SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Pastimes : Triffin's Market Diary -- Ignore unavailable to you. Want to Upgrade?


To: Triffin who wrote (83)2/6/2000 8:06:00 PM
From: Triffin  Respond to of 869
 
GG: HUNT: CREE

CREE: Competitive position and Tornado
Warning long post on CREE: part 1 of 2

First let me begin by saying that with the current database on CREE that I am very bullish on CREE so all of the following needs to be taken with a grain of salt.

Competitors:

There are competitors, links to which exist on the CREE thread. I tend to believe that CREE does indeed enjoy a lead on the competition, but don't have any independent information that suggests that lead is 6 years. I think that if CREE came up with the 6 year number by comparing where they were 6 years in the past to the current capabilities of the competitors, then they have over estimated their lead. A lot more information is available (including CREE's patents)today that help those competitors to better understand how to make SiC which compresses the development time.

It remains to be tested how much of a hurdle CREE's patents introduce in stretching that leadtime out or raising the competitor's costs of production. I've read many of CREE's patents and it is my belief that for growing commercially usable bulk SiC single crystals they have an effective barrier if it withstands a test in the court system.

I also think that they have patented the most cost effective processes fundamental to building devices in SiC. If these also withstand legal challenges, I think competitors will be forced to pay royalties for virtually all devices built in SiC or adopt a more expensive process technique.

In no way do I think that CREE's patent positions give them a lock on Blue LED's. This is CREE's predominant $ machine at the moment, so this is not trivial. OTOH it is comforting somewhat that CREE's hegemony in this market is the result of excellent execution leading to delivering what I believe to be the lowest cost/performance ratio product. I think their market leading position is likely to be secure for the next two years.

Relative to the blue LED's it is important to realize that CREE's products are composed of two materials: SiC and GaN (or GaInN). The biggest advantage SiC has is that it allows a good electrical contact to be made to the GaN that fits better to industry standard mounting techniques as detail by unclewest and others. This may turn out to be the Achilles heel. There is a lot of work taking place to find a way to use buffering layers to permit mounting GaN onto Si. The hurdles to doing this are huge, but IMO not insurmountable. If a method were to be found, the shift for blue LEDs to Si substrates would be extremely quick(maybe 1-2 quarters) since the wafers are readily available in 8 inches and undoubtedly much less expensive than SiC.

I think the above scenario is highly unlikely in the next two years but I'm not active in this area so don't take my word for it.

In the Blue LED arena I've also seen longer term threats coming from techniques to pattern Si or using something called porous silicon. I think these would be potentially cheaper than putting GaN on Si substrates. I don't understand the mechanism yet so I don't know if it is applicable to the blue end of the spectrum, but there are references to them "emitting light across the visible spectrum".

For these reasons I don't think that CREE becoming (if it were to do so)the Gorilla in the Blue LED space would offer the security that we expect to have with a gorilla. In short for this specific application (Blue LEDs) other discontinuous inventions may be too close for comfort.

Blue LEDS do appear to be the door to the bowling alley for all the other "pure" SiC device applications. Cree is gaining tremendous know-how in manufacturing and subsequently processing SiC. The combination of the other physical properties of SiC make it unlikely to be replaced(SiC is not yet established) in the high temperature, high power and to some extent high frequency applications.Indeed I think the only possible replacement material within the next 10 years would be AlN, further out maybe BN comes into play. Both materials are very difficult to make (far more so than SiC, I know this from personal experience.) The knowledge base on both materials is extremely limited and I am unaware of a semiconductor device ever having been made(I mean laboratory not commercial) out of BN. I don't see diamond competing because of the poor oxidation resistance. It could fulfill some application, but lower volume will likely relegate it to higher costs than if SiC could universally supply those applications.

I should point out that even if devices were today available in AlN and BN, it is not clear that they would have compelling advantages over SiC other than very high temperature oxidizing environments.

It is my belief that CREE has a very good chance to have their devices be qualified for some of these applications within the next two years (before threats hit on the LED front). A power/high frequency device was released last year that offers a discontinuous innovation that fits well in the burgeoning telecommunications market. Other devices are being sampled for design qualification.

Whether CREE is able to secure the beach in these areas will be shown in the next 18-24 months.

Tornado Status:

I think that relative to Blue Leds it is safe to assume CREE is in the tornado. They can not fulfill the demand for products that are consumer related (cell phones, games, vehicle lighting). Unfortunately, for me anyway, it is not a forgone conclusion that this will make CREE the gorilla of blue leds or white leds since displacing technologies may be on the horizon. Don't forget that the commercialization of blue LEDs is in its infancy with alternative technologies in place.

Because of the above, I judge the current growth to be welcome, but not proof positive that it is breeding the type of gorilla for which we are searching. Never the less, I am struggling to convince myself not to invest more $ until I see them solidifying a base in the non-LED device business.

I doubt I will ever be brave enough to make a Lindysque 100% commitment or even an unclewest 51%, but if CREE can demonstrate the beginnings of a tornado for the high power/ high frequency devices, I would feel comfortable to at least 3x my allotment.

Part two has a few details about the actual competitors.

---------------------------------------------------------

CREE Competitors and Value Chain, Part 2
Warning this is a relatively long post.Virtually all of this derives from web based information with very little independent confirmation. In many areas I am purely guessing and I've tried to note those instances with a question mark.

Also if after reading through this, you are thoroughly confused as to who is part of CREE's value chain and who is a competitor, then you have understood perfectly. The situation is indeed confusing since CREE often competes with with some of it's customers. I think the best we can say is that CREE is at a point where they need to insure the market is jump started, analogous to QCOM and the phone sets. I've seen nothing that indicates in which direction (devices or substrates) CREE intends to consolidate it's power. I prefer the devices, but they haven't asked me (yet!) for my opinion.

Secondly before delving into the competition it is good to fully understand the value proposition of SiC and CREE's competitive position.

