Reply by Max February 25, 20042004-02-25
On Tue, 24 Feb 2004 22:07:21 -0800, Albert Lee Mitchell wrote:

> Maybe it depends on what the definition of "is" is. 15 i/o pins on a >microcontroller eats quite a bit of cpu resources when most of those pins >have to be bit-banged
Sorry, that's just not true. The demo board I mentioned is quite capable of running a simple web server with only a cheap PIC16F877 @5MIPS, doing HTTP, TCP, UDP, ICMP, IP, ARP etc. in addition to driving the Realtek chip. If you need more horsepower for your APP, swap it for a PIC18 @10MIPS, or re-design with whatever MPU you fancy.
>plus a lot of embedded micros require quite a few of >those I/O pins when external memory is needed which would eliminate its >use for an embedded controller/server.
If external memory is being used, then you'd presumably memory-map the PHY, as Gerard suggested, which results in zero overhead on I/O pins (or one pin for CE if you're not externally decoding address spaces).
>>>I prefer to design for the future, not the past. >> >> You're never required to deliver real working systems today, then? >> Nice job. > > Sarcasm is not a good way to have dialog, smartass.
Well, what sort of response do you expect to that sort of meaningless pomposity? I fail to see what is wrong with hanging a Realtek PHY on a cheap MCU to achieve 10-baseT connectivity, given that it meets the requirements of the project in hand. You seem to be suggesting that all projects should be over-engineered to allow for some nebulous degree of future expansion, irrespective of whether the client requires it or not (and, presumably, irrespective of cost). Frankly, I think that's an untenable position to take. -- Max
Reply by Albert Lee Mitchell February 25, 20042004-02-25
On Tue, 24 Feb 2004 23:10:25 +0000, Max wrote:

> But even using a true microcontroller, the overhead of the RTL8019AS > isn't all that great, since you don't actually need most of the > address lines, and they can just be pulled. An example is Microchip's > POC board for 10-baseT, which uses only five address lines, or 15 I/Os > in total - reasonable on a 40-pin chip with 33 I/Os: > http://www.microchip.com/download/tools/picmicro/demo/pdemnet/39563b.pdf
Maybe it depends on what the definition of "is" is. 15 i/o pins on a microcontroller eats quite a bit of cpu resources when most of those pins have to be bit-banged plus a lot of embedded micros require quite a few of those I/O pins when external memory is needed which would eliminate its use for an embedded controller/server.
>>I prefer to design for the future, not the past. > > You're never required to deliver real working systems today, then? > Nice job.
Sarcasm is not a good way to have dialog, smartass. -- Regards, Albert ---------------------------------------------------------------------- AM Research, Inc. The Embedded Systems Experts http://www.amresearch.com 916.780.7623 ----------------------------------------------------------------------
Reply by Max February 24, 20042004-02-24
On Mon, 23 Feb 2004 08:11:21 -0800, Albert Lee Mitchell wrote:

>> Well, you can memory map the Realtek so it will cost you one additional line >> (CS). You also can use the READYIO and connect it to the wait pin (another >> line). It's possible to use the DTC (data transfer controller) to retrieve >> the data from the RealTek in background. > > Of course you can but not at the cost of a single pin. To interface >directly with a Realtek, not just the PHY chip, requires address and data >pins which eat all the h/w resources of a microcontroller.
I think Gerard was referring to designs which already interface to an external memory bus. In those circumstances, the CS line could well be externally decoded, along with all else, in a PAL or whatever. That's the benefit of memory mapping everything - zero overhead. But even using a true microcontroller, the overhead of the RTL8019AS isn't all that great, since you don't actually need most of the address lines, and they can just be pulled. An example is Microchip's POC board for 10-baseT, which uses only five address lines, or 15 I/Os in total - reasonable on a 40-pin chip with 33 I/Os: http://www.microchip.com/download/tools/picmicro/demo/pdemnet/39563b.pdf
>If no h/w >resources remain why bother? Nobody needs a webserver that's not >extensible. For instance how would you interface a LCD and keypad after >all the I/O has been eaten in the Realtek interface?
They aren't. Many pins on complex chips aren't really essential to basic operation, and good embedded designers learn to cut out the dead wood.
>I prefer to design for the future, not the past.
You're never required to deliver real working systems today, then? Nice job. -- Max
Reply by Albert Lee Mitchell February 23, 20042004-02-23
On Mon, 23 Feb 2004 15:46:23 +0100, Gerard Zagema wrote:

