I am attempting to determine the trace width needed for routing a differential signal of 100 Ω on a PCB I am designing.
This is a hobby project and the differential traces are for gigabit ethernet.
I'm very tempted to just forget about it, and just go with that manufacturer's numbers, but I am curious why different calculators spit out such wildly different numbers.
I found some EE exchange posts which provided unsatisfactory answers:
The second link seems to hint that some calculators take some extra parameters into consideration.
Having never routed differential signals the size difference these calculators are coming up with is a little alarming but perhaps this is normal?
Parameters:
- Trace Separation: 4mil
- Trace thickness: 1.4mil
- Dielectric Thickness: 3.5mil
- Er: 4.05
The manufacturer provides a stack up as well as a 2D calculator for determining trace width: here
The manufacture's calculator computes 3.49mil trace width.
When I attempted to validate this against other online calculators I find that some agree and some differ substantially.
EEWeb: Says the same numbers from my manufacture would be 160ohms!
Everythingrf: Agrees with the manufacture.
Can someone explain why they are different and what I need to consider when deciding trace width using these tools?
Best Answer
JLCPCB and EverythingRF use the same formula which only applies to circumstances where the trace-width-to-dielectric-thickness ratio is less than 1 (W/H < 1).
(EEWeb has a good explanation on their webpage but I didn't check if they use the same formula).
In your case the ratio is quite close 1: Trace width (W) is 3.49 mil, and dielectric thickness (H) is 3.5 mil. The ratio goes beyond 1 as the separation gets higher. So if you find that the dielectric thickness is less than or equal to the required trace width then you should review the stack-up.
JLCPCB's first stack-up recommendation for 1.6mm PCB seems correct to me but the 2nd one doesn't.