x
Optical/IP

AMCC Shrinks Sonet Equipment

Applied Micro Circuits Corp. (Nasdaq: AMCC) today announced a new silicon chip that promises to bring relief to carriers that are rapidly running out of space to house equipment (see AMCC Unveils New Framer).

The new chip, called Orinoco, promises to shrink the size of Sonet/SDH equipment by orders of magnitude. It does the work of more than a dozen of today’s chips -- enabling an end-to-end, copper-to-optics OC48 (2.5 Gbit/s) Sonet system to occupy a single circuit board rather than an entire rackful of equipment, as is currently the case.

Orinoco also promises to help equipment manufacturers speed up development times because they’ll be able to buy a ready-made chip rather than having to develop their own subsystems.

Orinoco takes 12 DS3 (45 Mbit/s) or E3 (34 Mbit/s) channels and aggregates them onto a single OC12 (622 Mbit/s) Sonet/SDH channel. This process is called framing.

However, framing circuits only account for about 20 percent of the real estate on the Orinoco chip. The rest of the space is devoted to so-called mapping circuits that reduce or eliminate timing errors such as jitter and wander -- in short, making sure everything stays synchronous. Timing errors are mostly caused by the fact that DS3s run at a different clock speed from STS1s -- the basic unit of Sonet -- which has a clock speed of 51.84 Mbit/s.

"We had many customers using our Nile chip [a DS3-to-OC12 framer] with external PLLs [phase-locked loops] to smooth the DS3 clock [take out the timing errors]," says Amit Banerjee, a senior marketing manager at AMCC. "These customers were asking us if we could eliminate the need for costly, hard-to-design-with PLLs."

Banerjee won't say how long it took AMCC's engineers to meet that challenge. However, integrating both analog (PLL circuits) and digital (framing) functions on the same chip was a major hurdle, he says. "It's relatively easy to design a chip with one or two PLLs on it -- people do it," he notes. But 12 PLLs, one per DS3 or E3, is a different matter, because the software design tools have to handle an incredibly complex and time-consuming simulation. Mixing analog and digital technologies on the same chip is also challenging -- all the more so because design engineers from different disciplines have to work together, Banerjee says.

According to Banerjee, Orinoco has been sampling for several months. One of the customers testing it out is Cyras Systems Inc., which is being acquired by Ciena Corp. (Nasdaq: CIEN) (see Ciena To Buy Cyras for $2.6 Billion).

"Integrating these PLLs onto a single chip to meet jitter and wander specifications for mapping DS3s into Sonet is a commendable achievement," says Sunil Tomar, Cyras's VP of engineering. The Orinoco could be the not-so-secret weapon that will help Cyras make its multiservice provisioning platform, the K2, the smallest on the market (see Cyras: The Next Cerent? ).

AMCC must be hoping that today's Orinoco announcement will put some pep back into its share price, which took a battering last Friday (see Market Pain Returns).

-- Pauline Rigby, senior editor, Light Reading http://www.lightreading.com

Peter Heywood 12/4/2012 | 9:02:43 PM
re: AMCC Shrinks Sonet Equipment There's got to be some snags to using AMCC's Orinoco. What are they?
Ajo 12/4/2012 | 9:02:20 PM
re: AMCC Shrinks Sonet Equipment A high density device like this has simular
problems to most any high density device. The
high framer count T1 devices that have recently
come onto the market have the same problems:

1. The standard connectors now take up more space
than the integrated circuits themselves. Until
customers start accepting new types of connectors,
the huge wire-wrap and BNC's are going to partly
negate the space savings of high density devices.

2. The whole system has to be designed with very
high density in mind. Not just for the above
reason, but also from a fan and power supply
perspective.

3. The customer has to purchase all the ports (and
support circuitry like LIU's, transformers, and
connectors) even if they only want 1/3 the number
of ports. This means the the part has to be
very competitively priced so the customer doesn't
mind paying for all those unused ports. Or the
vendor has to provide an alternative, lower port
count solution.

4. Solutions like these tend to not be cheap,
causing further confirmation of item 3 above.
There are a number of reasons for this, including
that they are proprietary solutions with little
competition, high package cost (due to I/O count),
and probably have low yields (after all,every
portion of every port must work).

5. Protection switching can be quite a challenge.

6. Although the devices may save money in the long
run, short term they increase engineering design
time and complexity. Purchasing also doesn't get
the volume discounts they were getting with the
lower port density parts.
[email protected] 12/4/2012 | 9:01:48 PM
re: AMCC Shrinks Sonet Equipment These are poor arguments. If you increase the density of any component there are trade offs.

(1) This has always been the case in Telecoms. Try re-partitioning your design maybe.
(2) This is called system design.
(3) Customers usually know how many ports they want and look for the optimum soltuion. It's not like they have to buy the biggest.
(4)Compared with less integrated parts there is usally cost parity at least. The "4x bandwidth for 2.5x the cost" rule affects even the semiconductor pricing in the end.
(5)Protection switching is no more of a challenge here than in any application.
(6)There is no such thing as a free lunch.When you consider the overall reduction in hardware with this type of integration, you can get year on year savings of 25% over the part's integration roadmap lifetime.
Ajo 12/4/2012 | 9:01:26 PM
re: AMCC Shrinks Sonet Equipment Agreed - there are always trade-offs. That
was the whole point to my response. Mr Heywood
asked what some of them might be, and I see that
you did not add any to my list.

In response to yours,

1. Re-partitioning the design? What in the world
are you talking about? The only way to get more
BNC's is to put them on the side or front or top
of the box, something that customers have
expressed a distaste for. Now, if they are
willing to be flexable, sure, we'll find a way to
do make it happen - but with the customers current
desire for DS3's to come out the back, the
connectors take up more space than the IC's. And
I disagree this has "always been the case in
telecoms." It has only recently become the case,
brought on by high density IC's.

2. Hmm. I said the system has to be designed
around high density devices (obviously implying
that old systems may not be able to support
such high densities), and you come back with
"this is called system design." Allllrighty then.

3. Once again, I said the 12 port won't be cost
competitive when the customer only wants 3 ports,
and you come back by saying basicly the same thing. That was my point exactly - you will miss
the business unless you engineer, make, stock,
market, and sell two different versions. All of
this decreases the effectiveness of doing the
high density design.

4. As time goes on, yes you will be correct. But
right now you are not. For the reasons I gave
(plus one I forgot: you often end up paying for
features you don't want or need), the high
density solutions tend to cost about the same
per port as medium density solutions. When it
first comes out, the only advantage of the high
density solution is space and maybe power. When
a newer, even higher density solution comes out,
it will push the previous high density solution
to your 4x for 2.5x.

5. The challenge is explaining to your customer
that it is ok that 11 other ports will take a
traffic hit because 1 port failed. At what point
does it become unacceptable? Hitting 27 ports of
traffic because 1 failed? Hitting 83 ports of
traffic because 1 port failed?

6. Once again I see you're agreeing while at the
same time, missing the point. Yes, long term you
will probably more than make up the difference.
But in the short term, the high density solution
may cost more.

I don't believe you showed any of these to be
poor arguments, and in fact, you restated several
of them. Sure, some are stronger than others,
but its when you look at the whole list that you
question just how high densities should go before
customers will not want your solution.
[email protected] 12/4/2012 | 9:00:33 PM
re: AMCC Shrinks Sonet Equipment Worth a look at the upside as stated here.

Single-Chip Aggregates A Dozen DS3/E3/EC-1 Lines Into One OC-12
http://tm0.com/sbct.cgi?s=1167...
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE