[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: packet drops in the ETRAX100LX ethernet
Thanaks to Per Zander,
I have tested ETRAX100LX revision 2 and 3,
PHY as Broadcom BCM5221 in auto-MDIX mode,
and RJ45 with transformer as Speedtech P52-111-1AG9.
When using revision 2, all is OK.
And when revision 3 and 10BASE-T hub or PHY chip in 10BASE-T mode,
it's OK, there's no packet loses.
But when using revision 3 and 100BASE-TX mode,
there's many packet loses and network bootloading is impossible.
I have tested for 6 PCBs of revision 3,
the result is same,
I have fixed the H/W as Per Zander's guide,
but it's not operated in 100BASE-TX mode.
----- Original Message -----
From: Per Zander <email@example.com>
Sent: Thursday, June 06, 2002 11:42 PM
Subject: RE: packet drops in the ETRAX100LX ethernet
> The older model "AXT6212 17854 A0147" is mask revision 2
> and the newer "AXT6212 18816 A0214" is mask revision 3.
> The differences between the two versions of ETRAX 100LX (part numbers
> 17854 and 18816) are limited to small changes in the USB interface,
> and there are no design related differences that could effect the
> Ethernet operation.
> We have performed extensive Ethernet communication testing on both
> mask revisions. The specific batch A0214 have been used in some of
> these tests. No problems have been observed.
> There may be some batch dependent variations in parameters like power
> consumption, signal threshold levels and output buffer speed, due to
> normal process variation. This kind of variation is taken into account
> in the specification of ETRAX 100LX, and should not cause any problems
> in a properly designed circuit.
> The problems you describe indicate that there is something marginal in
> the design of the circuit board, that makes it sensitive to batch
> variation. Typical problem areas are:
> - System clock (20 MHz).
> Jitter, noise and/or improper termination on the clock signal may
> affect the operation of the internal PLL in ETRAX 100LX. Too low
> amplitude of the system clock (e.g. because of EMI suppression)
> may also result in batch dependent problems.
> - Transceiver clock.
> Crosstalk and/or improper termination of the 25 MHz clock to the
> transceiver may cause packet loss.
> - PLL loop filter.
> Improper layout of the PLL loop filter may introduce noise that
> affects the stability of the internal PLL.
> - Insufficient decoupling or improper grounding.
> This may cause batch dependent problems, since variations in power
> consumption and output buffer speed vill affect the amount of
> ground bounce and power ripple on the board.
> Best regards
> Per Zander