Migrating from 10 Base-T to 100 Base-T Ethernet...

Hello,
I am starting an investigation to determine the easiest migration path
to move an existing product from 10 Base-T ethernet to 100 Base-T
ethernet.  The current product is a small embedded system running
Linux (2.4.9 kernel) on a StrongArm processor, using the CrystalLAN
CS8900A chip for the MAC/PHY.  This chip is in a 100-pin QFP package.

I'd love to hear that there is a drop in replacement for this chip
that will allow us to move up to 100 MBps without much trouble, but
I've looked around and can't see to find any after a first pass.  I
did find an application note from SMSC entitled "Migrating from a 10
Mbps Ethenet Controller to the SMSC LAN91C111" which does support 100
Mbps, but it isn't exactly "pin-compatible" from what I can see.

Any other solutions out there?

Thanks in advance for the help...

John O

Hi John,

The SA1110 StrongARM part has gone end of life as well.  Perhaps now's a
good time to start with a white board?  My company has a product that has an
SA1110 in it as well.... so I understand the firmware/software investment in
a particular architecture.

If you're using an OS.... I believe it will take care of the differences
between the speeds.  At least Linux will....


Edwin


"john orlando" <john@jrobot.net> wrote in message
news:cc4b5d44.0407270727.63fcfc7b@posting.google.com...
> Hello,
> I am starting an investigation to determine the easiest migration path
> to move an existing product from 10 Base-T ethernet to 100 Base-T
> ethernet.  The current product is a small embedded system running
> Linux (2.4.9 kernel) on a StrongArm processor, using the CrystalLAN
> CS8900A chip for the MAC/PHY.  This chip is in a 100-pin QFP package.
>
> I'd love to hear that there is a drop in replacement for this chip
> that will allow us to move up to 100 MBps without much trouble, but
> I've looked around and can't see to find any after a first pass.  I
> did find an application note from SMSC entitled "Migrating from a 10
> Mbps Ethenet Controller to the SMSC LAN91C111" which does support 100
> Mbps, but it isn't exactly "pin-compatible" from what I can see.
>
> Any other solutions out there?
>
> Thanks in advance for the help...
>
> John O

Hi Edwin,
I know the StrongArm is in end of life.  I'd love to bump up to the
XScale, but it isn't going to happen in the next release of our
hardware unfortunately.

As for Linux taking care of the differences in speeds:  I am hoping
that whatever hardware/chip replacement suggested to bump up to 100
Mbps would have a device driver already set up for the SA-1110 running
Linux...I was hoping to not have to re-invent that wheel.  Yes, we're
running Linux, but that only helps out if such a device driver already
exists.

Any other pointers out there?

TIA...again,
John

"Edwin Bland" <edwinb@socal.rr.com> wrote in message news:<F3LRc.3239$Bt5.1734@twister.socal.rr.com>...
> Hi John,
> 
> The SA1110 StrongARM part has gone end of life as well.  Perhaps now's a
> good time to start with a white board?  My company has a product that has an
> SA1110 in it as well.... so I understand the firmware/software investment in
> a particular architecture.
> 
> If you're using an OS.... I believe it will take care of the differences
> between the speeds.  At least Linux will....
> 
> 
> Edwin
> 
> 
> "john orlando" <john@jrobot.net> wrote in message
> news:cc4b5d44.0407270727.63fcfc7b@posting.google.com...
> > Hello,
> > I am starting an investigation to determine the easiest migration path
> > to move an existing product from 10 Base-T ethernet to 100 Base-T
> > ethernet.  The current product is a small embedded system running
> > Linux (2.4.9 kernel) on a StrongArm processor, using the CrystalLAN
> > CS8900A chip for the MAC/PHY.  This chip is in a 100-pin QFP package.
> >
> > I'd love to hear that there is a drop in replacement for this chip
> > that will allow us to move up to 100 MBps without much trouble, but
> > I've looked around and can't see to find any after a first pass.  I
> > did find an application note from SMSC entitled "Migrating from a 10
> > Mbps Ethenet Controller to the SMSC LAN91C111" which does support 100
> > Mbps, but it isn't exactly "pin-compatible" from what I can see.
> >
> > Any other solutions out there?
> >
> > Thanks in advance for the help...
> >
> > John O

john orlando wrote:
> I am hoping that whatever hardware/chip replacement
> suggested to bump up to 100Mbps would have a device
> driver already set up for the SA-1110 running Linux...