The absolute single best tutorial on SiC semiconductor technology that I have found is this 142 slide presentation from mid 1998. About half of it is technical, but the other half has good graphics and easy to understand conclusions.

nina.ecse.rpi.edu

If you don't want to go through all of the slides at least do the following:

Read the first 35 or so

Slide 42 Grumman wafers up to 3 inches

Slide 49 Japanese institutions active in this area

Slide 53 A SiC diode tolerated a forward current density of 800 amps per sq. cm (that's impressive)

Slide 111 SiC timeline

Slide 133 Cree's latest (remember this is mid 1998) results

Slides 135+ Applications

Slide 141 Conclusions

As far as CREE's weaknesses, the only thing that I can find that is recent exists in progress reports from Cornell's work under an Office of Naval Research Contract. In short they have produced better/ more consistent devices on sapphire substrates than SiC. Cree is at least one of their vendors, but it is not unknown for companies to give junk to the universities. You have to wade through these reports (which are mostly technical) to find the pertinent comments. I've extracted several key ones(with some paraphrasing) in case you wish to bypass the reports at:

iiiv.cornell.edu

Due to poor quality of substrates these devices are not yet reproducibly good on SiC... by working with two or more vendors it is expected that those problems will be worked out in the next few months...

The 2 inch substrates from 1 vendor has visible defects even cracks at the surface, while another has an excellent surface but has high net donor density[It is not clear if these are from Cree or a competitor. The latter issue infers a quality control problem or that the vendor purposefully sent junk.]

Eleven of 16 wafers commercial 2 inch SiC wafer were returned to the vendor due to large defect density [Confirms the difficulty of making good SiC wafers]

This is the follow up to using multiple vendors after culling through the wafers to select the best SiC wafers from CREE and another:

The alternate vendor's 1 3/8 inch wafers are compared to CREE N+ wafers. Both show high interfacial charges relative sapphire I paraphrased the actual paragraph heavily to make it understandable. The Cornell folks have done a lot of fine tuning work to get the sapphire devices to perform. Originally some devices on SiC were better than those on sapphire. I think they focused more on sapphire since the substrates were more available. I think that if they apply the same effort to SiC substrates, they will get similar results ultimately. This summary is very recent so it is clearly still difficult to get good SiC wafers.

Now to the who's who! I've not included those firms which are resellers.

Substrate Competitors

Sterling/Epitronic/ATMI
TDI
SiCrystal AG
II-VI
HOYA
Novecon
Nippon Steel
Grumman?
Westinghouse?
Rhombic?
American Xtal Tech?
Aixtron? (indirectly)
Emcore? (indirectly)

Value Chain:Substrate Customers

TDI?
Grumman?
Ge/Emcore/Gelcore?
ABB
Vishay
Infineon (Seimens-Osram)
Diamler
Matsushita
Various Universities and Government Labs

Sterling etal
sterling-semiconductor.com

Sterling and Epitronics stem from a Mother Company called ATMI. While I don't know anybody personally at Sterling or Epitronics I do know the upper management at ATMI. I respect their professionalism so by transference I deem these two entities as posing a viable competitive threat to CREE. As best I can tell Sterling is selling substrates made by Epitronics. I think it is mostly SiC films deposited onto Si. Sterling may have had technology transferred from Epitronics to allow them to produce in house as well. It seems they mostly produce 35 mm wafers and will do 50 mm (2 inch)as special order. They claim less than 100 micropipe/cm2 defect levels. [Cree has achieved 0.8/cm2 densities but the commercial level is also on this order]

TDI

tdii.com

Their site is very informative with nice graphics so I encourage a visit.

They claim to have a new process that improves the quality of SiC wafers by essentially patching the defects and then growing a much improved layer on top. They are able to deliver 35-41 mm wafers with defect densities of <15/cm2. I'm assuming that they likely start with CREE wafers.

SiCrystal AG

sicrystal.de

This German firm offers SiC wafers up to 50 mm with defect densities of 2x those of CREE. A wafer currently costs slightly over $1000 (vs. Cree's $495)

II-VI

ii-vi.com

Two Six is a mini conglomerate that recently has started to claim the ability to produce SiC wafer substrates. They originally projected being able to deliver wafers up to 50 mm by the end of 1999. It doesn't appear to me that they are really in production yet. Although they seem to be behind the power curve here, I wouldn't discount them. In the distant past I was aware of programs in the optics regime where they were able to set the standards. Some of the processes used are applicable to growing SiC layers.

Hoya

hoya.co.jp

Now here is something intriguing. They claimed to be able to supply SiC on 150mm (6 inch) Si wafers. They used to have a site where you could sign up to get free wafers for researching. About a week ago when I decided to apply for a free sample, the site disappeared. In fact I've noticed that most of my links to japanese sites on this topic have disappeared. Coincidence? I don't think so. We'll need to watch the Japanese collective closely. The only additional info on Hoya that I can offer is dated 1994:

itri.loyola.edu

Nippon Steel

I have no link to a Nippon Steel site but this visit report from nearly two years ago states that they had one of the most advanced SiC wafer programs:

ttec.org

Japan General

This is old but indicates that they are not sleeping over in Japan:

atip.or.jp

Novecon

novecon.com

These folks will supply SiC wafers from Russia. The irregularly shaped (high quality) Lely substrates are not competitive since there is no way to commercially process these to make product. They are only useful for research. Otherwise, they can deliver 1 inch wafers with typical defect specs.

They also mention various patents here which gives an Idea of potential competitors/customers to CREE:

novecon.com

I wish to point out that much of the early work on SiC semiconductors was done by Russian Scientists. Indeed Cree seems to have a good relationship with many and has coauthored papers and funded efforts. Check these out:

ehis.navy.mil
ru.com

If I remember correctly unclewest or somebody on this thread was involved in military intelligence and can probably comment of the Russian's capability. I personally have always respected them for having done so much with so little. Indeed I've always thought that since they had so little resources they were forced to better think the problem through before beginning to attack it. These folks are probably a very powerful resource to CREE.

Grumman

sensor.northgrum.com

I really don't have much on them except to say that the Cornell slide show indicated that they had produced a 3 inch substrate by mid 98. I don't think they are selling anything commercially. I also believe the focus is mostly military and doesn't appear to be a hot area for them since none of their open job positions are focused in the SiC area. No SiC based products are listed on the web site. They seem to be mostly device focused so maybe they are a CREE customer, but I've also seen references to their ability to deposit epitaxial SiC.

Westinghouse

I've mostly only seen them referenced in reports like these:

nctn.hq.nasa.gov

strategies-u.com

Generally I would say that Westinghouse is spectacularly unsuccessful in commercializing semiconductor devices over the last 20 years or so. These announcements refer to a device launched several years ago, but I can't even find a Westinghouse web site to investigate it further. If somebody knows of their site address please post it.

