WTRS Executive Interview

Interview with Bob Metcalfe, Chairman & acting CEO, Ember Corporation.

Jan 23, 2006

1. Can you tell us a bit about your background?
I was born in Brooklyn, New York. Then I went to MIT, then I went to Harvard, eventually getting my Ph.D.. Then I worked at MIT and then I moved to Palo Alto Research Center where I invented Ethernet on May 22, 1973. And my life became about Ethernet for the next several decades. I started 3COM Corporation on June 4, 1979 for the purpose of promoting computer communication compatibility. Hence the name "3COM". Apparently I'm still in that business.
Ethernet and 3COM turned out to be successful. I retired from 3COM in 1990 and became a journalist for 10 years. I was a publisher pundit between 1990 and 2000, writing for InfoWorld where I had an Internet column. It was great fun. As an editor, I declined editing privileges after I wrote a column. One day I wrote about the IBM 3270, which you may remember, and an editor put a comma in 3,270 and I received gazillions of emails reminding me that I was a jerk and that I didn't even know how to spell 3270.
My book, Internet Collapses, is still available at Amazon.com. On 1/1/01, I became a venture capitalist, which is what I am now. And at the moment I am a general partner at Polaris, but also interim Chairman and interim CEO at Ember Corporation, which leads us to this conversation.

2. Who is Ember?
When I first became a venture capitalist, I was asked to advise a grad student on his Ph.D. dissertation, which turned out to be about embedded networking. His name was Rob Poor, and we agreed that after he finished his Ph.D. we would then talk about starting a company, which we did. It was Ember, and we started it roughly 4 years ago.
Ember is now a company of roughly 45 people, although it's growing at the moment. We are headquartered in Boston, Massachusetts, I am sitting there right now, and we also have a prominent office in Cambridge, England, which is where our semiconductor design center is. This is our CMOS radio group. Ember's mission is to network embedded microcontrollers with standards-based wireless semiconductors and software. We were working on this before there was a ZigBee, so for a while EmberNet existed before ZigBee. And now as ZigBee has come along, EmberNet has been tracking and in fact leading the standard. At the moment we ship two configurations of EmberNet. One is certified as a standard by ZigBee, and the other one works -- that is, it has accumulated features that are not in the standard that our customers over four years have indicated that they need to do their applications. Our competitors, in trying to find some competitive advantage against us, which they have trouble doing, they make stuff up about us. In particular they try to insinuate that EmberNet is not ZigBee, and of course it is ZigBee. It's certified, one of the few stacks that are certified by ZigBee to be compatible. The slight sliver of truth that they build their ugly accusation on is that Ember pre-dated ZigBee and has features in it which extend ZigBee. In fact we believe that these features will be standardized in time. So quite a bit is made of the fact that EmberNet is not just ZigBee. These vicious competitors try to imply that we are not ZigBee-compatible, when we actually are completely and are certified so prominently by ZigBee itself. So how they get away with this, I don't know.
ZigBee is our strategy, it is prominent in everything we do, so the fact that prominent industry analysts would ask such a question is heartbreaking because it means that we are not doing a good enough job in getting the message out.

3. What is EmberNet?
Most recently we have been calling EmberNet, "EmberZNet" just to emphasize that it has ZigBee compatibility. It is certified ZigBee-compatible, that's why we put the Z in the middle of the EmberNet. Customers who adopted EmberNet prior to the availability of the ZigBee standard may still be using the old EmberNet. But the latest versions of EmberNet, which are now shipping, and to which all of our customers are invited to upgrade, offer them a choice of EXACT ZigBee compatibility, which is better than anybody else's compatibility. The difference is that we also offer other features for our customers, if they want them. The principal difference between our two stacks today is that we have a tree stack, which is ZigBee certified, and a mesh stack which is going to be ZigBee certified when ZigBee has the mesh capability. ZigBee has not yet gotten around to having an official mesh routing function. We have had it for some years, in fact we started with it. That was Rob's Ph.D. dissertation. Since some of our customers use it, we cannot stop using it. So we issue it separately. This way our customers can be 100% certified ZigBee compatible, or they can use what is essentially ZigBee-plus. This is the same problem that Ethernet had 20 years ago. When you have a standard that is evolving, or in this case converging, and you are introducing products along the way, you try to guess at where the standard is going so you are standard with what has been approved and are ready to be standard with where it is going. You are constantly converging on a moving target. Everyone is doing this, its just that we have been at it longer, so we have a more competent stack that is more field tested and robust and feature-rich.

4. Why did you choose a wireless mesh solution for EmberNet?
Meshing of network nodes is a very important technology that adds capacity, redundancy, lower power, and other advantages. We don't think there is a ZigBee without a mesh, really.
We are just waiting for ZigBee to standardize the mesh, and frankly I don't know the timing of that so I cannot answer that question. In some cases our customers have chosen to go non-mesh, and generally that has led to problems. They have the Wi-Fi model. Wi-Fi is not meshed, everybody talks to the access point. And that means that if something gets between you and the access point, you are out of luck. And that is why we think ZigBee is adopting a mesh; in order to have a much higher reliability and much coverage. This is because star networks are just not robust enough, that is why the mesh is required.

