I’m finalizing the length tuning of HDMI signals on my PCB (KiCad v9), but getting strange results on one of the pairs. All three pairs plus clock are routed similarly, but in one pair I’m reporting a skew of over 300mil…?!
When I use the single length tune function to measure, it thinks one trace is 3588, and the other 3808 mils. Further checking shows that the trace that it thinks is 3588 is actually reported as 3822 (full length), which is more in line with the expected skew.
How have I “lost” that much trace is my design?
Guess it’s back to the old fashioned hand-tuning for now…
Or to some issues in the length tuning track following algorithm that are fixed for 9.0.3. Can you try with the latest testing builds? If the issue is still there, it’s probably due to the calculations differences fixed for v10 in the link above.
The issues remains on newer builds. It seems there is something in the trace; if I rip it up and re-route, it’s OK. That happened on a few different interfaces I was routing. It’s either OK, or way off.
In that case, there may be colinear / overlapping track segments. Does running the track cleanup tool and removing colinear segments help? If not, can you attach the project (or a minimised example) in a gitlab issue please?