You will not find a drop-in upgrade, since there are virtually no
non-PCI 10/100 controllers on the market.  Here are the only ones I've
found:

SMSC: http://www.smsc.com/main/catalog/lan91c111.html
Asix: http://www.asix.com.tw/emb-non_ax88796.htm
Davicom: http://www.davicom.com.tw/eng/products/dm9000.htm

The Asix chip is about 2 years on the market, and very close to NE2000
compatible, so you will find lots of drivers available with source. 
(The driver will require some tweaking, mostly because of a PHY init bug
and the extra 100Mb settings.)

SMSC has been around a while, but I see it mentioned even less often
than the Asix.  The Davicom seems to be a clone of the SMSC chip.


But the real question is... what's pushing you to upgrade to 10/100?  Is
it a marketing thing, or is there an issue you've run into (which might
have an alternate solution)?

Best of luck,
Richard

Richard <rh86@no.spam> wrote in message news:<41199DD5.46C948AA@no.spam>...
> john orlando wrote:
> > I am hoping that whatever hardware/chip replacement
> > suggested to bump up to 100Mbps would have a device
> > driver already set up for the SA-1110 running Linux...
> 
> You will not find a drop-in upgrade, since there are virtually no
> non-PCI 10/100 controllers on the market.  Here are the only ones I've
> found:
> 
> SMSC: http://www.smsc.com/main/catalog/lan91c111.html
> Asix: http://www.asix.com.tw/emb-non_ax88796.htm
> Davicom: http://www.davicom.com.tw/eng/products/dm9000.htm
> 
> The Asix chip is about 2 years on the market, and very close to NE2000
> compatible, so you will find lots of drivers available with source. 
> (The driver will require some tweaking, mostly because of a PHY init bug
> and the extra 100Mb settings.)
> 
> SMSC has been around a while, but I see it mentioned even less often
> than the Asix.  The Davicom seems to be a clone of the SMSC chip.
> 
> 
> But the real question is... what's pushing you to upgrade to 10/100?  Is
> it a marketing thing, or is there an issue you've run into (which might
> have an alternate solution)?
> 
> Best of luck,
> Richard

Ahh yes...the real question.  No, it isn't a marketing thing.  Our box
interfaces to another box that recently bumped up to 100 Mbps, and
thus we were investigating the possibility of us migrating as well. 
Our hope was to have this last revision of hardware be the last one,
and thus any potential expandibility issues for the future were being
investigated.  We aren't currently even approaching a multi-megabit
throughput (there are bursts of course that could probably take
advantage of the higher speed, but shaving a microsecond here and
there isn't really that important in this application).

I think I have my answer:  there isn't a drop-in solution, but there
are a couple of solutions that would maybe require a few weeks of
hacking away to get them working.

Thanks to all for the help!

John

john orlando wrote:
> Our box interfaces to another box that recently
> bumped up to 100 Mbps, and thus we were
> investigating the possibility of us migrating as
> well.

Well, I'm still unclear on the need for 10/100, since the 100Mb device
should auto-negotiate to 10Mb just fine (e.g., if you used a crossover
cable).

However, here's a hack that requires more hardware, less coding, meets
the need, and any marketing value as well...

Stick a 10/100 switch controller chip between the CS8900 and the
external jack.  Configure the CS8900 for full-duplex operation and let
the switch controller negotiate the external connection at 10/100
full/half.

Full-duplex can be a big performance boost, but it's a static setting
for 10Mb chips - you can really only turn it on when you *know* it's
connecting to a switch port at 10/full.  By adding the switch
internally, you get a performance boost, keep the existing driver code,
and gain an upgraded external interface.

The switch chips I've seen are virtually turn-key, so it might be worth
a look.  (E.g., Realtek RTL8309SB)  I've seen some consumer products
that even did auto-crossover, which would be a neat add-on if the switch
supports it.

Have fun!
Richard