[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: 100 MB Ethernet problem



I notice the spec from Broadcom's datasheet and it is
the first trigger to me to think about it may be the
source of the problem. The clock is not jitter free
even with the 0.022uF cap in place. I agree that using
a crystal or oscillator is a better idea. But it is
too late for us to adapt this because the layout is
done.

Eric

--- David Kilp <David.Kilp@xxxxxxx.net> wrote:
> I would also add the following in case it helps
> anyone else with similar
> Ethernet issues we had in getting a custom board
> going:
> 
> We also found issues with the 25 MHz ref clock. 
> The circuitry from the the Axis dev board (well the
> Axis 82 
> dev board we started with anyway) is probably not
> going to meet 
> the specifications for the Broadcom chip. If you
> read the specs 
> they specify a +/- 50 ppm clock with a 50% duty
> cycle (+/- 10%)
> and edge rates from 1ns to 4ns(10% to 90%). 
> If you use the clock circuitry of the development
> board I think 
> you will find that you will have a difficult time
> meeting 
> those specs especially as regards variation in
> parts, temperature, etc.
> We switched to a master clock oscillator circuit -
> one of 
> those all-in-one devices in a single package such as
> from CTS 
> rather than the discrete oscillator circuitry that
> the dev board used. 
> Once we did that and followed all the Broadcom
> Ethernet 
> pc-board layout recommendations (VERY IMPORTANT!!!)
> all the 
> mysterious Ethernet problems went away.
> 
> Regards,
> 
> Dave
> 
> > -----Original Message-----
> > From: owner-dev-etrax@xxxxxxx.com 
> > [mailto:owner-dev-etrax@xxxxxxx.com] On Behalf Of
> Zhuorong Deng
> > Sent: Tuesday, June 29, 2004 9:45 AM
> > To: Mikael Starvik
> > Cc: dev-etrax@xxxxxxx.com
> > Subject: 100 MB Ethernet problem 
> > 
> > 
> > Hi, Mikael:
> > 
> > I found the problem, which is the 25MHz Ref clock
> to
> > the PHY is unstable with the PLL filter components
> in
> > the recommended value from the datasheet. I have
> to
> > change the 680pf capacitor to be 0.022uf to make
> the
> > clock being more stable. After this, the unit
> works
> > fine with the 100MB network because I can see file
> can
> > be transfered at the rate of about 33MB/sec with
> ftp.
> > flash update image is also working with 100MB
> network
> > now.
> > 
> > I also post this to dev-etrax@xxxxxxx.com and hope it
> > will help someone with the similar problem.
> > 
> > Regards,
> > 
> > Eric
> > --- Mikael Starvik <mikael.starvik@xxxxxxx.com>
> wrote:
> > > In the output you can see that the Rx frame
> counter increases.
> > > This indicates a receiver problem as you say. I
> > > would say
> > > that this is a hardware problem (such as bad
> routing
> > > and/or
> > > bad impedance). I am not really a hardware guy
> so I
> > > can't
> > > really give any pointers on where to start.
> > > 
> > > -----Original Message-----
> > > From: Zhuorong Deng
> [mailto:zhuorongdeng@xxxxxxx.com]
> > > Sent: Friday, June 25, 2004 5:32 AM
> > > To: Mikael Starvik
> > > Subject: RE: 100 MB Ethernet problem 
> > > 
> > > 
> > > Hi, Mikael:
> > > 
> > > I attch the result of the ifconfig with this
> mail.
> > > It
> > > is from a unit/hub combination that works with
> ping
> > > but fail with ftp. It does not show many packet
> > > error
> > > or packet fropping.
> > > 
> > > If I use another hub with the unit, the ping
> will
> > > fail. Most of the time, I don't see the Rx
> packet
> > > count increased when the unit is being pinged.
> > > Sometimes I can see one or two packets coming in
> and
> > > then I will get the Reply for one or two times.
> It
> > > looks like the Rx end is having trouble. Any
> > > suggestion?
> > > 
> > > Regards,
> > > 
> > > Eric
> > > 
> > > 
> 
>