Rhombic

I list them with some trepidation since the web site content seems to have crossed into the realm of "hype". You can judge for yourself. They seem to think they have a technology that can improve the quality of compound semiconductors, but appear to have not progressed beyond the concept phase (indeed their website almost seems to be seeking money and probably exceeds what the SEC would allow).

rhombic.com

American Xtal Technology

axt.com

I've seen references to them announcing last year having grown their first SiC crystals, but there is nothing about it on the web site other than the comment buried in this news release:

axt.com

They also claim to be involved in the blue led market by supplying GaN and SiC substrates.

Emcore and Aixtron

emcore.com

aixtron.com

I lumped these two together since they supply machines (and these days that usually means an associated process to run on the systems)to deposit layers like GaN or SiC. They both have partnerships, but the Emcore/GE Gelcore partnerships is potentially the greater threat to CREE from the device standpoint. I'm not able to determine if they usually use CREE wafers as a substrate, but would think they try to deliver a process which is substrate vendor neutral. I would also assume they deliver processes to their customers that circumvent CREE's patents. Usually purchasers of such systems have the equipment vendor indemnify them from patent violations.

Customers

I assume these firms are likely substrate customers which means they are device competitors. I'm not including the already mentioned ones above.

GE/Emcore, Gelcore

gelcore.com

This joint venture is focused on the lighting potential.It is not clear whether they intend to use sapphire or SiC or both. My guess is that if adequate quantity and quality of SiC is available they will use it because of the packaging advantages. GE can also bring tremendous material processing technology to this JV. Even if they are today a SiC customer one should not discount their ability to produce it in the future.

Additionally, I have seen GE referenced numerous times as having made advancements in SiC devices. This is the only GE site I've found about SiC devices:

crd.ge.com

ABB
abb.com

Although ABB is probably best know for power plants, one of the B's stands for Brown which is/was a leading power semiconductor manufacturer. Consequently their interest in SiC is understandable. I don't see them really trying to compete in the substrate arena although I did see a reference to them successfully growing SiC layers.

Vishay

vishay.de

Clearly states the device is on SiC. Can only assume it may be from Cree.

Infineon formerly Seimens/Osram

infineon.com

You'll have to burrow down through their site. Start with products and then to optoelectronics of LEDS and then finally blue Leds (if I remember correctly).

They have had a strong relationship with CREE over the years. I also know that on the materials research side they have ongoing programs with SiC, GaN (+alloys) and AlN. I wouldn't be surprised to find out they have licensed some of Cree's technology. They also led the charge in having the blue/white leds used for cars.

Matsushita

I don't have a direct link to Matsushita, but I have this reference:

ttec.org

I think it is also interesting to realize that Panasonic is a Matsushita brand and they are reportedly buying the blue/white leds from CREE.

I'm going to stop here. That really doesn't cover 100% of the customers (or probably competitors)since I didn't mention pure end customers like Nokia nor the government labs or universities. Nevertheless I think it is clear that CREE does not (yet anyway) have the world to themselves. They still need to execute to get into an Intel like position. I don't think they can adopt the Qcom model without taking on a greater amount of risk. It wouldn't be bad to maintain the position of predominant wafer supplier as long as possible while they try to capture as much of the device market as possible.

I hope this helps better understand CREE's competitive position.

Fatboy

EOM -------------------------------------------------------



To: Triffin who wrote (83)3/9/2000 7:17:00 PM
From: Triffin  Respond to of 869
 
GG: WIND PRODUCT OFFERINGS

Wind River Networks: Transforming the Way Networking Products are Developed

From the devices that attach to the edge of the network, to the servers that provide content and applications, and all the infrastructure plumbing in between, embedded software is critical. Wind River is a leading supplier of the underlying embedded technology that runs many of these smart, connected devices, particularly for the Internet infrastructure and intelligent server input/output (I/O) industry segments.

Market competition is driving networking equipment manufacturers not only to cut costs and reduce development time but also to dramatically increase functionality and performance levels. Wind River?s Networks business unit was formed specifically to help manufacturers accelerate the development of network infrastructure and high performance content server platforms by supplying a higher percentage of infrastructure and content server solutions.

The networking business unit offers an ever-increasing array of software building blocks for constructing both infrastructure devices and intelligent I/O applications for servers. Integrated packages from Wind River Networks that contain the common components for a particular device type are made even more valuable as they are already optimized for Wind River?s VxWorks© RTOS. These targeted solutions enable original equipment manufacturers (OEMs) to focus on product differentiation.

Internet Infrastructure
The networking market is predicted to have reached the $36.8 billion mark when 1999 totals are calculated (Cahners In-Stat). This is attributed to the roll-out of broadband services to business and residential users and the resulting new networking applications and services that it creates. These, in turn, will drive investment in the carrier core. However, qualified engineering resources are scarce and time spent building common checklist items detracts from a vendor?s ability to adopt new, emerging technologies.

In addition, most of today?s networking infrastructure equipment is built using application-specific integrated circuits (ASICs), in-house implementations of Internet standards and proprietary management applications. ASIC chips take up to 18 months to develop and cost between $200 and $400 per chip (The Microprocessor Report). For that reason, networking OEMs are beginning to create extendable network device platforms that can support higher levels of product differentiation than could be achieved previously with ASICs.

To accomplish that, OEMs are using network processors, which can be reprogrammed, in conjunction with integrated, modular commercial off the shelf (COTS) network device software components such as the network protocols and management applications from Wind River.

Today, Wind River offers over 40 key networking protocol implementations with partner support for many more. In addition, Wind River Networks offers packaged solutions like Tornado for Managed Switches (TMS), which includes the RTOS, drivers for specific target network processors, and the core network protocols and management capabilities required in an Ethernet switch or router. With this base platform, OEMs can quickly and easily build complex networking equipment.

Key Customers

Cisco
Lucent
Nortel
Redback Networks
Foundry Networks
Technology Partners

Intel
MMC
Allayer
Trillium
Harris and Jeffries
Rapid Logic
Standards

IEEE
IETF

Intelligent I/O
Today?s PC servers are powered by CPUs boasting unprecedented levels of performance. However, these powerful servers are hampered by current PC I/O architectures, which use the main processor to handle I/O events and severely degrade system performance. Key factors in the Internet explosion are the need to provide access to huge amounts of content, the move to e-commerce and other Internet based services. Storing and providing that content and many new types of on-line applications and services are PC servers.

