| . |
. |
The Problem with
Being Able to do Anything
Serious development on the VME started in the
year 2000 under General Manager Katsunori Seno.
Almost immediately after that, we got together
with the LSI Design Division.
Seno: During a period of company-sponsored
study abroad in the US, I participated in research
on dynamic reconfigurable circuit technology.
After returning to Japan, I decided to
begin work on Sony’s own development of this
technology based on that experience. Although
the concept of using software to
change the hardware configuration already
existed, it was not the case that technology
suitable for practical application had been
established. The definition of ‘capable of being
applied to anything’ was, inversely, being
the problem. For example, it was clear that
developing something that could handle application
areas from portable players to communications
equipment to medical equipment
would fail and that if we didn't focus on a specific
application area, optimization would not
be possible. Without such specialization, it
would not be possible to get aspects such as
power consumption and price down to the
point where the products would be adequately
competitive. With regards to this point, I was
lucky to have met Mr. Mamoru Yamazaki’s
group.
Yamazaki: At that time, we had narrowed our
focus down to silicon audio and were looking
into LSI development. While the Memory Stick
Walkman had already been released as a
product, even though, unlike the MD, which
was at its zenith at that point, there was no
mechanical system whatsoever, its batteries
did not last as long as those in MD players.
We were sure that if we could only produce a
product that achieved long battery life, it's
value as a product would increase.
Seno: Mr. Yamazaki happened to hear about
the VME and visited our group. “We're aiming
at developing something like this.” “That's exactly
what we're trying to develop.” We agreed
to cooperate immediately.
Yamazaki: Wasn’t that at exactly the time that
you, Mr. Seno, were thinking about promoting
the VME?
Seno: Since things only have value when they
are used in an end product, we were looking
for ways to commercialize the technology.
Since silicon audio has no mechanical components
and also does not require a large
LCD, the VME’s effectiveness in reducing
power consumption would be clearly revealed
in the dawn of its adoption in end products.
The best possible place for the VME to make
itself comfortable had appeared.
An Immoderate Goal...
The Microwatt Project was started in October
2000. The goal of that project was to reduce power
consumption during ATRAC3 audio data playback
to 5 mW by the fall of 2001.
Yamazaki: At that time, the estimate for systems
using a Sony DSP was that power consumption
would be in the 20 mW class. Thus
reducing this 20 mW to 5 mW represented a
high hurdle. Some of us had the thought that
maybe this was a rather irresponsible goal.
Looking back, it now seems a rather good
number to have chosen. If we had chosen 10
mW as our target, we would not have been
able to aim so uncompromisingly for such extreme
performance.
Seno: Since we in the VME group had the
target of 5 mW, we put aside all thoughts of
other application areas temporarily, allowing
us to focus all our efforts on audio.
Yamazaki: We were fortunate to be able to
work together on this as a joint effort from early
on in the project. Next, our encounter with the
Network Walkman Division was also perfect.
The test chip that included the VME was completed
in July 2001. We mounted it on a board,
and the instant we pressed the on button, it
produced music. The power consumption was
also perfect: it achieved a 4 mW level, exceeding
the goal.
Seno: Although we only realized this later,
that point corresponded exactly to the final
period when the devices to be used in the
Network Walkman were being decided.
Yamazaki: If we had not been able to prove
the correctness of our logic or show how complete
our implementation was, it probably
would not have been adopted.
Seno: That was the point when the VME, LSI,
and product development efforts were all tied
together.
Yamazaki: After that, we adopted the required
specifications for the end product and created
a final LSI. Since we were all, however, intent
on winning in the marketplace, the development
effort was truly dizzying. Missing the last
train became the expected end of the day, and
we became quite familiar with late-night TV.
We completed the prototype two months later,
and that one also worked on the first try. It felt
like a circus trapeze act, with the star being
handed off from one trapeze to the next; for
some reason, we had no failures.
Considerations Extended
by VME Adopters
What would be the best way to take advantage of
the VME? This requires careful attention to detail
at the microwatt level in the LSI design.
Yamazaki: First of all, please don’t think that
it's just a matter of just including a VME. When
the power consumption is at this low a level, it
can easily fluctuate by 2 or 3 mW due to minor
changes in the layout or other aspects.
Extremely detailed studies are required to
shave down the power to the microwatt level.
Seno: It is our LSI design technology capabilities
that have allowed us to drive it down to
this level.
Yamazaki: A lot goes in at the system level
as well. Although the CPU sleeps while the VME is operating, just sending a command
to do nothing consumes power. Therefore we
had to implement a variety of schemes, such
as putting the CPU to sleep completely yet
being able to wake it up immediately with a
signal from the VME. The VME performs
more than 1000 reconfigurations per second,
and since the put to sleep then wake up sequence is performed that many times as well,
how efficiently that could be performed was
crucial.
Towards Mobile Products
that are Still Unknown
The VME continues to evolve and LSI technology is looking to its next stage.
Seno: Since the new version of the Network
Walkman was released in April this year,
Yamazaki and I have been getting a lot more
requests for interviews. It is quite rare that
the LSI technology gets as much attention as
the end product.
Yamazaki: This time, we started development at the stage where the end product in
which it would be used was not yet decided.
We will perform market research to determine what is desired and consider various
architectures, including various combinations with the VME. We get a lot of satisfaction from the fact that we ourselves can propose what is possible and that the end product designers recognize that. Of course, a big
part of this was the good fortune of the timing matching perfectly with a time when a
switchover to a new product would be occurring in the marketplace.
Seno: Although Sony has a commanding
lead in accumulated dynamic reconfigurable
circuit technology, we intend to continue to
evolve the VME. While performance itself
will be an object of these efforts, fleshing
out this technology so that it can support even
more applications is also important.
Yamazaki: What sort of LSI performance
characteristics are desired to directly lead to
end products that are attractive users is still
a subject of ongoing research. These will, of
course, differ with the customer and with the
geographic area around the world. The point
that they can flexibly respond to such needs
is a strength of LSIs that include the VME.
This is something Seno and I have discussed
a lot. We really want to be involved in the
creation of new mobile products that can propose new lifestyles for the end user.
Seno: Although we have rather different
personalities, we do share the same value system.
Yamazaki: We are both intensely dedicated
to our work. But Seno has the character defect that he insists on dressing well even
when discussing these issues over drinks. I'm
always rather disheveled by the time we go
out drinking. |
|
Vol.42
