Charlie Oppenheimer is president and CEO of Digital Fountain (www.digitalfountain.com), a small, privately-held company based in Freemont, CA. The firm's innovative DF RaptorTM forward-error-correction (FEC) technology is an effective means of overcoming network impairments and providing high-quality service in commercial and military wireless-communications systems, especially with streaming data signals.
MRF: What attracted you to join this company?
Oppenheimer: There are two things you look for in an entrepreneurial situation. One is a very big profit. Big markets are where big opportunities are made. The other is some distinguishing characteristics in the company. And this company had both of those in place. This technology can be used in some many different systems: cellular, Internet, defense communications. It is limitless. The technology at its core is the result of the scientific breakthroughs of two leaders in the field, Dr. Michael Luby (co-founder and chief technology officer) and Dr. Amin Shokrollahi (chief scientist) of Digital Fountain.
Just to give you a sense of the accolades that these guys receive: in February there was a dedication down in San Diego for a new school in information sciences. They invited scientists for all around the world to join in a several-days-long celebratory event. The opening plenary panel with people like Andy Viterbi and Mike Luby. The moderator asked some questions to open things up and when he got to Mike Luby he asked how it was that he and Amin were the ones who are outside of the traditional information theory community but you were the ones who proved Shannon's theory of an information-ideal code. And to have them recognized among that class of audience just said so much about the technical basis for the company.
MRF: Did you see in the opportunity presented by Digital Fountain that you were a missing piece?
Oppenheimer: To be accurate, I actually joined the company a little before that. I wasn't looking at it as simply an opportunity to run the company. But it was really that the technology could be used in so many different arenas that I felt the company had the potential to be the Dolby of communications.
MRF: Do you see a little Digital Fountain "DF" logo on equipment the way the Dolby Labs logo is so pervasive on audio and broadcast equipment?
Oppenheimer: We are seeking to make the technology pervasive across so many applications. If at some point that it becomes that way there may well be value for our customers that feature it in that way. When companies put that Dolby logo on their products, it conveys a certain level of quality and performance. We certainly would like to be able to convey that type of value to our customers' products.
MRF: There is a certain level of confidence that comes from the "Intel Inside" logo or the Dolby Labs mark on a product. There is a feeling that you are buying quality, and that would be something to aspire to for a DF logo.
Oppenheimer: There are lots of companies that are trying to get their technologies licensed and they look to get their logo featured. We're a little different in that we are first focused on the application. The market perceptions will follow our deliver, not the other way around.
MRF: Not fully understanding your technology, perhaps you could explain an element about it: is it necessary for you to understand the nature of the transmitted errors in order to correct for them?
Oppenheimer: Actually, no. You don't need to know a lot of information about the signal. In the field of forward-error correction, there are two general classes of codes. There is a class known as error coding and a class known as erasure coding. Error coding is used everywhere. Any digital communications system in the world has error coding in it. Every hard drive has error coding. Error coding allows you to correct information that was somehow corrupted in transmission. Erasure coding is a complement to that. It allows you to recover data that was lost in transmission. You use it with error coding. But its use in communications is much newer. The technology that we are licensing is erase coding.
Really all you need is to know that there was an amount of information lost. If you know the total amount of information that needs to be conveyed, and you know how much was lost, then what our technology allows you to precisely determine the content of what was lost. Unlike error coding and some other erase coding approaches, this technology performs independently of the amount of loss, the pattern of loss, whether it is burst or random, but it doesn't matter to our technology.
What makes this so special? There are three areas in which the company focuses the application of this technology: mobile broadcasting, video over IP, and defense communications. n the mobile world, there is a tremendous surge in interest in delivering multimedia content over cellular telephones. It is a great new service opportunity for the carriers. But there is a problem. To send a stream to a cell phone, it uses between 10 and 40 times the bandwidth-of a voice communications. Wouldn't it be nice if the carriers could charge 10 to 40 times the price of a voice call for this new service? But that won't work. Economically, it is challenging if you were to envision large numbers of subscribers getting multimedia services all on a unicast, point-to-point basis. The cellular capacity would go quickly. All of the carriers are looking to supplement their cellular infrastructure with a broadcast overlay. The advantage of a broadcast approach is that you pay for the infrastructure once and the more subscribers you have there is no more incremental cost to the system. The problem is that broadcasting by definition is in one direction. If I am delivering high-value multimedia content, I want to be sure that everyone is getting it. Subscribers are not going to pay for services they are only partially getting. The one way to make sure that everyone is getting the content is by using this erasure code. It allows you to recover data that was lost.
MRF: Why is your approach better than other methods?
Oppenheimer: There are other erase codes. The DF Raptor technology uses less bandwidth overhead. It is less computationally intensive. And it provides greater loss protection. That is why we were so successful last year in winning two standards decisions: the 3GPP decision for MDMS service and the DVBH decision, also a mobile broadcast standard.
MRF: Obviously, there are also some politics involved in these decisions.
Oppenheimer: We are fortunate that it is such a good community that recognizes the value of what this does for the services. For the video over IP, we are seeing tremendous investment around the world, and our approach brings high value to those services as well. For the video over IP providers to succeed, the quality delivered to the television has to be every bit as good as we already enjoy on the television now. With a cable system, or a satellite delivery system, essentially what you have are dedicated communication highway "lanes" for every channel. Every channel has its own lane, and no one else gets into that lane. With video over IP, it is a highway and everybody shares all the lanes. Again, we have to make sure tat all the data gets there or the quality is not going to be that good.