As Internet-based connectivity explodes, there is greater demand for I/O bandwidth. New technologies are required to improve overall system performance and to provide more scalable I/O capacity. Both of these goals can be achieved with a concept called intelligent I/O, which offloads portions of the server I/O functions to a dedicated I/O processor (IOP).

A dedicated IOP improves overall system performance of the PC server by reducing the load on the main CPU and by moving intelligence closer to the data. Scaling is made possible by adding IOPs as I/O needs change, which system vendors offer their customers as expandable I/O options.

To help make all this possible, Wind River is a major contributor to an industry initiative called I2O. This movement, in which Intel is another major contributor, has produced a specification that defines an architecture for intelligent I/O and related technologies such as InfiniBand. Wind River Networks offers key components that deliver on the promises of I2O, including the only I2O real time operating system (IRTOS), the IxWorks? IRTOS. System vendors and I/O subsystem manufacturers use an operating system and a toolkit developed by Wind River that is optimized for intelligent I/O applications. Together, this software helps networking equipment vendors create products that accelerate I/O functions in storage and network access.

Why Wind River?

Value to the Market
Shorter time to market
Increased development productivity
Worldwide engineering and technical support
Scalability
Partnerships
Value to the End User
Leveraging reuse helps keep costs down
Access to the latest technology
Reliability
Ease of use
Technical Advantages
Reliability and proven performance
Modular approach
Powerful development tools
Off-the-shelf run-time components
Support for industry standards
Portability
Products
Wind River's proven networking solutions, allow customers to pick the best components for their unique needs. With Wind River's core embedded technology as a foundation, companies can leverage their own design expertise to create leading-edge networking and Intelligent I/O systems faster and more powerful than ever before.

Tornado? and VxWorks©
Wind River's Tornado II embedded software development environment is the most powerful and complete platform for building real-time embedded applications and devices. It is comprised of three highly integrated components: Tornado Tools, the only embedded development tools that include a target hardware simulator for immediate product development; the VxWorks RTOS; and a full range of communications options for target connection to the host. VxWorks RTOS is proven in millions of applications today and is the de facto standard operating system for data and telecommunications products.
VxWorks Network Stack
VxWorks Network Stack is a complete, robust TCP/IP stack that includes features such as Dynamic Host configuration Protocol (DHCP) that allows a networked device to dynamically acquire an IP address.
Embedded Networking Protocols
Wind River offers over 40 protocols in source code form for routing, bridging/switching, VPN, WAN and legacy LAN access.
Epilogue? SNMP
The Simple Network Management Protocol Version 3 agent includes Master/Sub agent support and a MIB compiler for networked devices.
Wind© Web Server
This server is scalable from 4 to 20 KB and includes Server Side Push that links the web page to the embedded application so that an IMC device can populate information dynamically.
Tornado for Managed Switches
With TMS Wind River introduced the industry?s first software solution for developing Layer 2 and 3 switch and router based network devices. It is comprised of a set of highly integrated components: Tornado Tools, a comprehensive embedded development environment; the VxWorks RTOS; a complete set of Layer 2 and Layer 3 switching and routing protocols; and the management services required for enterprise class network devices.
RouterWare Embedded Protocols
RouterWare embedded networking protocols provide comprehensive, efficient software implementations of Internet and international network standards to address critical time-to-market needs. The RouterWare products include 40+ protocols, delivered in source code form, grouped into five families: TCP/IP, IPX, bridging/switching, VPN/tunneling, and WAN data link.
Tornado for Intelligent I/O
Tornado for Intelligent I/O provides the tools customers need to build intelligent I/O platforms, right out of the box. By combining Tornado for Intelligent I/O with Intel?s family of fully integrated I/O processors, customers have a complete intelligent I/O solution.
WindNet?
WindNet is an advanced series of networking products. These products extend the networking features of Tornado and enhance the networking capabilities of embedded microprocessors.
WindNet OSPF
WindNet OSPF is an off-the-shelf implementation of the OSPFv2 routing protocol and algorithm. The product?s level of modularity and integration enables developers to merely drop the module into their design.
WindNet PPP
WindNet PPP provides a robust implementation of standard Point-to-Point Protocol (PPP) functions delivered on a versatile remote access framework. This provides an ideal platform to add functional extensions and create advanced embedded PPP solutions.
WindNet SNMP
WindNet SNMPv1/v2c release 2.0 agent brings industry-standard network management into the world of embedded real-time systems. Its advanced and extensibile design makes it easy to add remote manageability to networked devices.
WindNet STREAMS
WindNet STREAMS provides an industry-standard platform for integrating a wide variety of networking protocols into an application. Numerous protocol modules have been developed for the STREAMS framework. These modules can be directly leveraged for use in WindNet STREAMS, greatly accelerating the time-to-market for networked embedded applications.
iVasion? Products:
WinPoET?
WinPoET, the industry's first PPP over Ethernet client for Windows, provides end user access to high-speed broadband networks using the Microsoft dial-up user interface. Compatibility with existing PC hardware and software, all xDSL modems and use of Ethernet NICs lowers provisioning costs.
WinPoET SDK
With the SDK you get all the pieces you need to create your own PPP-over-Ethernet client. You can change the launch and system tray icons, the look and feel of the connection dialogs, and the information presented in the system tray applet.
WinVPN? Client
WinVPN? Client provides remote users with low-cost access to LAN resources through Internet based Virtual Private Networks (VPNs). WinVPN Clients create tunneled, authenticated connections across the Internet that can be optionally encrypted. Because remote users dial in through their local ISP, long distance telephone charges are sharply reduced or eliminated.
Envoy SNMP Agent
Envoy is a highly portable memory-efficient implementation of the SNMP protocol. Envoy is specifically designed for OEMs and system integrators that require a fast, small agent or the basis for a management station.
MIBway for Rapid Control
Nearly all network devices have SNMP capability. With MIBway for RapidControl, developers can add Web-based management to SNMP-managed devices without throwing away the work already done for creating, testing and refining their SNMP management information bases.
RapidControl for Applets
Rapid Control for Applets allows device vendors to build dynamic, real-time Java-based management interfaces to a network device using standard JavaBean-based applications. RapidControl for Applets requires no embedded Java Virtual Machine (JVM) and no Java programming expertise.
RapidControl for CLI
Today, every networking device design process includes creating a command-line interface (CLI). RapidControl for CLI does most of the work so that engineering can spend more time on the critical product engineering issues.
RapidControl for Web
A properly architected Web-based management system improves ease of use, reduces cost of support, adds universal accessibility, and increases brand awareness. RapidControl for Web builds powerful Web-based management systems for embedded network devices quickly and cost-effectively.

