Dp83848C with auto-nego crashs

Thierry,

Could you please provide the complete register set (addresses 0x00 - 0x1F) for the good condition and the problem condition?  When you read the registers, please read them twice.  Some of the registers latch their initial status so the current value is only available on the second read.

Patrick

Hi Patrick

To read the register, I dit that : 

int _stat_PHY[0x1F]; int _c; int _c2; for (_c = 0;_c<= 0x1F;_c++) { _stat_PHY[_c] = read_PHY(_c); for (_c2 = 0;_c2<=1000;_c2++); _stat_PHY[_c] = read_PHY(_c); for (_c2 = 0;_c2<=1000;_c2++); }

I uploaded a screenshot with the registers

--------------------------------------------------------
On 200 boards the Ethernet is OK

and on the installation where there are 20 boards, there is a problem

I taken one board to test in my office

The board rans during 1 week, no problem.
There is only probleme when I unplug the cable and plug again or in power on


and on the new board there is no problem

Thank to help me

At 100Mbits is OK

_{I uploaded also the registers for a good board}

--- ATTENTION

It's a new picture

the register #16 switchs between 0x4E15 and 0x6E15. 

nobody has an idea ??

The difference between 0x4E15 and 0x6E16 is the presence of receive errors.  However, both cases exhibit false carrier events.  It would appear that the receiver is having difficulty locking to the receive signaling. 

It may be that there is some marginality in the design.  Could you share the schematic? 

Patrick

Hi

I uploaded the page with DP83848C.

This schematic works well sinds 5 years. There is only 20 boards where there is a problem.

While the board is never reseted and the cable is not disconnected there is no problem

There is only a problem in power on or when the cable is disconnected and connected again (after 5 times)

No problem if I force 100 MBITS.

always nobody has an idea ??