Fast and easy way to a LPLC Ethernet Remote IO Rack

C

Thread Starter

Curt Wuollet

Hi all I was following a PR article on Linux Today for Zflinux systems and found something very interesting. It is a complete design for a Linux Internet Appliance with Orcads, Gerbers, and the whole nine yards. It has everything we need for an Ethernet Remote Rack except DIO. It has the PC busses pinned out so we could add a few 8255's, some buffering, and have the elusive LPLC EIO rack, running Linux, ready for programming in a very short time for very little effort. How many people are interested? I could really use someone who has OrCad access to download and do a Postscript plot of the thing so we can get an idea about size and architecture. You can find the things at www.zflinux.com in the downloads section cleverly disguised as an Internet Appliance reference design. It even has 8 D/A and 8 A/D channels. Should be a slam dunk if we can get the resources. On a related note, my company has taken over the development of the translator board for the DIO48. This means I should have artwork next week and it will belong to the project. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C

Curt Wuollet

Hi Ken Per another post, I didn't get Eagle as lite is too small for the project and standard and pro are too spendy for me to do out of pocket unless I do a lot more board work. pcb lacks an autorouter and schematic capture but the price is right. I'm sure someone on the list can try it and get back to us. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
W
Curt, Doesn't this make the cost per point rather prohibitive for a small group of I/Os? I have downloaded the package, but the only file I can look at is the BOM. It's got hundreds of parts! I don't know how much the ZFLinux chip costs, but this board must cost over $200 to build. Maybe I'm out to lunch, but I was hoping for a <$50 BOM and much less complexity. Willy _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
Hello, Rabbit Semiconductor makes a Ethernet Development kit for under $200 and it gives you IDE, code, schematics, etc. Their processors are dirt simple to interface to as far as hardware. They even sell the hardware, including memories and the ethernet chip on their site. ~Ken _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C

Curt Wuollet

Hi Willy It would be cost prohibitive for small point counts. I would expect that there are things we could do without though. Opto's Ethernet "Brain" is real spendy too and does Modbus/TCP rather poorly. Optimation's rack is quite a bit better on price but it doesn't speak anything we are real interested in. There are pluses however, it could run a _lot_ of IO. And being a whole PC it could run LPLC if the guys don't get too carried away. Looking at it as a standalone LPLC it's fairly cheap. I suppose it's better to compare it to a PC than to just intelligent IO. I started out looking at the uCsimm. That's expensive and not very powerful. Most SBC's are awful spendy or don't have networking or some other essential. I hope you can find something at $50.00 too. I have found SBC's that would be cheaper, but you would quickly burn the difference in porting Linux and adding enough stuff to do what we need. That's why it would help if someone can dump the artwork, I'm sure we could reduce the parts count if we know what we're looking at. I said "Fast and Easy" I didn't say cheap :^). Like all engineering.... cheaper, faster, or better....pick any two. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C

Curt Wuollet

Hi Ken I haven't seen a Linux port yet. We have a problem developing stuff for our OSS PLC on a proprietary runtime. Of course, if it's cool, a Linux port should happen any day now :^). I'll take a look around for Linux on Rabbit. I too was excited when I saw it a while ago. I could be wrong, but I think we want Linux on both ends. The really cool thing about the Zflinux stuff is that so much is already done. Hacking internals to fit on a tiny micro is for younger guys who can hack till 3 AM and not crash in their cornflakes :^) Having something now doesn't mean we can't have something better. (cheaper) later. We aren't in it to corner the market or make a million. With all the embedded Linux development going on things will just get better and better. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C

Campbell, David (Ex AS17)