EOM ----------------------------------------------------



To: Triffin who wrote (83)3/13/2000 9:21:00 PM
From: Triffin  Read Replies (1) | Respond to of 869
 
GG: HUNT: ELON

To: Uncle Frank who started this subject
From: janet_wij Monday, Mar 13, 2000 12:20 AM ET
Reply # of 20091

Echelon Corporation (ELON) Project Hunt Report
1. Company Overview

Echelon Corporation develops, markets and supports a family of hardware and software products and services that enables original equipment manufacturers (&#8220;OEMs&#8221;) and systems integrators to design and implement open, interoperable, distributed control networks. The company offers its products and services to OEMs and systems integrators in the building, industrial, transportation, home and other automation markets. The company provides a variety of technical training courses related to its products and underlying technology as well as customer support to its customers on a per incident or annual contract basis.

The Company derives its revenues primarily from the sale and licensing of its products and, to a lesser extent, from fees associated with training and technical support offered to its customers. Product revenues consist of revenues from sales of transceivers, control modules, routers, network interface devices and development tools and from licenses for network services software products. Revenues from software licensing arrangements have not been significant to date. Service revenues consist of product support (including software post-contract support services) and training. The Company recognizes revenue from product sales at the time of shipment to the customer.

2. Revenues

Total revenues grew to $9.8 million in the third quarter of 1999 from $7.1 million in the third quarter of 1998. Total revenues for the nine months ended September 30, 1999 grew to $28.4 million from $23.6 million in the same period in 1998. One customer, EBV, the sole independent distributor of the Company's products in Europe since December 1997, accounted for 29.0% and 23.7% of total revenues for the quarters ended September 30, 1999 and 1998, respectively, and 27.5% and 22.5% of total revenues for the nine months ended September 30, 1999 and 1998, respectively.

Product. Product revenues grew to $9.3 million in the third quarter of 1999 from $6.4 million in the third quarter of 1998. Product revenues for the nine months ended September 30, 1999 grew to $26.7 million from $21.3 million in the same period of 1998. The 45.2% increase in product revenues between the two quarters and the 25.4% increase in product revenues between the two nine-month periods were primarily a result of an increase in sales of control and connectivity products, and to a lesser extent, an increase in sales of LonPoint products and network services products, partially offset by the decrease in sales for development tools products.
Service. Service revenues decreased to $473,000 in the third quarter of 1999 from $717,000 in the third quarter of 1998. Service revenues for the nine months ended September 30, 1999 decreased to $1.7 million from $2.4 million in the same period of 1998. The 34.0% decrease in service revenues between the two quarters and the 28.0% decrease between the two nine-month periods were primarily due to reduced customer support revenues partially offset by a slight increase in training revenue.

12-month sales growth as of:

Dec. &#8216;98 13.8%
Mar &#8217;99 14.8%
June &#8217;99 13.0%
Sept &#8217;99 18.7%
Dec &#8217;99 23.5%

Gross margins have grown over the last 8 quarters and in Sept. &#8217;99 stood at 60.5%.

3. Neuron Chips & Dependence on Key Manufacturers

The Neuron Chip is an important component used by the Company's customers in control network nodes. In addition, the Neuron Chip is an important device used in many of the Company's products. Neuron Chips are currently manufactured and distributed by both Motorola, Inc. ("Motorola") and Toshiba Corporation ("Toshiba"). The Company has entered into licensing agreements with each of Motorola, Toshiba and Cypress Semiconductor Corporation ("Cypress"). The agreements, among other things, grant Motorola, Toshiba and Cypress the worldwide right to manufacture and distribute Neuron Chips using technology licensed from the Company and require the Company to provide support and unspecified updates to the licensed technology over the terms of the agreements. The Cypress agreement expires in April 2009, unless renewed. The Toshiba agreement expires in January 2001, unless renewed. The Motorola agreement expires in January 2001, and Motorola has announced that it will discontinue distribution of Neuron Chips after January 31, 2001.

4. Market Acceptance of Interoperability
The future operating success of the Company will depend, in significant part, on the successful development of interoperable products by the Company and OEMs, and the acceptance of interoperable products by systems integrators and end-users. When products or subsystems from multiple vendors can be integrated into a control system without the need to develop custom hardware or software, they are "interoperable." The Company has expended considerable resources to develop, market and sell interoperable products, and has made such products a cornerstone of its sales and marketing strategy. The Company has widely promoted interoperable products as offering benefits such as lower life-cycle costs and improved flexibility to owners and users of control networks.

5. G/K Characteristics

A. Is there a discontinuous innovation or a proprietary open architecture?

Yes. LonWorks is a discontinuous innovation whose technology is open and proprietary. The most dramatic market successes are (or began as) open, proprietary technologies. Their proprietary nature offers an economic rationale for their existence, and their open availability makes them accessible to all, with no advantage accruing to any particular user. Examples include DOS Windows, Unix, Ethernet. LonWorks is an open technology. Its enabling component (the Neuron chip) is manufactured and sold worldwide by two competing semiconductor giants, Motorola (now Cypress) and Toshiba.

What does the technology encompass?
This is perhaps the least understood (and appreciated!) aspect of the LonWorks story. Many who are only casually familiar with the technology tend to view it with a very narrow perspective: perhaps that of a protocol definition or a power line communication transceiver, or a specialized chip for control systems, or whatever. Such constrained views miss the whole point of the technology, and certainly the principal reason for its success in the marketplace. For it is all of these things, and much, much more.
Designing, constructing, installing, and maintaining a control network requires attention to a great many issues, including: an appropriate communication protocol, physical layer transceivers for the selected medium (or media, in many cases), capable microprocessors for both application and protocol execution, device controller hardware & software for the required sensors and actuators, timing hardware, RAM and ROM storage, non-volatile parameter storage (such as EEPROM), an appropriate programming language model & a compiler/debugger for same, run-time libraries, a real-time distributed operating system, an interoperability solution, physical node identification (for installation, tracking, etc.), network management software, on-line diagnostics, protocol analyzer, host application program interfaces, installation and configuration tools, etc., etc.
LonWorks technology offers a ready-made, highly integrated solution to *all* of the above concerns, much of it in the form of a single, low-cost silicon component from two of the world's largest semiconductor companies. Available today. On the chip. Off the shelf.