Sometimes, I hear people say that they are going to design their IP network so that there is going to be no packet loss. And actually, that is silly. It is not really an ideal network. You want to have some packet loss. To have no packet loss, you are essentially saying that there would never be a traffic jam on the lanes, that there would never be congestion. There would never be any interference. You would have to so greatly overprovision that network as to completely negate the fundamental economic benefit of a packet network in the first place. Which is that it is shared and you get to share those resources among all these different applications. If you are never getting packet loss from congestion, then you have way overbuilt your network. You have spent too much money on it.
You want to have some packet loss, but packet loss is a problem for the applications. With video, you want to make sure that every packet of the video and the audio is getting there. That is what our technology does. In these high-bit-rate video applications, we offer low bandwidth overhead, high loss protection, and the least computational requirements.
As an example, in Japan, DF Raptor, is rapidly becoming a de facto standard. One of our licensees, Sumitomo Electric Networks, got into the business of building servers and set-top boxes and incorporating our technology. Their first customer, NTT, was not convinced that they needed any erasure coding protection. So they went to test it. This goes back to December 2003. They did a trial in which 300 set top boxes were installed in consumers' homes with a special feature. Our technology could be randomly turned on and off while customers were watching their televisions. They measured three different things: was DF Raptor on or off, how much packet loss was there, and how long did the viewer watch the television. Even when packet loss was less than one tenth of one percent, the viewing times doubled when DF Raptor was turned on. That was the end of that discussion. NTT was so taken with the effectiveness of our coding that they published the results of this trial.
MRF: Have customers who have seen what you offer turn in the other direction and investigate how "unrobust" they might make their network and still provide good quality because of your coding?
Oppenheimer: It is interesting that you bring that up. Not yet, but we are starting to get around to that question. In essence, what we are seeing is that the coding makes the network much more forgiving. And if you are a network engineer and you are responsible for delivering quality of service, and you now have something that is much more forgiving, even with electromagnetic interference, you would want this technology in order to be able to save money when building the network.
MRF: Where does your technology fit in with the defense community?
Oppenheimer: With defense, you have an even more complex problem. That is, you are dealing with wireless communications in the most impaired environments imaginable. There is nothing that we do that makes the signal more or less secure. Not only is it an impaired environment, but you may also have an enemy trying to jam your transmissions. And so the demonstration that we show in defense is a connection where we have 99-percent packet loss, and the data is still getting through. It takes longer for the data to get through with that kind of impairment, but it does get through. We don't design for a 99-percent loss scenario, but the fact that this technology allows us to get the data through may save someone's neck. That at the core is why we established this relationship with Northrop Grumman. They have seen that this technology can be a cornerstone of the nextgeneration communications infrastructure throughout defense, intelligence, Homeland Security, and space communications. And defense presents unique challenges. It is not only a community where you need to be on the inside, but also there are special requirements where you need to have people with security clearances, have secure facilities, and that is really a difficult burden for a small company. By being a technology supplier to Northrop Grumman, and for their partners, where they can provide a lot of the applications support and guidance, just gives us a huge leg up. Every defense that we go after now is with Northrop Grumman and/or one of their partners.
MRF: Do you approach the technology differently for military than for commercial customers?
Oppenheimer: The emphasis is different. In commercial, people are looking to figure out first how they can have the lowest possible capital operating costs. First and foremost. In defense, reliability is first. The points of importance are different.
MRF: Is there a place for Digital Fountain's technology in an optical communications network?
Oppenheimer: We haven't identified a real need so far. Every optical system has the error coding that I talked about before. The bit error rates are really not that high on optical communications networks. But if it is an optical communications network that is supporting IP, you now have packet losses because of the nature of the shared media. That is different than the errors coming from the optical network itself, and our coding can be effective. Now in satellite-communications systems, there is a need, especially in applications that are mobile. For fixed satellite applications, we can be used but it won't provide that dramatic a difference. But for a mobile application, you have problems of signal strength, fading, etc. That is why XM Satellite Radio and Sirius are both licensees of ours. Honda is a licensee of ours, but also an investor. People wonder why an automobile company would be an investor in a network transport company. But if you look at where automobiles are going, cars are becoming more and more information appliances in their own right. Honda happens to be particularly forward thinking in this and they feel that our technology is going to make a difference in the types of applications they are going to be able to do in a car.
MRF: What brings you to our offices today?
Oppenheimer: We have a new software library we are announcing called the DF ToughStreamTM Engine that is designed to simplify the use of FEC for streaming applications, such as IPTV and Internet TV. A user can specify the amount of correction and other parameters, but all the mechanics are taken care of in the software. We are now giving operators the ability to integrate this technology into their systems quickly.
MRF: Can the customer experiment with the parameters?
Oppenheimer: We built an evaluation kit that allows customers to play with the parameters and see the effects of the code.
MRF: Can you explain your three-tiered licensing agreement?
Oppenheimer: There are really two stages. There is the development stage and the sales stage. In the development stage, your engineers are working to integrate the technology and we have a development license that supports that activity. You can build whatever you like. When you are ready to sell something, we have an OEM license that gives you the ability to sell your products with our technology. There is also a short-term evaluation stage to experiment with the technology, but the second and third stages are for seriously working with our technology. Sometimes, our customers end up with a different application for the technology then the one they started with.