100Base-X and Controlled impedance

[jdhar]

I am working on a prototype board that has an FPGA and National DP83843 10/100 PHY on it. I was wondering if it's necessary to attempt to use controlled impedance for any parts of the PHY, or if it's ok just keeping trace lengths as short as possible? If controlled impedance is required, I'm assuming I need to at least have a GND plane on the other side of the trace, correct? And what paramaters do I have to obtain from a prototype manufacturer if using controlled impedance.

Thanks, Jai.

 

[sigint]

10/100 PHY requires Differential impedance control of 100 ohms as well as length and phase equalization, you need to read the specification closely or have a long talk with the design engineer.

SiGiNT

 

[jdhar]

In reading the national datasheet, I have not seen this anywhere. Also, what signals require controlled impedance? The only high-speed signals that I can see are the MII signals, clocks and TP/TD twisted pair signals. I was hoping that i could treat these as lumped models, but I can't find their rise-time anywhere!

 

[sigint]

The specification mentions several times that controlling the impedance is necessary, (ie: PG. 39), however it doesn't give you a value, poorly written IMHO, It also says it's not recommended for new designs.

Ethernet 10/100 is always routed for 100 ohms differentially controlled impedance, as to the values to control pair to pair and + to - skew, it's usually up to the engineer or a min/max is given in the data sheet. the 802.3 specification may be some help.

SiGiNT

 

[jdhar]

Thanks for your help. I'm aware that it's not recommended for new designs, but this is for a personal-project, and the signal integrity requirements for their drop-in replacement (a Gigabit-capable chip) seem much more stringent. I guess I will have to keep the traces as short as possible and length match - maybe that will help..

 

[supra924]

The control impedence is required since you are going to connect it to the long wires where youare not aware of impedence and preferably to use diode termination.

 

[jdhar]

Is it possible to implement 10/100 phy without controlled impedance routing? I have seen 50 ohm terminators placed on a lot of application examples.... maybe I could do tha ton my board even tho it's not impedance controlled?