Is PCI Express not for embedded designs??

Even Cardbus was a stretch; XpressCard is, in the quest for higher performance, a vastly more complex interface designed to be implemented by an ASSP.

The PC is fast becoming a closed architecture anyway; for how many cards do you think you'll find detailed technical information readily available?

I'm not quite sure what you're expecting to find... adding a micro to the back end of a PCI/(e) bridge won't involve or require any 'libraries' or 'design tools' pertinent to PCI/(e)...

You need to find a PCIe bridge chipset like PEX8311 that provides a local bus interface to which you can glue an embedded micro. From there you'll be twiddling memory/registers in the bridge to talk PCIe...

Regards,

--
Mark McDougall, Engineer
Virtual Logic Pty Ltd, 
21-25 King St, Rockdale, 2216
Ph: +612-9599-3255 Fax: +612-9599-3266

Their only part to have PCIe is the 8641, and they don't seem very close to its finishing - but it may be a monster in the context anyway, you are probably after something smaller.

To reiterate on one of Lewins points, being able to buy a wifi card - be it PCIe or whatever - does not mean you can use it, unless you run MS windows. I spent some time searching and there is just nothing on the market one could use without writing his own wifi firmware for one of the few wifi chip(sets), and your options will be limited to those chips which are documented, which is valid for some older ones only.

I suspect this will be the real hurdle, not interfacing a PCIe card - bridging PCI to PCIe somehow should be doable (??? unless the silicon for that is also party members only area, of course).

Dimiter

------------------------------------------------------ Dimiter Popoff Transgalactic Instruments

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

Altera and Xilinx have built-in support in their FPGAs for PCI Express. I'm less familiar with Xilinx, but Altera actually has even the PHY of PCIe built into their GX series FPGAs (including the new low cost Arria GX). They also have a development kit for it, demo designs and an evaluation board.

On the cheaper side, you can look for dedicated PCIe interface chips that connect to a uC in some parallel way.

Eli

Yes, there are plenty of ethernet chips available that can be glued to a controller without the need for a PCI interface.

Rene

--=20 Ing.Buero R.Tschaggelar -

& commercial newsgroups -

Yes, we thought about this way back when we first developed this device in the 1990's

The main reason for going with a PCMCIA slot was up-gradability. This has extend the life of the product going from wired Ethernet to

802.11b, a,g... Cost wise it's very easy to do a in-field update when all you need to do is pull a card and update the firmware. Using "off-the-shelve" hardware made maintenance easy. Most often, the only thing that would fail was the Ethernet Adapter and the end user could pick one up a the local EStore and fix it themselves.

Sounds to me like those days are gone forever...such is progress.

I don't expect to see the detailed information like Lucent used to publish (for free) in the good old days. I'm thinking that it will have to go as part of an Open Source Project and be released using the available drivers in most part from the Linux groups. I see they already have drivers for most all of the WiFi cards using PCIe.

Steve

But if you do have the 802.11 knowhow you may be able to use one of the many USB based bridges? Perhaps at least some of them would be some device you are familiar with from the PCMCIA days, accessed over USB? Have you managed to use an off the shelf wifi adaptor (PCMCIA or whatever)? If you have, I would be very interested about some details (even if they are publically available, I still don't know where to look, although I once spent a few hours of searching).

Dimiter

------------------------------------------------------ Dimiter Popoff Transgalactic Instruments

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

We build a private label niche-market device. What we discovered is in extremely electronically noisy environments, like manufacturing with welding equipment, there was only one WiFi chipset that could handle it. It was built by Lucent called the Hermes One. We tested every other chipset we could find and all failed except this Lucent/Hermes combo.

There were several companies re-branding this setup (Compaq, Cabletron, Buffalo, Roamabout, 2Wire, Dell TrueMobile to name a just few)

The Prisim based devices didn't work at all.(Linksys, Netgear and all other "Small-Office / Home-Use Only" devices)

Back in the day, all you had to do was contact Lucent and ask for the complete design specs. They would have you sign an NDA, then send you a 20+Meg PDF file with everything you needed, start to finish.

All of the USB devices we've tested to date are useless toys, meant for maybe a small office with a clear line-of-site to the AP. My guess is low power is more important to them than Signal Strength. They advertise 300' at 2MPS, but it's more like 30' at 1MPS. These work more like BlueTooth than real, Industrial Strength WiFi. .. :(

I'm beginning to think this is going to be the same as it was when we did the 16-Bit PCMCIA card years ago. First we built the prototype, got it sort-of working then received the white papers explaining all these things we needed to do, but really didn't need at all.

We threw the books aside, tweaked the firmware for a few days and have been selling them ever since.

At best from the first prototype to the working model was no more than a week, with just two people working on it, one hardware, one software.(Of course there have been updates through the years as we added the new security protocols, fixed bugs and whatnot.)

99.9999% of the capabilities this new PCIe does, there is no need for in a "simple" embedded design.

Speed is not important when you are talking to a WiFi system that maxxes out at 27Mbps on a good day, but spends most of its life at

0-1Mbps.

Are we really going to have to add a $39.00 PLX controller chip so a micro can talk to an $8.00 PCIe card?

I mean if the only device a micro like a Coldfire is ever going to talk to is a WiFi card, shouldn't be able to do so on its own?

Am I over-simplifying this?

Steve,

I do not think you are oversimplifying it, things are hardly acceptable as they are today.

But there may be another option - not use PCIe at all. I have looked through a few chip(sets) which are made for single package wifi, really tiny. They are serially interfaced one way or the other - SPI in many cases, beats me why exaclty this but one can live with that. Then some of them advertise "enough power" to work without an additional amplifier, and some indicate how to connect one. I suspect the later makes the difference between those which will be good for you and the the rest (clear line of sight-style, I might prefer them as in most applications I have I would just go for the lower EMI). One of them - which can only do only "b", however - is documented and available (Thanks to Guy Macon who got the datasheet and made it available). No firmware for it to be found I know of - but stocked at Digikey... I think I found some data on another one - STM - but not everything, this may take more fights. Chances are these are compatible to something old you already know of - that Intersil thing or whatever - at this or that level. Since a PCIe board is unlikely to be easier to get documentation on than these single-package things, the choice seems easy to me.

Dimiter

------------------------------------------------------ Dimiter Popoff Transgalactic Instruments

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