How extensive is the LonTalk protocol?
Protocols today are generally designed around--and described by--the ISO standard "Open Systems Interconnection Reference Model", which encompasses a full set of protocol features, and classifies them according to seven functional categories (referred to as "layers"). Thus the "seven layer OSI model".
The LonTalk protocol implements the entire seven layers of the OSI model, and does so using a mixture of hardware and firmware on a silicon chip, thus precluding any possibility of accidental (or intentional!) modification. Included are not only such expected features as media access, transaction acknowledgment, and peer-to-peer communication, but also more advanced services such as sender authentication, priority transmissions, duplicate message detection, collision avoidance, automatic retries, mixed data rates, client-server support, foreign frame transmission, data type standardization and identification, unicast/multicast/broadcast addressing, mixed media support, and error detection & recovery. For an overview of the LonTalk design, and the benefits of fully functional protocols, refer to "LonTalk Protocol Rationale", available from Echelon Corporation. The full LonTalk Protocol Specification is also available; due to its size, a charge of $50 applies.

How is LonTalk an Open Protocol?
The market for control networking has only recently evolved from one of early adopters and experimenters to maturity. Everyday, developers are finding more uses for the protocol, yet at the same time are inventing uses never envisioned by the people that developed the dedicated silicon upon which all current LonWorks networks are based, the Neuron Chip. Therefore, Echelon has opened up the LonTalk protocol to allow any company to port it to the processor of their choice. This means that applications requiring 16 or 32 bit processing power, can now host the protocol in native mode and eliminate the need for a microprocessor interface program.
The protocol is also currently in committee review by the Electronics Industry Association for possible recommendation as a standard for home automation. Additionally, the protocol is part of the American Society of Heating, Refrigeration, and Air-Conditioning Engineers' BACnet control standard for buildings. This is now known as ANSI/ASHRAE 135-1995.
Although it is possible to implement the LonTalk protocol on a generic processor architecture, a silicon realization (i.e., the Neuron chip) is the best approach for most control applications. This is for several reasons:
&#61623; The Neuron is actually 3 8-bit inline processors in one. Two are optimized for executing the protocol, leaving the third for the node's application. This ensures that the complexity of the application does not negatively impact network responsiveness and vice versa.
&#61623; Use of the Neuron chip guarantees an appropriate hardware execution environment for the protocol. If there is a downside to a full-featured, seven-layer control protocol, it is the need for ample computational power to execute it. The protocol is thus implemented with a mixture of hardware and firmware (the instructional set of the Neuron processors was designed specifically to execute the LonTalk protocol, allowing the firmware portion to occupy a mere 8K bytes of on-chip ROM).
&#61623; The creation of a custom silicon device also allows the inclusion of additional functionality to facilitate control node design. The Neuron chip, for example, incorporates watchdog timers, on-board diagnostics, 35 device controller types, a distributed real-time operating system, run-time libraries, three types of memory, and even a 48-bit software-accessible serial number (which, guaranteed by the chip's manufacturers to be unique, provides an always-available installation address for any Neuron chip-based node).

intellicom.se

In summary, LonWorks technology is a commodity solution to the many problems of designing, building, installing and maintaining control networks, of all sizes, in all industries. The Neuron chip is the enabling technology for, and the core component of, each and every LonWorks node.

Jack Hicken, editorial director of Instrumentation & Control Systems, wrote, &#8220;My gut feeling is that LonWorks will do for control networks what the microprocessor did for computers.&#8221; (My emphasis) Like the microprocessor, LonWorks technology exploits the very high volume production of a highly generic and massively integrated solution. (FAQS)

The principal suppliers of LonWorks technology are Echelon Corporation, (creator of the technology, and provider of transceivers, connectivity products, development tools, and training), Motorola (until January 1, 2001), Toshiba and Cypress Semiconductor (Neuron chips) (from FAQ)

Over 2000 corporations use LonWorks technology today, and their numbers are growing rapidly. Because many of them view their use of the technology as strategic, they are averse to any public reference.

B. Is there a Discontinuous Innovation?

Yes. In the building controls industry, two enabling technologies have converged, the Internet and LonWorks. LonWorks Network Services (LNS) seamlessly ties these technologies together. LNS is an Internet ready, out of box solution, with no need for custom engineering where it is possible to install, configure, monitor and control a system all on LNS. LNS is a real, open, multi-vendor and multi-product solution that enables a system to.

1. Speak peer to peer, a node from manufacturer A can speak to a node from manufacturer B.
2. Possible to use a common tool to configure.
3. Can substitute a node from manufacturer A to manufacturer B.

This integration allows a single user interface, lower failure, competitive prices, easier and faster specification and design and very, very low technology risk.

C. Does it have the potential to grow into a mass-market phenomenon, become a standard?

Yes. There have been myriad examples of Echelon wins cited on this thread, including the adoption of LonWorks as an ANSI standard, and the production of the iLON server. I will concentrate on the building controls industry and the comparative adoption of BACnet and Echelon as standards, and how Echelon is winning market share in this space.

BACnet is a committee-created ASHRAE standard, and has been described as a framework for an open system. Their stated purpose is &#8220;to define data communications protocols for monitoring and control.&#8221; There is a 501-page book of BACnet specifications; however, since every engineer designs systems differently interpretation becomes problematic. According to an Echelon claim, &#8220;You don&#8217;t see a lot of BACnet product,&#8221; because of conformance procedures.

They contrast the number of control nodes installed that are accessible through the two different protocols:

Echelon LNS=Millions
BACnet=Thousands

The real test then becomes the degree to which these two protocols provide interoperability. LONMARK took a different path to insure seamless interoperability. Members of the LonWorks Interoperability Association decided on technology with the neuron chip embedded, and where the Lonsoft protocol is an open, published device specification which has become a de facto standard supported by shipping real products.