> Curt Wuollet wrote: > Hi Willy > > It would be cost prohibitive for small point counts. *snip* > That's expensive and not very powerful. Most SBC's > are awful spendy or don't have networking or some other essential. > I hope you can find something at $50.00 too. Quoting one of the press releases on zflinux's web site: "The MachZ is priced from $60 in low volume to mid $30's in high volume quantities on an OEM basis." I have been working my way through NatSemi's chip data sheet pages and found that the minimum part count for a NIC is 1 x controller, 2 x SRAM, 1 x EEPROM + x passive components. (Add in the voltage convertor if you want BNC). SRAM for the packet buffers, EEPROM for the configuration (including MAC address). Passives for the signal issolation. You will break the $50 limit just attempting to add network support. If you are serious about Linux on an embedded system you might want to throw the kernel into EEPROM (that is going to cost a few dollars). > I said "Fast and Easy" I didn't say cheap :^). > Like all engineering.... cheaper, faster, or better....pick any two. Just about got the three corners of project management there. price, quality and time, it is always a tradeoff between those three. Does anyone know of an ethernet &lt;=> serial convertor that can be accessed from the ethernet (eg: a reverse RAS box?). David Campbell _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
Actually Curt, The Rabbit 2000 people seem to have put together a nice software/hardware package for TCP/IP development and provide schematics, debugger, etc. .. I don't think that it gets much easier. Lets face it, you just have a main loop copying IO information and the ethernet stuff would happen on an Interrupt routine and get buffered for the main loop to process. I did a similar thing for the CAN bus with a siemens Micro and it went rather nicely. I would be able to help here, but I don't yet understand TCP/IP and ethernet. I think running linux on the IO end *just* for remote IO is kind of like hitting the proverbial fly with the sledgehammer and a lot of people wont go for it. I think you will have just as much work trying to port it into a tiny micro (even the zflinux) and getting around the various issues that you would have been better off using an embedded micro with no OS on it. Having said that, it is kind of like all the topics on here ... ultimately it remains to be decided by the one doign the actual work. :eek:) ~Ken _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C
I've got samples from Crystal that have a lower parts count. You would still need the magnetics. It's gotta be doable I can buy NICS for $9 of course, they buy in huge quantities. I question their use of NSC's chip when the Crystal is supported. "Campbell, David (Ex AS17)" wrote: > > > Curt Wuollet wrote: > > Hi Willy > > > > It would be cost prohibitive for small point counts. *snip* > > That's expensive and not very powerful. Most SBC's > > are awful spendy or don't have networking or some other essential. > > I hope you can find something at $50.00 too. > > Quoting one of the press releases on zflinux's web site: > "The MachZ is priced from $60 in low volume to mid $30's > in high volume quantities on an OEM basis." > > I have been working my way through NatSemi's chip data sheet > pages and found that the minimum part count for a NIC is > 1 x controller, 2 x SRAM, 1 x EEPROM + x passive components. > (Add in the voltage convertor if you want BNC). SRAM for the > packet buffers, EEPROM for the configuration (including MAC > address). Passives for the signal issolation. > > You will break the $50 limit just attempting to add network > support. > > If you are serious about Linux on an embedded system you > might want to throw the kernel into EEPROM (that is going > to cost a few dollars). Or just EPROM for a lot less. > > I said "Fast and Easy" I didn't say cheap :^). > > Like all engineering.... cheaper, faster, or better....pick any two. > > Just about got the three corners of project management there. > price, quality and time, it is always a tradeoff between those > three. > > Does anyone know of an ethernet <=> serial convertor that > can be accessed from the ethernet (eg: a reverse RAS box?). Chase Research and a lot of others. Or of course something like the design above or an old 486 running Linux. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
W
Hi guys, The Crystal Ethernet footprint goes on less than 2" of board space and has just a couple of parts. The CS8900 is about $10 in <100 quantities. The magnetics are REELY cheap for through-hole parts. Besides, if we're looking for the cheapest parts, Winbond and some other companies have chips that are already supported by Linux and have minimal parts counts, too. Take a look at a $9.00 NIC, they're that cheap because they only have a few parts. Processor? How much power do you need for an I/O block? There are already ports of Linux to many other low-cost 32 bit microcontrollers. If you want to make this Open-source hardware project be easier in that regard, it could be shared with other groups, written about in some of the Linux journals, etc. What about the Mot MCF5272 ColdFire processor? It has enough guts to run Linux, has an Ethernet controller built-in (albeit minus the PHY), and costs $11.00 (unknown quantity). It also has an SDRAM controller, so we could use a glob of cheap DRAM, and a huge cheap 8 bit Flash part, moving all executables into RAM. I think we can do it for $30 or even less, with a lot of bang for the buck. It just can't be a PC. No way; but it doesn't need to be. What about making it so it looks the the DIO48? Then it would already drive the other boards being discussed. Any comments? Willy _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C

Campbell, David (Ex AS17)

> Curt Wuollet wrote: > I've got samples from Crystal that have a lower parts count. > You would still need the magnetics. It's gotta be doable I > can buy NICS for $9 of course, they buy in huge quantities. Yup, looking at Crystal's web site this is indeed the case. > I question their use of NSC's chip when the Crystal is > supported. Sorry, I went to NatSemi as they put all their datasheets on the web. Other chip manufacturers like you to write in for them. As for the terminal server, it seems a waste (of space and resources) to use a 486 even though it is the cheapest option. It is something that will eventually bite you (due to on-going maintenance and support). David Campbell _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
A

