4Links provides boards, chips, IPR and consultancy .... for links.
Thank you for finding this site. We'll be tending from now on to add news items on the 4Links site, rather than here. So please read up some of the history on this site, but look at the news on the 4Links site at www.4Links.co.uk
In June 2000, we mentioned that 4links is now incorporated! See the announcement on the 4Links.co.uk site. And in July we announced that we would be moving to Bletchley Park, home of the great code-breaking work during WW2, which is setting up an incubator unit for innovative Hi-Tech companies, and who are as keen to have 4Links as we are to be moving there. We're now there! (or Here!)
Following on from the SpaceWire-PCI boards shipped in early May, we made further shipments in June and will make more in July. We've had a very nice thank you, both to 4Links and to Dundee University, from one of the branches of NASA who have bought these boards.The SpaceWire-cPCI board was used to check out some corner cases of the software driver. We still have to finish mechanical details like the front panel, and hope it will be possible to ship these boards this month.
Between business trips and a very enjoyable holiday in the french Alps, I (Paul) was away for much of June. With Chris Hyde and Steve Bickerdike in the office, we were covered by someone here every day I was away. They had a few unplanned emergencies to deal with and dealt with them all very well. While on holiday I did a little work towards a lower-cost physical layer for 1355, with one or two twisted pairs instead of the current four pairs for signals. The work went well.
Plenty has been happening both in the space industry and elsewhere, partly as a result of this presentation at February's SpaceWire meeting at ESTEC. Apart from presentations from the space community, this meeting also had presentations on use of 1355 for robotics and for home networks. Paul Walker gave three presentations overall, one about the "macrocells" from 4Links, one for the 1355 Association, and one looking at ways to build highly fault-tolerant switches for space. After the main meeting, we had a "Core Working Group" session on the latest draft of the standard. Steve Parkes had done great work in producing this draft, but I (Paul) still had a number of concerns about it. I was very grateful both to Brian O'Neill and Barry Cook for their support in the discussions.
Towards the end of February, the 1355 Association finally came under the wing of ISTO, the IEEE Industry Standards and Technology Organization. ISTO put out a press release, including a quote from Paul. The quote tries to show the enormous breadth of application of 1355. We did not yet name the company who talked about home networks at ESTEC, and it's a great company to have with us --- take a look at their presentation to find who.
The C113 looks good, and we're adapting it to four links to fit the "4LinkBox" (aka "matchbox") that seem to work with Spartan chips so far. We even had the C101 talking to 4LinkBox to 4LinkBox to 4LinkBox to 4LinkBox to C101, and hot plug in this chain was the best we have ever seen. We've designed the 4LinkBox to take Xilinx 4000 series and Virtex as well, so this board should make a good development platform both for 4Links and for other link users as well.
Paul's 4Links Christmas Newsletter 1999 was rather provocative, but some very nice comments have been received about it. It finds a few examples in real life of Occam/CSP channels, and of course the links themselves are such channels. With keen interest in the Space industry for formal methods and generally making sure things are going to work, it is possible that there may be a growth in interest in these channels.
The C113 is a three-port interface for 1355 which includes larger FIFOs than the C112 and an 18-bit wide interface. An improvement over all 1355-DS implementations that we know of is that it reliably reconnects after disconnection --- we use some extra protocol but totally compatibly with the standard (Thanks again to Barry Cook for this protocol and for the C113 design).
Paul gave a presentation on 1355 at the IEE's Open Forum for their Electronics and Communications Division, on 1 July. The talk was the last in the main lecture theatre before the evening lecture by Sir Peter Bonfield, CE of BT. (There was a gap of half an hour between the two talks, so we weren't really on the same platform!) It's not often we have the chance to be in such illustrious company, and was a great opportunity for people to hear for the first time about 1355. Judging from the applause after the presentation on 1355, and the people who wanted to talk afterwards, it went pretty well.
Back in December 1998, orders came in for a few PCI-1355 boards, but that had been quiet since until a few weeks ago. To meet the demand we've ordered more PCBs, enough to use all the STC101 chips in stock. For a while I tried to put customers off, by telling them the board is slow and that the bugs in the C101 limit the board's usefulness. But, for the time being, it is the only 1355 board on the market that plugs into PCI, and the demand indicates a renewed interest in 1355. We've now secured components to ensure that we can to extend the life for a few more tens of boards.
In December 1998 also, we reported Xilinx introducing Rad-Hard FPGAs which could use the 4Links C112 design for Space applications.
Intel's announcement of USB2 threw 1394 into turmoil, but it is still bus-based and the industry is going to have to bite the bullet sometime and change to the modern paradigm of switched networks. Of course 1394 has a great bandwagon, and of course Intel have huge influence, but the applecart has been upset and this presents another opportunity for links. The 4Links Xilinx implementation of a 1355-DS-SE link has now been shipped to the first customer. The xilinx chips talk to other link components at the full 100 MBaud of 1355. The 4Links circuit is called the C112 Asynchronous Receiver/Transmitter (ART), includes two 16-byte FIFOs, an improved start-up mechanism, and the autonomous flow control and autobaud of 1355. It fits the slowest version of the smallest (and therefore cheapest) Xilinx FPGA, which Xilinx say will cost just $2.95 in volume. The design is available under licence. The current implementation appears to have good margins at 100 MBaud, and simulations suggest that higher speeds will be available from the faster versions of the Xilinx FPGA family. An evaluation/demo board is available with the Xilinx chip and link connectors, so you can see how well this works. A ModelSim simulation file is also provided in ModelSim's IP-secure form, for evaluation of the design with your own circuitry.
