Part Number:TM4C1294NCPDT
Hi. We are using TM4C1294 MCU in internal PHY mode for Ethernet.
Problem: Linking sometimes takes up to 30 seconds, sometimes doesn't even link, rare cases it gets link in few seconds.
My observations:
With oscilloscope I see FLP(fast link pulses) and their timing seems ok. But what is strange is what happens right after few FLP bursts. It seems like the PHY starts sending some data. When I check the tivaif_transmit function(which is furthest I can get with transmition in software) it shows no activity - so this data(if it is data) is not originated from software.
Also I noticed that I get very similar pulses at the same time on both Tx and Rx pair. I don' know if FLPs should be sent out on both pair at the same time, but I doupt it - so pulses at the same time would mean that it is interference caused by crosstalk?
This test mentioned above was done without cable plugged in the RJ45 socket.
I also checked PHY registers, and what seemed strange is that LPANABLE and EPHYANLPA registers change their value at the same time very few seconds.
LPANABLE = 4, EPHYANLPA = 0
LPANABLE = 5, EPHYANLPA = 16865
From these registers and all other observations I think what is happening is that MCU detects its own FLP pulses and think's it can autonegotiate and then fails, and this happens all the time and interferes with the real autonegotiation process.
I already tried different RJ45 mag jack but with no positive results.
In attachment you can see some images with FLP and the "data" or whatever that is send afterwards.
Can you please help? We have customers, waiting for solution... Thanks in advance.
(Please visit the site to view this file)