D. Are there high barriers to entry and high switching costs?

Yes to both. Echelon has created what seems like insuperable barriers to entry. A company would first have to create a competing discontinuous innovation. Then they&#8217;d need to work at getting the buy-in from the disparate group of manufacturers who make up Echelon&#8217;s value chain. Something Echelon has been working on for the past ten years. Echelon has a robust trade group in the LONMARK Interoperability Association. According to a pdf document on Echelon&#8217;s site entitled &#8220;The Open Control Technology Standard for Buildings,&#8221; there are:

&#61623; Over 3,500 OEMS with Development Tools designing products with LonWorks
&#61623; Over 1,400+ products now available&#8212;commercial, residential, industrial
&#61623; Millions of nodes installed
&#61623; Over 240 members in LONMARK Interoperability Association
&#61623; Vendors representing 57% Share of the Buildings Control Systems Market at ASHRAE &#8217;99 with shipping products

E. Have value chains developed?

Echelon&#8217;s LONMARK sponsor members pay user fees to carry its logo, and they are a veritable who&#8217;s who in the controls industry. Some of these LONMARK sponsor members are Carrier, Honeywell, Johnson Controls, and SIEBE, thus enabling Echelon to leverage its standards. I would say a gold-plated value chain has developed in the building control industry. As an aside, the relationships they have developed across industry lines and with competitors within various industries would seem to preclude a takeover. Echelon is the arms merchant to all, but does not compete in the core businesses of most of its LONMARKS sponsors and partners.

F. Have they crossed the chasm?

Yes, and the head pin is building automation. In addition to the 57% of non-residential building controls companies shipping Echelon enabled products, they also claim 240 LONMARK Interoperability Association members from 17 Countries, or 82% of the building controls market.

G. Existence of Hypergrowth?

From December 5, 1990 to December 31, 1998, shipped 6.95 million nodes.
From December 1998 to June 1999 shipped 1.5 million nodes.

In the words of the gentlemen (who did not introduce himself by name) presenting on the video, &#8220;That&#8217;s the kind of growth you see in a platform that&#8217;s beginning to catch fire.&#8221; Sounds to me he&#8217;s talking about hypergrowth, however, if we use as our measure Ken Oshman&#8217;s goal of billions of neuron chips we&#8217;re not there yet. But the growth over the past year may be characterized as micro-hypergrowth. They shipped those 6.95 million nodes over an eight-year period, and were on track to ship 3 million in 1999. Allowing for a modest ramp up in shipments they could achieve 100% year-over-year growth in shipments. Also, their iLon server began shipping this quarter. How that will impact revenue will be telling, since they state in their 10Q that &#8220;the company recognizes revenue from product at the time of shipment to the customers.&#8221; 100%+year-over-year revenue growth is quite likely.

6. Summary and Analysis

Echelon has crossed the chasm as defined in the RFM at page 29: &#8220;The chasm is a hiatus in market development during which there is no natural customer for the discontinuous innovation.&#8221; Decidedly, Echelon has plenty customers.

Echelon has knocked down the building automation pin in the bowling alley, and they have the &#8220;capability to support a new value chain&#8217;s formation&#8230;&#8221; (p33 RFM) What we are witnessing with Echelon is a new marketplace coming into being. Also at page 33 of the RFM: &#8220;The bowling alley stage is thus the earliest moment that a high-tech company can truly define itself as a going concern, because for the first time it has committed customers and a protected marketplace to support it in the years to come&#8230;.&#8221; And for us Gorilla Gamers, &#8220;it is also the earliest investable moment in the gorilla game.&#8221;

Home networking is another story. Here, Echelon is currently crossing the chasm, and, of necessity, slowly I might add. Their progress here is absolutely dependent upon, and must await, permanent always-on broadband connections into the home. The most recent statistics show 1.9 million households having broadband access as of 1999. The Strategis Group forecasts more than 25 million high-speed households by 2004. devices.internet.com As broadband rolls out, Echelon is optimally situated with products and partners to capitalize on home networking. According to the RFM &#8220;each new niche creates a moment of micro-hypergrowth&#8230;. The reason niche markets are called the &#8216;bowling alley&#8217; is that ongoing market growth will be niche to niche, each entry into a new niche being eased by leveraging the prior niche&#8217;s solution set and reference base of customers&#8230;.&#8221;

Echelon is poised to play a significant role in the home networking arena. Their iLON server which currently sells for $1200 is expected to be sold in future at price points acceptable to consumers on Main Street. Meanwhile, witness Echelon&#8217;s ties to manufacturers of consumer electronics, white goods, home lighting and security for the home. They are well positioned. It is still early in this game, and I expect their home networking growth will mirror the explosive growth of broadband. Howardroark in a posting on the Motley Fool board reported that Echelon expects volume product shipments by 2001 in the home networking market, and that Ken Oshman feels &#8220;pretty good&#8221; about LonWorks becoming the leading standard for home networking. In the meantime, manufacturers are already making and shipping LonWorks enabled appliances.

My conclusion: Echelon has already crossed the chasm in building automation, and I&#8217;m awaiting tornado formation in the near term. This part of their business is in &#8220;micro-hypergrowth&#8221; as defined by the RFM. Echelon is not yet a gorilla, but on the strength of the building automation market alone, I would say this company is a &#8220;YPGWA,&#8221; young potential gorilla with attitude, whose near-term execution and earnings must be closely monitored for any signs of tornado-like activity. Analysts are expecting profitability in Q2. Echelon bears very close watching.

Disclosure: I am long this stock.

7. References
echelon.com Echelon's home page
lonworks.echelon.com Echelon Lonworks page
biz.yahoo.com SEC 10-Q
echelon.com Open Control Tech Standards for Bldgs.
intellicom.se„ssor
lonmark.org
westermo.se
cesweb.org
biz.yahoo.com
ASHRAE Journal, April 1999, article entitled "Specifying Interoperability," Paul Ehrlich, P.E. and Ofer Pittel,P.Eng.
Posts #2824 and 2827 from The Motley Food from howardroark and fenway 78 respectively re January&#8217;s Conference Call.

EOM ------------------------------------------------------

To: unclewest who wrote (19762)
From: 100cfm Saturday, Mar 11, 2000 9:57 AM ET
Reply # of 20091

Elon
Had an interesting conversation re Lonworks with my Trane Rep yesterday afternoon.

1. Trane is abandonning their proprietary automation system known as Tracer. The new system rolling out will be known as Summit which is Lon compatible.

2. All Trane products and devices will have the Lon chip/board in them. The Summit control panel will have terminals that will accept other Lon systems to be tied into it ie: alarm systems, lighting systems and sprinkler systems.

