Davicom Ethernet PHY

Jul 18, 2006 6 Replies

I'm using AT91SAM256X with the Davicom DM9161A PHY in a configuration directly copied from the Atmel eval board (AT91SAM7X-EK).



I'm running the same software in both the Eval board and my board, and the eval board works 100%, so I think I'm looking at a hardware problem.



The problem is that I'm dropping data intermittently. Pinging 1000 times with a 256 byte packet gives me about 30% dropped. If I shorten the packet size, the error rate drops. At 32 byte long pings I'm at 2-4% dropped.



Been banging my head against this now for a couple of days, and not making much headway.



So my questions to this wise community; where should I look for problems?


Some more details :



I've looked at all the power pins for noise and my board is not significantly different than the Atmels eval board is.



I got a big improvement on the throughput when I changed my bandgap resistor from 6.2k (from the Atmel schematic) to 6.8k (which they really stuffed on the board). Is this a clue to something?



I have two deviations from the Davicom design guidelines;



1) My ground plane under the Rx/Tx pair is not AGND. It's DGND.
2) I'm running the Tx/Rx pairs close to each other. Do you thing this could be enough to cause these problems?

Any ideas would be much appreciated. Thanx.


I tried to get the Davicom DM9161A working on my ATmel at91RM9200 based board and had problems, never solved, with the auto negotiation sequence and a particular netgear hub. Davicom tried the hub and also found problems, though different, in the end gave up and used a Micrel KS8721 part.

Same Bad report here. We have the 9161 working on our designs and the same problem with the netgear hubs, never solved. We couldn't even get the 9161A to work, so in that regard I would tend to second the Micrel motion.

Chris

Thanks,

Doesn't sound too encouraging, although my problem seems different. I have no problem with the auto negotiation against my Netgear switch (FS116). It syncs up and does the ARP stuff on first try.

I'm thinking my problem is in the layout between the PHY and the magnetics. There is only 20mm of wires there, and while laying out the board I was thinking that those 20mm can't really have much effect compared to 30m of CAT5. Is this a true statement or is this piece of the design very critical?

Thanks again.

I'm not familiar with the PHY you are using, but looking at the error type indicated in the status registers might help.

Also, I notice that Davicom has layout guidelines for the 9161. You might see how close your board comes to following those if you haven't already.

Regards, Steve

"Anders" skrev i meddelandet news:Aqevg.123611$ snipped-for-privacy@newssvr13.news.prodigy.com...

You are of course aware that the RMII interface is no longer supported by the AT91SAM7X? You have to use the MII interface!

I have seen problems where the Davicom PHY has been set to fixed 100 Mbps Full Duplex. I have been told that if the other end uses Autonegotiation, then a 100 % standards compliant switch will fail. Some switches support this anyway.

Also look at the Ethernet clock frequency, if it is slightly off the nominal frequency, you will see errors and changing capacitances might help.

Best Regards, Ulf Samuelsson This is intended to be my personal opinion which may, or may not be shared by my employer Atmel Nordic AB

I was not, but luckily I'm using MII :)

The autonegociation seems to work just fine. It's the data transfer that's intermittent.

That's a good tip. I'll check on it. Thanks !

BTW, to get moving with my application code, I've forced the PHY to connect at 10M. Now I'm getting a perfect throughput. Sounds to me like it could be related to crystal frequency.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required