The availability of 100 Mbaud asynchronous communications, for the cost of the cheapest Xilinx FPGA, opens up a host of new opportunities. If you can use these communications, for whatever application, contact info@4Links.co.uk.
4links was started by Paul Walker, and has worked closely on a number of projects (including the C112 ) with Barry Cook of Keele University. Paul is also Editor to the 1355 Association
Partly as a result of being editor to the 1355 Association, Paul was invited to edit a special issue of "Microprocessors and Microsystems" (Elsevier Science) on 1355. The ten papers made a double issue of the journal, which has just been published. The papers show the vast amount and the quality of the work that has gone into 1355. In spite of that and its success in the Space market, 1355 is still rarely known and we hope the special issue will go some way to make it more widely known. In the editorial, Paul suggests that the problem is partly that 1355 uses the new paradigm of switching. Plenty of other technologies such as Ethernet and ATM are going switched, but none of these technologies was specifically designed for connections between processor chips. As just about everything we buy now has a processor chip, an interface for connecting these chips together is probably better than any interface designed for a particular application or particular chip.
The bus that has been conventionally used to connect microprocessors is a paradigm due for extinction. Like all paradigms due for extinction, it is putting up a pretty good fight, and the last bastion of the bus will probably be 1394 (aka FireWire). The ballot for 1394a closed in June, and Paul was horrified at the complexity that has been added to an already complex standard, and at the woolliness of the recommendations for compatibility between different variants of the standard. The final appendix, giving advice on performance of a 1394 network, and in the light of 1394's being trumpeted as a consumer product, was a farce.
By comparison with 1394, the special issue of Microprocessors and Microsystems demonstrates an outstanding quality of work, however small the band of us doing it. The editorial goes on to use Paul's analogy with manufacturing. However unlikely the change to something like 1355 looks now, this analogy shows that the change is utterly inevitable. Perhaps the equally guaranteed problems that will be thrown up as 1394 gets into the market will bring that inevitable day forward.
Heartfelt thanks to the authors, the reviewers, and the publishers for their contributions and patience on the special issue. Also to Peter Thompson who invited Paul to edit it. The issue is Volume 20, nos 7,8, dated 30 March 1998, but actually hot off the press in July/early August.
Here are a number of 4Links products and proposals to make more use of the 1355 link technology.
The 4Links PCI-1355 board (180k .pdf), (photo 54k .jpg, heavily aliased) is a simple interface to two links that can be used in any modern PC. The board uses the SC101 chip, which has now been withdrawn, so before too long the board needs to be replaced. As commented above, the SpaceWire-PCI board is coming on, but there is probably also a need for a simpler, lower cost, replacement for the existing 4Links PCI board. The conference paper Transputing without Transputers (56k .pdf) describes some of the design decisions taken for the PCI-1355 board.
During the delay for the C101s to arrive for the boards, Barry Cook designed a PLD version of the DS-SE link, and managed to squeeze it inside a cheap 44-pin PLD. 4Links is selling the CW1355-C111 chips (65k .pdf) as programmed PLDs, as a licence for the PLD program, or as a licence to the design for customised products. The C111 has had some excellent reviews from customrs recently. The attached application note (65k .pdf) refers to a slightly earlier version of the chip, but is still useful. We have found a keen interest in making link circuits available as FPGA designs, so that users of transputers and links, let down by the original supplier, can continue with the technology and keep it within their own control. Hence a great deal of work for specific customers on special versions of the FPGA circuit.
An important potential application of links is for the network in the home, and a couple of patent applications were filed during 1996 on UTP cabled links and on multi-cast routing, both useful or necessary for the home network. Out of this work came a paper (100k .pdf), presented at the 1996 WoTUG conference in Japan. Also here is a copy of the transparencies (200k .pdf) used in the presentation. A response (90k .pdf) has recently been submitted to the VESA Home Networks "RFI" for network protocols etc, which is based on the routing protocols and suggests use of "JavaPP" for configuration and management. "JavaPP" --- perhaps now called "JCSP" --- has been developed by the universities of Oxford, Kent, and Twente, to integrate the CSP primitives as a special class library within Java, to make it easier to write Java programs with thread communication and indeterminacy.
While in Japan back in '96, Paul gave a tutorial on the IEEE 1355 standard, based mainly on the 100 plus slides (file includes index of Tutorial slides) prepared by Colin Whitby Strevens who chaired the P1355 working group. Partly as a bit of lighter relief, but also because modern manufacturing started in Japan, the session included a short presentation of 1355's analogy with modern manufacturing techniques (24k .pdf). Here also is a table showing the analogy, that was used in the editorial for the 1355 special issue of Microprocessors and Microsystems.
The tutorial also included a number of slides about applications of links and 1355, provided by members of the 1355 Association. The Association's web site is now alive and running, including a huge amount of information and even letting you use some special software on the web site to give you a set of labels for the routing switches of your network. Please have a look.
The announcement about the demise of the T9000 has generated interest in Transputing without Transputers and perhaps also in Hardware for Virtual Channels. Thes are both now a bit old, but are still relevant today.
NB: Please Note Our New Address NB:
phone +44 1908 64 2001
Just for a change, August's's picture is from another holiday in the USA! (htm + 11k jpg).
Last Updated 1 August 2000
Disclaimer and Copyright notice. Unless otherwise noted, the information held on this site is Copyright 4Links for technical help, or C P H (Paul) Walker, or is held jointly between Paul Walker and Barry M Cook. Information may be freely copied, provided the copying is done in full, and provided that any copyright notice or attribution is included with all copies. Other web sites are warmly invited to reference this site --- please email notice that you have done so to firstname.lastname@example.org. All information is provided in good faith, but without warranty of any kind.