Andrew Kohlsmith

> > I said "Fast and Easy" I didn't say cheap :^). > > Like all engineering.... cheaper, faster, or better....pick any two. > Just about got the three corners of project management there. > price, quality and time, it is always a tradeoff between those > three. Yes... I thought the phrase was "On time, On budget, On spec... pick two of the three" :) > Does anyone know of an ethernet <=> serial convertor that > can be accessed from the ethernet (eg: a reverse RAS box?). There are a few "dumb" Ethernet/485 gateways for ModBUS already. I believe Omega makes one for $99. A personal project of mine is a much smarter gateway which maps the 485 component's address spaces into its own and has customizable polling and forwarding abilities and is based on the Coldfire 5272 processor (5206 with ethernet basically). Regards, Andrew _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C
Hi Willy Go for it! No, actually you are right, we could make one with ColdFire or StrongArm or the other popular and ported embedded type Micros for much less. The problem is that we would need fairly serious resources both fabrication and programming. It doesn't have to be a PC to run LPLC either although if it lacks an mmu we're toast. There is a good possibility we could find another "almost there" hardware project and get the same boost we would using the MachZ. What I'm trying to avoid is starting with silicon and doing everything that 100's of other folks are doing all over again. Perhaps the most valuable (and hardest) lesson I have learned from OSS is that I don't _have_ to do everything myself. Doing 10% isn't as satisfying as doing 100% but things sure happen faster. I do guage the doability of things from what I myself can do because I have been pushing this for a long, long, time and I have yet to get any offers of resources to get it done. With the MachZ, I can generate the schematics, someone with OrCad access could do the hacking and slashing of the exixting artwork and the programming would be the same as a PC, maximizing the chances that it would happen. I could screw it up myself and let the real programmers rescue me :^) Your point of view is exactly the same as mine until I make the doability a factor. Netappliance designs for Linux are going to be coming out of the woodwork, There should be something we can hijack. Of course, if you have the resources....:^) Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C
Hi Ken Like I said, I could be wrong. If all we want is an Ethernet I/O rack with the ability to run the protocol of our choice, and we have the talent to implement those protos on that platform, you're right. After all, I'm pretty sure the Optimation rack is a PIC design. I am also interested in a rail mount home for LPLC and access to the Linux serial and networking tools so it's not _just_ for EIO. There is also the problem of ownership. If WOT was going to make these and sell them, I'd be right with you, I'd contract the talent I don't have and cut the best deal I could get with Rabbit and deal with licensing, etc. People would need a Rabbit license to extend and enhance our efforts. Of course, we can simply use the commercial offerings. Our only ROI is to be able to provide an entirely free (as in free speech) automation solution. Only Linux or ECOS provide that freedom as far as I know. Fortunately, there is intense interest and activity in hardware to run Linux in the embedded market. Also, we are blessed? with an industry with spectacular profit margins, an ideal environment to tolerate using Linux and powerful processors instead of the least cost approach. For low volumes, development would predominate and having an x86 should minimize that effort. Life is full of trade offs. What's it worth to have a solution now, before the proprietary interests further poison the Ethernet well? It could be priceless for universal connectivity, but, I could be wrong. Sorry to get political, but, changing this industry is going to take more than good engineering. Willy is looking at low cost and free solutions, I'm looking at free, fast and easy. The whole concept of "free (as in speech) hardware" is a new frontier. Whatever we come up with will be a first in this industry. It's something we really need to consider carefully. Don't get me wrong, Ken, I'm not shooting you down, we are all just feeling our way here. I like Willy's approach long term, but I want Ethernet hardware badly enough to compromise on cost. I can't compromise on freedom. There are a lot of subtleties with trying to make free hardware that aren't apparent until you get into the process. As soon as you have someone actually build the hardware, it's hard to keep it free and distributing just the IP doesn't help that many people. We can use all the help to figure this out we can get. Give the problems some thought, we can use your ideas. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
Hi Curt, I am with you here. If you want a standalone Controller with an ethernet port to communicate witht he rest of the world, the ZFlinux chip is definitely worhwhile to look into. I was not thinking of that with regards to my suggestion. I guess from what you were saying that you would like to have an intelligent "bus coupler" hardware design that would also double as a standalone controller. This I was not aware of. As far as the issue of open and free (as in speech) hardware/software for the remote IO bus coupler (assuming it is *just* a bus coupler and not a controller) any micro would do because you do not need any software on it except ANSI C. For what a bus coupler needs to do, it doesn't require any liscensed software. The rabbit2000 products are taylored to embedded design and I wouldn't think that they would charge royalties for us for using their libraries, but I do not know that for sure. I would be willing to bet that they want to sell their silicon nad development environments and leave liscensing out of it. I may be doing some rabbit2000 work at home in the future and I will look into the ethernet interface some more. ~Ken _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
C
Hi Ken Please do, and report back. It is an option if it fits our GPL requirements. I thought that Rabbit had a runtime executive. If not, and the libraries are at least LGPL we can look a lot harder at it. Free (as in beer) isn't quite good enough for a pure OSS project. I'm sure something will happen soon to make this all come together. If someone can make a Linux PDA for $150.00, we ought to be able to make this happen. Regards cww _______________________________________________ LinuxPLC mailing list [email protected] http://linuxplc.org/mailman/listinfo/linuxplc
 
Top