WIZnet Developer Forum

WizFi630-EVB impedance matching

hello,
I already started to develop a gateway with this wireless module, ref WizFi630, and it seems for me to not have great difficulty to adapt WizFi630-EVB circuitry to our needs. It happens that I’m with some doubts with the correct dimensiong of the parameters of the high-speed signal lines(ethernet signals), and I don’t want to make mess with PCB layout. In did that is the only part of the pcb projcet that I didn’t finish.

I kindly ask if it is possible to access the gerbers of this evaluation board, ref WizFi630-EVB, or the pcb project. I don’t want to copy the board, I just to want to confirm my doubts and my calculus about lines impedance(I assume it’s 50ohms), and differential impedance of the pair lines.

If it is not possible to access the gerbers or the PCB project I would like to ask if it is possible to know the following parameters of the lines in this pcb board.

  • h (thickness of the FR4 dielectric)
  • Er(electric permitivity of the dielectric)
  • spacing between differential microstrip lines.
  • spacing from gnd to microstrip lines
  • width of the lines
  • thickness of the copper laminate

Best regards,
Vitor Barbosa

Hello Vitor

I just already sent to you some files related with WizFi630-EVB.
Please refer to them.

Hello,

in the ethernet signals from WizFi630, according to the pcb line dimensions my line impedance calculus for:
w=0,2mm (width of the lines)
d=0,2mm (space between differential lines)

And to Z0=100ohms(±4ohms) I have:
h=0,5 to 0,6mm (thickness of the dielectric) (my estimation)
er=4,2 to 4,7 (electric permitivity constant typical on FR4 laminates)(my estimation)
Zdiff=130ohms (differential line impedance on line pairs)(my estimation)

Is it possible to confirm it?
Best regards,
Vitor Barbosa

If you get through impedance calculus and keep routing rule of pair-line,
I think that is no problem for operation. But, We can’t guarantee.

Thanks

OK,

I have worked a big part of my life in PCB design for high-speed signals(from 622MHz to 10GHz) and a 100MHz signal on FR4 should have a wavelentgh in the order of 1.2m to 1.6m, depending on th type of dielectric. So, in this conditions(100MHZ,…)if the line distance from the module to the transformer is less then 10cm a reflection from a mismatched line should be negligible in the time domain.

Anyway, as I ignore what are the default parameters for ethernet drivers I have searched a lot and apparently it is assumed among different manufacturers typical values for the ethernet drivers:

  • single ended impedance Z0=50ohm for eache line of a pair
  • differential impedance Zdiff=100Ohm between the lines of each pair

I don’t know if this is true for wiznet ethernet drivers but I want to think that this is also recomended for them.

So, my conclusion for the Wizfi630-EVB differential lines is that they work not for having correctly matched/dimensioned lines but for exhibiting negligible reflections.

Based on gerbers analysis, this EVB only has two routing layers and no power planes, so for having a PCB stack with
w=0,2mm(microstrip line width)
d=0,2mm(differential microstrip line spacing)
t=17.5um
er=4.2
h(dielectric high) sould be 0.12mm for a Z0=50mm and Zdiff=100ohm. This h value is not reasonable for manufacturing on a standard FR4 laminate.

These simulations and calculus were made on free Saturn PCB toolkit and confirmed with Txline.

The only thing I would like to know, as it not detailed on WizFi630 documentation, is if it is recommended to have Zdiff=100Ohm and Z0=50ohm on differential pairs. According to WizFi630 schematic this assumption seems to be coherent.

Best regards,

Vitor Barbosa

[quote=“ssekim”]Hello Vitor

I just already sent to you some files related with WizFi630-EVB.
Please refer to them.[/quote]

Hi ssekim,

Could you please send me those files via email as well? I’m doing the same kind of work and I could really use them.

Cheers,

Pablo C.

Hello Pablo

I just sent the files to you.
Please refer to them.

Copyright © 2017 WIZnet Co., Ltd. All Rights Reserved.