"--and Patrick for seeing ...

More...



patrick.mannion

9/11/2010 11:11 AM EDT

Hi all, the link is now up and running again (no registration required). ...

More...

Extreme Design: Realizing a single-chip CMOS 56 Gs/s ADC for 100 Gbps Ethernet

Ian Dedic, Fujitsu Microelectronics

8/25/2010 7:38 AM EDT

To provide a long-haul, 100-Gbps, optical transport network with maximum reach and immunity to optical fiber non-idealities, the industry has settled on dual-polarization quadrature phase-shift keying (DP-QPSK) as a modulation method, which means that a coherent receiver is required. The biggest implementation challenge resulting from this decision is the need for low-power ultra-high-speed ADCs, and their technology requirements define the way that such a receiver can be implemented.

A 100-Gbps coherent receiver needs four 56-Gs/s analog/digital converters (ADCs) and a tera-OPS DSP which dissipate only tens of watts. This paper discusses the forces pushing towards a single-chip CMOS solution, and the challenges in realizing this.

Without suitable ADCs – especially with low enough power consumption – it is impossible to produce a 100Gbps coherent receiver suitable for a commercial optical network, as opposed to prototype systems only suitable for demonstration in the lab.

These ADCs need sampling rates of at least 56 Gs/s and resolution of 6 bits or more, with power consumption of no more than a few watts each to fit within the power constraints imposed by the system. To do this with sufficient dynamic range for input signals up to 15GHz and higher, it was thought that this would require technology such as very advanced SiGe or ultra-small-geometry CMOS (40nm or smaller). By extrapolating from historic advances in ADC design, it was predicted at the end of 2008 that suitable ADCs would not be available until 2013.

However, the development of new circuit techniques means that these ADCs actually became available in 2009 using 65nm CMOS. This brought forward the date at which single-chip 100-Gbps coherent receivers became technically and economically feasible, and has caused a significant change in the industry roadmap for these devices.

This Extreme Design article discusses the underlying design issues, how the design was implemented, and the Fujitsu’s first customer evaluation technology containing a two-channel 56-GSa/s version of the ADC. It is presented in pdf format (no registration required); to read it, click here.)

(Editor's Note: to see the previous articles in the "Extreme Design" series, click here.)

About the author
Ian Dedic is with Fujitsu Microelectronics Europe GmbH, Maidenhead, Berkshire, United Kingdom.

.





Subhajit

8/25/2010 10:28 AM EDT

The link to the 2-channel 56 GSa/s version of the ADC does not seem to work.

Sign in to Reply



iniewski

8/25/2010 8:00 PM EDT

What is SNR (or equivalent number of bits) for this converter? Kris

Sign in to Reply



DickH

9/7/2010 11:39 PM EDT

iniewski,
I chased this up as much as I could afford the time, and came to the conclusion that ENOB is less than or equal to 3.

Whether that's right or not, I can't be sure. I may be looking at the wrong Fujitsu papers (probably out of date). Would the author care to comment?

And could Somebody fix the link so that we can actually read the article?

Sign in to Reply



Frank Wiedmann

9/8/2010 5:50 AM EDT

You can find more information about this ADC at http://www.chais.info . In the OFC paper, they report an ENOB above 5.7 that is almost constant with input frequency (+/-0.2 from 1GHz to 15GHz) for a -6dBFS sinewave input.

For the full technical details of the design, you can read the patent application: go to http://ep.espacenet.com/ and search for EP2211468.

Sign in to Reply



DickH

9/8/2010 5:00 PM EDT

the link still doesn't work, and the reference in the pdf listing all the previous articles points back here. C'mon guys, fix it...

Sign in to Reply



IJD

9/10/2010 10:34 AM EDT

I've asked our marcom guys to contact EETimes and get them to fix the link -- the article's a *lot* more readable than the patent (unless you really enjoy explanations like "first switching circuit means connected to second switching circuit means"...)

Sign in to Reply



IJD

9/10/2010 11:44 AM EDT

If you don't mind registering, what I suspect is the paper (I didn't put it up at EETimes!) is the Technology Backgrounder at http://www.chais.info -- but if you really hate registering, you can get it from http://bit.ly/cUmXfT

Sign in to Reply



patrick.mannion

9/11/2010 11:11 AM EDT

Hi all, the link is now up and running again (no registration required). Enjoy!
Best regards,

Sign in to Reply



BicycleBill

9/11/2010 4:12 PM EDT

Thanks to all for pointing out "when good links go bad"--and Patrick for seeing it was fixed and letting us know.

Sign in to Reply



Please sign in to post comment

Navigate to related information

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)

Feedback Form