3. Bacnet seems to be relagated to the function of communicating between different systems ie: Honeywell, Johnson and etc. In short Lon will be on the terminal and device level and Bacnet will be on the system level.

4. Carrier is also moving over to Lon from their proprietary system known as CCN according to my Trane Rep.

5. Make no mistake that these are definning moments. Trane and Carrier account for the Lion's share of all commercial/industrial HVAC equipment installed worldwide.
Since Lon's main thrust right now is building automation, this is huge. The third and fourth tier players McQuay and York are sure to be following close behind. Will look into it on Monday.

6. Lon is in the early adoptors hands right now. I say this because when I asked mr Trane is his equipment being specified with Lon yet the answer was no. That is consistent with my findings, I have not seen LOn actually specified in any specifications I have read and I read close to a dozen a week.

7. It seems the discription of the adoption cycle in the FM is right on with Lon. The early adoptors are jumping on board but the pragmatists( the engineers and architects) are still not sure. Once the pragmatists feel comfortable with Lon they will start including it into their specifications and then the flood gates will open up or should I say the wind velometer will start whirring.

8. Trane is holding a Lon interoperability seminar the week of the 20th for engineers, building owners and architects. I will be attending for any technical info I can pass on but mainly to take the pulse of the pragmatist who will be attending. I will gauge their sentiments before and after the seminar and see if any have a timeline for incorporating Lon into thier specifications.

In sum, I will go back to what Merlin said about when the right time to invest pre tornado is, "when the news makes it obvious that the tornado will form"(sorry mike if thats not exactly how you said it). To me the fact that Carrier and Trane are abandonning their proprietary systems that were so precious to them and incorporating Lon into every device and piece of equipment makes it obvious to me that a tornado will form without doubt. I immediately doubled my position in Elon when he left.
One last thought in re to the various discussions of what the number of neuron chips and id nodes that will be sold. When you think that any decent sized building will have upwards of a thousand Lon nodes it's very easy to see that a millions of nodes per day can be sold worldwide.

EOM ------------------------------------------------------

To: Uncle Frank who started this subject
From: Magic212 Monday, Mar 13, 2000 8:53 AM ET
Reply # of 20091

Leading Swedish Telecom Selects Echelon's LONWORKS Networks For Connected Homes
Monday March 13, 8:28 am Eastern Time

Company Press Release

biz.yahoo.com

STOCKHOLM, Sweden--(BUSINESS WIRE)--March 13, 2000--Echelon Corporation (Nasdaq:ELON - news) the leader in
networking everyday devices, today announced that Telia AB, the leading Swedish telecom company, has chosen
Echelon's LONWORKS© system as a standard for networking everyday devices in Telia's Connected Home initiative.

Telia's Connected Home initiative involves collaborating on the development of network-delivered services via networked
devices in the home. The initiative will involve typical products and services for the home including networked kitchen
appliances, interactive entertainment center, home control and security systems, and home offices. Telia is currently
working to roll out the concept to a small collection of homes in Europe and will conduct initial field trials in select areas of
Scandinavia during the year 2000.

``After extensive evaluation, we concluded that Echelon's LONWORKS system is the leading technology for networking
everyday devices in the home,' said Leif Bengtsson, vice president of Telia Research. ``Echelon's LonWorks networks
provide the open, interoperable platform we need to be able to interact with everyday devices in the home and provide
valuable new services through both broadband, power wiring and wireless communication.'

In a demonstration being shown this week at Computer World Expo, attendees can use a Wireless Application Protocol
(WAP) enabled mobile phone and Jini services to monitor and control LONWORKS devices located in the Telia Research
booth.

``We are pleased to be selected by Telia as a device networking standard for their Connected Home initiative,' said Frits
Bruggink, Echelon's vice president, Europe, Middle East and Africa. ``In Scandinavia, and throughout Europe, we see a
growing momentum behind the networked home and a growing recognition of LONWORKS networks as the defacto
standard for connecting everyday devices in the home. While many parts of the world are talking about smart homes,
they are becoming a reality here in Europe though initiatives such as Telia's Connected Home.'

About Echelon Corporation

Echelon Corporation is the developer of LONWORKS networks, recognized internationally as a standard for open,
interoperable control networks. With thousands of application developers and millions of devices installed worldwide,
LONWORKS is the leading cross-industry standard for networking everyday devices in building and home automation,
industrial, transportation, and public utility applications. Echelon offers a full range of off-the-shelf hardware and
software products to support the development, installation and management of intelligent, open and interoperable control
networks.

Products that have been verified to conform to Echelon's LONMARK interoperability guidelines are eligible to carry the
LONMARK logo, an indicator that a product has been designed to interoperate over a LONWORKS network.

Echelon is based in Palo Alto, California, with international offices in China, France, Germany, Japan, Korea, The
Netherlands, and the United Kingdom. Further information can be found at echelon.com. Further
information about the LONMARK Interoperability Association can be found at lonmark.org.

About Telia AB

Telia AB is the leading supplier of telecom-based information services in the Nordic/Baltic region. At the same time,
``seamless' pan-European and global services are being realized through powerful alliances and partnerships in Europe
and around the world. Annual turnover in 1998 reached 51,240 SEK and the number of employees amounted to
approximately 30,600. Further information can be found online at telia.com.

This press release may contain statements relating to future plans, events or performance. Such statements may involve
risks and uncertainties, including risks associated with uncertainties pertaining to the timing and level of customer
orders, demand for products and services, development of markets for Echelon's products and services, and other risks
identified in Echelon's SEC filings. Actual results, events and performance may differ materially. Readers are cautioned
not to place undue reliance on these forward-looking statements, which speak only as of the date hereof. Echelon
undertakes no obligation to release publicly the result of any revisions to these forward-looking statements that may be
made to reflect events or circumstances after the date hereof or to reflect the occurrence of unanticipated events.

Note to Editors: Echelon, LONWORKS, LONMARK, the LONMARK logo, and the Echelon logo are trademarks of Echelon
Corporation registered in the United States and other countries. Other marks belong to their respective holders.

Contact:

Echelon Corporation
Lyn Balistreri, 650/855-7402
lbalistreri@echelon.com
or
Telia AB
Asa Dahlgren-Sundquist, +46-8-520-67-154
Asa.D.Dahlgren-Sundquist@telia.se

EOM -----------------------------------------------------