Howdy.
My brand new Ethernet driver employs a TLK106L with a Pulse HX1188FNL. It worked like a charm - if it wasn't for the cable lenght. For testing, I am using 30m pieces of CAT5e SF/UTP 24AWG cable connected together with a CAT5e inline couple to form 30m, 60m, 90m, etc.
For testing, I employ three different configurations: DUT is always my TLK106L circuit. On the other end of the line is either another one of my TLK106L boards, an HP PC network card, or another cirucit with DP83849IVS.
Now here is the result: 30m and 60m work very well. 90m is the problem. On the scope I see the TLK106 establishing a connection and then dropping it again after about one second. The reason is quite obvious: Signal levels have been attenuated to less then 100 mV (for the shortest signals that are only 16ns in length, the longer ones are not as bad).
My feelings go complete with the TLK106 - how can it get something out of these little signals. BUT: The PC's networking card and the DP83849IVS indeed work together. It's only my TL106 that is failing. To make things worse, the DP83849IVS unses an HX0068ANL as transformer, which is not as good as my HX1188, which has less insertion loss: -1dB compared to -3dB of the HX0068.
Our design is very close to the TI eval board layout and we payed special care to line impedances and differential impedances. So, I do not see ringing or something else that suggest a flaw in our design.
So here is my question:
Is a maximum cable length of below 90 m acceptable for a TLK106L? The datasheet talks about 150m error free cable length!