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

Re: [802.3_100GNGOPTX] Next Generaton 100 Gbs/ PMD technology alternative commentary



Hi Yurii

Absolutely.  

We're not addressing the costs or competitive landscape of any particular market or standard.  We're mostly just trying to provide a frame of reference for people to evaluate the various kinds of opportunities, and encouraging people not to think of being in silicon as a 'magic bullet' that instantly solves cost problems, especially for low- and medium-volume products.  Even at millions of parts per year, a CMOS mask set can turn out to be non-negligible, for instance, if you're trying to hit a very low cost point per part.

Best,

M
On Apr 22, 2012, at 10:15 PM, Yurii Vlasov wrote:

Hi Michael,

Thank you for clarifying the goals and the intended audience of your very useful article. I completely agree with you (and with Chris in this regard) that it is important to distinguish hype and fiction from reality. Nobody wants to return back to 2000. You guys have done a great job in bustering myths. In doing so let' s try not to unadventeruously create new ones.

My intent of commenting on Chris's endorsement of your article for the whole 802 Study group is to exactly emphasize what you just confirmed: that your article was written for a different mostly academia audience. I believe you and the authors (with or without industrial experience alike) would agree with me that as such it should be considered with caution when its conclusions are used for the purpose different from what it was intended for. And especially when used as a possible base for rigorous assesment of cost benefits for a very specific (100GE 4ch 25G) and very large volume (millns/year) market that this Study group is focused on. 

Yurii

Sent from my iPad

On Apr 22, 2012, at 9:21 PM, "Michael Hochberg" <michael.hochberg@xxxxxxxxx> wrote:


Hi Yuri
Lity. 
I thought I'd weigh in on this.    Our goal was to try to address some common misconceptions about the field, rather than to push any particular kind of solution.  All of the authors on this see the overall success of this field as our goal. Tom, the lead author, and I have been academics for some time, though every author on the article has been involved with at least one silicon photonics commercialization effort.  This article is an exercise in saying some moderately heretical things about silicon photonics, to help the community focus on areas and projects that play to the strengths of the technology. 

I was hoping to convey in the article that the points we're making are the product of pretty broad and diverse experience sets:  Tom and I were both founders at Luxtera, and Theirry is still there.  Patrick is at IME, which does extensive commercial and pre-commercial work in a variety of fields, including silicon photonics. Dennis is an academic who's spun off several companies, and Steve has been involved with the BAE efforts in silicon photonics and in electronics.  So I thought it was kind of surprising that all of these people could be brought together to agree on a set of points about what's going on in the field. 

I regularly run into people talking about how doing some individual, relatively low-volume device (say a high-performance modulator or a photodetector) is sure be much cheaper because it's in silicon, without having thought about the cost of packaging, yield, mask costs, etc.  This is the kind of hype we were trying do dispel with the article.  The article was written for a broad audience outside of integrated photonics, most of whom have little or no industrial experience, and aren't used to thinking in terms of system costs, yields, and so forth.   Obviously you and the members of this IEEE group have a much more nuanced view of these things.

If the only applications that make sense for silicon photonics are ones that require very large volumes, because they require the development of custom processes, it's a big problem for the success of the field.  By contrast, if a lot of applications emerge that can be addressed through the use of common processes, as occurred in the electronics industry, then silicon photonics is likely to have a bright future.  The high volume applications are important, but for there to be a healthy ecosystem there needs to be a wider swath of opportunities.  And it's critical to be able to start a project in the field without budgeting for the cost of full mask sets - this is a big barrier to entry for everyone, from startups to academics to research groups at large semi companies. 

I agree that if one can use common processes and foundry infrastructure, the up-front costs can be substantially reduced.   That's a big piece of the reason that we've been working so hard on making MPW runs available through opsis.  And I certainly agree that the cost of the silicon chip itself is rarely going to be the dominant cost in a final product.   One of the signs of success for silicon photonics will be a gradual increase in the fractional value of the chip compared to the package.   The best path to this outcome is building complex systems-on-chip.

I'm glad that the article is provoking some discussion.  I'd be very interested to discuss specifics of what assumptions and conclusions you see as being incorrect.  I think it could be a jumping off point for a very productive discussion.

Highest regards, 

Michael Hochberg
On Apr 19, 2012, at 9:36 PM, Yurii Vlasov wrote:

Hi Chris

I am just saying the word of caution keeping in mind that  a) Nature Photonics is anything but an industry journal b) Commentary in Nature Photonics is not going through the rigorous reviewing scrutiny as regular papers c) The authors are mostly representing academia or academia-like institutions or organizations (with a single exception of Luxtera engineer) . As such, this Opinion paper should be taken with caution. Even the same authors in different circumstances, for example representing a view of a company at the industrial forum, will be most likely giving different messages.

The main disagreement is in consideration of the cost efficiency - the most important topic for this Study group.  Silicon photonics in the foundry does not necessarily require up-front infrastructure investment. the manufacturing cost is not necessarily directly related to the foundry volumes and, most importantly, the cost of the final product is not necessarily defined by the cost of silicon photonics piece of it. I would be very curious, by the way, to hear real cost estimates from Luxtera, but I suspect that this will remain in my wish list for some time. My opinion on cost topic I tried to express in the IEEE magazine Feb issue and in the presentation I gave at the last Plenary meeting.  

Saying that I agree in general with the view on on-chip optics as a very far away perspective, on packaging as the major challenge, and on the idea that integration with the most advanced CMOS  node is not an advantage.  

Yurii

______________________________________
Dr. Yurii A. Vlasov,  Manager
Silicon Integrated Nanophotonics  | tel: (914)945-2028
IBM TJ Watson Research Center  | fax: (914) 945-2141
1101 Kitchawan Rd, PO box 218     | email:yvlasov@xxxxxxxxxx
Yorktown Heights, NY 10598           |
http://www.research.ibm.com/photonics