Hacker News Reader

Comment by evck | Hacker News Reader

Yep, it's an OSHPark render. I'm not quite ready to release the design files, so I figured I'd post that picture since it's nice an purple.

I'd also recommend OSHPark.

View on HN

Good luck with the waveguide! Have you done anything that high-speed with OSH Park before? I'm waiting on a PCIe board from them that I hope will do Gen3 speeds (8GHz) over 2.5-inch traces, but the lack of impedance control has me nervous.

I suppose the waveguide looks pretty short (under an inch?) so it might not matter much?

Haven't done anything high speed, but I'm not too worried since it's short. It might not be perfect on the first run, but hopefully it'll work.

8 GHz and multiple traces would be a bit more concerning without impedance control. Hope it works out.

To be clear, my differential pairs are impedance controlled in the sense that their trace/space yield the correct single-ended and differential impedance values at the nominal dielectric permittivity and thickness, but OSH Park doesn't additionally guarantee permittivity and thickness tolerances to guarantee that my nominally-50-ohm traces will be actually-50-ohm traces +/- 10%. That said, I've run the numbers (using the excellent and free Saturn tools [1]) for the stated Er range of 4.2 +/- 0.3, and a little variation in the nominal 7.8 mil thickness, and things still looked OK.

Fingers crossed!

[1] http://www.saturnpcb.com/pcb_toolkit.htm

Reply on news.ycombinator.com