5. Are you planning to migrate EmberNet to a standard radio technology like ZigBee?
Our plan and our strategy has always been to be ZigBee certified.
Our version 2.0 stack is ZigBee certified and the latest release, version 2.1 will also be ZigBee certified. Does that mean that we ship only the stuff that ZigBee has gotten around to standardizing? No, because our customers have products that they need to develop. They need to have those features in advance of the standard incorporating them.

6. Ember's chips run in the 2.4 GHz ISM band, do you see this as an obstacle to some market segments, such as Industrial?
We don't see it as an obstacle. We used to offer sub-Gig radios, but discontinued those. We are focused on 2.4 because it is available worldwide. There probably are some applications which are better served by radios below a Gigahertz, but we are not going to pursue those applications. You might want to go below a Gigahertz if you wanted more range and were not using a mesh. But the mesh capability offers the redundancy that gives you the same effective range as a sub-gigahertz radio. With a mesh you don't need to go below a gigahertz, and you still get the coverage.
We have just come out with a new 802.15.4-standard radio, which underlies ZigBee in the 2.4 GHz band and guess what else is in the 2.4 GHz band? Wi-Fi. We have very carefully designed into our radio the ability to co-exist with Wi-Fi. This is already proving to be a big selling point for our radios. We actually assume that our radios will be used in real configurations where there is a lot of interference in that unlicensed band (2.4GHz). Having the features in our radio to reject Wi-Fi noise and jump around it, which is our adjacent channel rejection feature, is very important. Some of the radios we compete with have no idea about this interference and they get completely shut down by interference from Wi-Fi.

7. What types of sales channels are in place today for Ember and how do you plan to develop them over time? Are you going to move out of the traditional semiconductor sales model?
We have news on that front. We recently announced a relationship with STMicroelectronics.
With that, you see the debut of what could be called a new channel strategy for us in that we plan to do a series of partnership announcements with microcontroller companies like ST. Our partnership with ST is really big and deep, but we plan to pursue others. The goal here is for us to be sure that our semiconductors and software can be purchased at the same time and place as our customers buy their microcontrollers. By partnering with ST, we second-source with them. That is, we are their second source and they are our second source. Anywhere you can buy an ST microcontroller, you can buy our products. In my whitepaper, on our website, you will see a list of the microcontroller vendors that we are aiming to partner with, now that we have partnered with ST. You will also see a general explanation of our world-view and general strategy. It's on the front page of our web site at Ember.com.
This is half of our channel strategy and it hinges on a particular product. You may know that we introduced the Ember 250 and the Ember 260. The 250 is a system on a chip solution with its own microcontroller and DSP, memory, and CMOS 802.15.4 radio, all in one chip. And when we conceived of that product we knew that it immediately puts us in competition with the makers of 10 billion microcontrollers per year. So we also came out with the 260, which is a very similar semiconductor to the 250 except that it is a network coprocessor containing our stack. Through its serial port you can hook it to any microcontroller, so it becomes your ZigBee coprocessor. And in my white paper I do list the four key benefits, but I will repeat them now:
1. The 260, through a serial port, can connect to any microcontroller
2. you don't have to put the stack in the application space of your microcontroller, because its out in the coprocessor
3. the mesh network and 802.15.4 processing are all floated from your application processor,
4. and finally, no one has to port a radio to your particular chip.
We can continue evolving and improving the radio independently. The microcontroller company and their customers don't have to keep up with changes in radio technology. The idea of a ZigBee co-processor is key for the customer and key for partnering with microcontroller vendors in that we can now offer them a way to put all of their microcontrollers on ZigBee by just buying or, in the case of ST, reselling the 260.
The other half of our channel strategy is that in addition to a series of partnerships with microcontroller vendors, we are continuing to engage directly with OEM customers who are designing ZigBee directly into their products. We have more than 100 design wins in that space and we are continuing to pursue those.

8. Based on your vast experience with the customer requirements in this space, how do you think end users will adopt these new wireless technologies?
The low-hanging fruit are applications which are already networked, only they use wires. Then ZigBee becomes a wire substitute. But the more interesting, and ultimately the more important, applications will be applications that were not possible before there existed low-cost, low-power radios like 802.15.4. The users who are already using networked controllers, like in their cars and in their security systems, will be the first because they will start having wireless versions of their existing applications. The more interesting ones, further downstream, will be the new users who will run across completely new applications that were not possible before.
A great example of that is from Raymarine. This is not a huge volume application, but it sure is interesting. I am able to remotely control my boat. Imagine a man falls overboard. The boat will know. And that was not possible before.
When ZigBee is used to connect the products of different suppliers together, so when you hook Philips Lighting and Honeywell HVAC together, that is multivendor compatibility. This is one of the big benefits of having a standard. This, of course, is further out.

9. Is there anything else you would like to tell our readers as a final note?
There are roughly 10 billion microcontrollers shipped every year. Very few of them are networked. All the companies involved in this ZigBee initiative are aiming to use a standards-based approach toward getting those embedded microcontrollers all networked.
That effort has been struggling for some years, as all standards-efforts do, and it seems to be gaining steam now. We see it in our own revenue numbers. We see it in the general goings-on in ZigBee. So we think that the networking of the embedded space is a happening thing. It is a huge space, so we don't think there is any particular need for the few vendors currently in the space to be bashing one another. We do not require any of them to fail for us to succeed.
###

More information about Ember here...

This interview ran in our Jan 23, 2006 newsletter issue.