> >> My problem with the Realtek parts is that they require most/all of the >> pins available on an embedded microcontroller and mos/all of the cpu >> resources. This is why we went with the WC3100 and Cygnal, most of the >> i/o and cpu is still available for applications. Just serving webpages >> doesn't do us much good unless the 'server' can do something of value like >> turn a lamp or toaster on/off at least. >> >> -- Regards, Albert > > Well, you can memory map the Realtek so it will cost you one additional line > (CS). You also can use the READYIO and connect it to the wait pin (another > line). It's possible to use the DTC (data transfer controller) to retrieve > the data from the RealTek in background.
Of course you can but not at the cost of a single pin. To interface directly with a Realtek, not just the PHY chip, requires address and data pins which eat all the h/w resources of a microcontroller. If no h/w resources remain why bother? Nobody needs a webserver that's not extensible. For instance how would you interface a LCD and keypad after all the I/O has been eaten in the Realtek interface?
> All the other pin's and ports are for the fun of the user so you can > switch/control more then you probably ever need for an embedded project like > this. The modules like ST2011 are used quite a lot now for ModBUS interfaces > and GUI's for PLC's. You don't need any more pins then the RS485.
I disagree. Bill Gates once asked, "Why would anyone want more than 640k of RAM?" History proves that users will quickly swallow every feature and ask for more.
> But we have also user who use this device for monitoring temperature and > controlling Fan's for potatoes conservation. They use four of the 8 ADC's > inputs and two outputs for relais.
That's a great start. Do you have the CPU resources to do a Fuzzy Inference between web requests or to do an interpolation of non-linear data? When your hardware _and_ software are maxed out it severely limits the possiblities. I prefer to design for the future, not the past. -- Regards, Albert ---------------------------------------------------------------------- AM Research, Inc. The Embedded Systems Experts http://www.amresearch.com 916.780.7623 ----------------------------------------------------------------------
Reply by Gerard Zagema February 23, 20042004-02-23
> My problem with the Realtek parts is that they require most/all of the > pins available on an embedded microcontroller and mos/all of the cpu > resources. This is why we went with the WC3100 and Cygnal, most of the > i/o and cpu is still available for applications. Just serving webpages > doesn't do us much good unless the 'server' can do something of value like > turn a lamp or toaster on/off at least. > > -- Regards, Albert
Well, you can memory map the Realtek so it will cost you one additional line (CS). You also can use the READYIO and connect it to the wait pin (another line). It's possible to use the DTC (data transfer controller) to retrieve the data from the RealTek in background. All the other pin's and ports are for the fun of the user so you can switch/control more then you probably ever need for an embedded project like this. The modules like ST2011 are used quite a lot now for ModBUS interfaces and GUI's for PLC's. You don't need any more pins then the RS485. But we have also user who use this device for monitoring temperature and controlling Fan's for potatoes conservation. They use four of the 8 ADC's inputs and two outputs for relais. Regards, Gerard Zagema
Reply by Gerard Zagema February 23, 20042004-02-23
> Yet another person learns how to hang a Realtek > ethernet chip on a cheap processor and port > a open-source stack to it... >
Well, StackTos is not an open project but it is our own OS. It is working now for several years in some hostile environments. But I know, it's always easier to criticize then to give some constructive information. To hang an Ethernet controller on a cheap processor that is what this discussion is about, if I'm right, so I don't understand the insinuations....or is it a compliment? In that case......Thanks. Gerard Zagema
Reply by Albert Lee Mitchell February 22, 20042004-02-22
On Sat, 21 Feb 2004 19:39:27 +0000, Owlmeat wrote:

> Casey wrote: >> Marko said... >> >>>You may consider Hitachi processors. Very powerfull devices, and lot of >>>tools. >>> >>>An example at: >>> >>>http://www.stacktools.com/page/cdir.c?dir=ucmodules/ST2011 >> >> >> Sorry, but that's one of the most uninformative web sites I've >> seen in some time. Might want to finish it some of it first >> before giving the URL out. > > Yet another person learns how to hang a Realtek > ethernet chip on a cheap processor and port > a open-source stack to it...
Well, to give him credit, at least he accomplished something. That was an invaluable lesson, if he understood the source code. I'm sure his next version will have significant improvements. My problem with the Realtek parts is that they require most/all of the pins available on an embedded microcontroller and mos/all of the cpu resources. This is why we went with the WC3100 and Cygnal, most of the i/o and cpu is still available for applications. Just serving webpages doesn't do us much good unless the 'server' can do something of value like turn a lamp or toaster on/off at least. -- Regards, Albert ---------------------------------------------------------------------- AM Research, Inc. The Embedded Systems Experts http://www.amresearch.com 916.780.7623 ----------------------------------------------------------------------
Reply by Owlmeat February 21, 20042004-02-21
Casey wrote:
> Marko said... > >>You may consider Hitachi processors. Very powerfull devices, and lot of >>tools. >> >>An example at: >> >>http://www.stacktools.com/page/cdir.c?dir=ucmodules/ST2011 > > > Sorry, but that's one of the most uninformative web sites I've > seen in some time. Might want to finish it some of it first > before giving the URL out.
Yet another person learns how to hang a Realtek ethernet chip on a cheap processor and port a open-source stack to it...
Reply by Casey February 21, 20042004-02-21
Marko said...
> You may consider Hitachi processors. Very powerfull devices, and lot of > tools. > > An example at: > > http://www.stacktools.com/page/cdir.c?dir=ucmodules/ST2011
Sorry, but that's one of the most uninformative web sites I've seen in some time. Might want to finish it some of it first before giving the URL out. Casey
Reply by Marko February 21, 20042004-02-21
You may consider Hitachi processors. Very powerfull devices, and lot of
tools.

An example at:

http://www.stacktools.com/page/cdir.c?dir=ucmodules/ST2011

Cheers,

Gerard Zagema