Super tight curves and conversions

Since KiCAD has added curves, I can already see sharp limits of what that engine can do. I would like to either be able to draw tighter, more dynamic curves or have a function that looks at a line from component to component as laid out by the user and attempts to convert all angles to curves without violating boundaries of other lines or components.

I see that there are two modes for curved lines. But when drawing in tight spaces, they are both almost useless. This is more an issue when doing anything in RF.

I am familiar with the work around, to draw objects and boundaries in the copper area. And yes this works but there are bunch of issues, like copper fill doesn’t respect objects DRAWN, that is the boundaries are for the 45deg angle lines, not the manually drawn.

Additionally, the default is to align the center of all tracks with the center of the trace. This forces most paths to exit the pad at a UP/DOWN/LEFT/RIGHT config. While that is fine in most applications, in RF, I would like to shift that behavior so that the trace can be draw in at the edges of a pad or an angle

I am working on assembling some good graphics. Sorry they are not yet ready at the time I post this.

KiCad’s support for curvy tracks is not very good. Sometimes I think about looking for and/or creating an issue for this to ask for improvement, but I make so little use of curved tracks that it is a bit difficult for me. Such a report would require quite a bit of preparation, and I’ve also just started experimenting with KiCad-Nightly V7.99 (Which can apparently assign net names to graphic arcs, but I have not worked with that yet).

Maybe you can find some comfort in:

Also note that video is now almost 2 years old. After that KiCad has gotten native support for teardrops. From what I remember from some short experiments, that “Round Tracks” plugin also need some space to work.

That is an argument I’ve heard many times before, counter arguments say that under 6GHz or so, differences are negligible.

1 Like

I looked at the plugins for curved tracks and teardrops. But they were more than 2 years old and in that time we have gone from 5.x to 7.x version. I have no idea how that might effect those old plugins.

I know other software having not even gone through a single major version update, will find plugins broken without explanation and it creates havoc in the program.

Usually this is correct for reasonable track widths.

High voltage circuits can benefit from smoothed corners

1 Like

That is because of corona discharge which is a valid concern for HV, but completely different phenomena.

It is, but corona discharge and also the formation of tin whiskers with lead free solder under high electrical fields are real problems

1 Like

Tin whiskers are also real but they are yet another completely unrelated topic. This creation of whiskers is a quite weird phenomena. I’ve also heard of whiskers forming on zinc plated metals, such as the insides of 19" racks that can cause shorts when breaking off and falling on a PCB.

this video link does a download and has a bunch of info about tin whiskers.

I have also noticed this myself. Not as long whiskers, but on the zinc plating of small DC motors which form a strange white powdery residue on the zinc plating.

1 Like

There are certain flux substances I forbid my coworkers from using since it remains active. They liked to use it for soldering wires. We always covered the wires in Heat Shrink to protect. But months to years later, the devices come back for service and the leads parts no longer work. Inside the Heat Shrink is a white powder. The wires has been destroyed at the contact. No green like with Copper Rust, just white. I clean the contacts, re solder, reseal and all is well.

The thing about RF is that the behavior swings a lot depending on the power and frequency. If a wire is matched to the impedance and length of the wave, the more multiples you have the more stable the signal gets. The shorter the wire, at higher power the more unstable it gets and the slightest deviation from whatever the actual rules as governed by nature, the stranger the results get. In particular VSWR, and Phase can start to swing wild if everything isn’t just right.

The work I am doing is with less than 1v on the receiving. I think parasitic is the biggest threat. That is that the very weak return signal might be effected or even drained by a captive or inductive coupling from a nearby trace or part.

True, the traces measure maybe 3mm at best, but the sharp corners create tiny antenna and strait parallel lines act as either inductors or capacitors. It should be trivial. But I have no way to quantify that just yet. That is, I do not have a complete project with curves and without to compare the results and determine if the difference is indeed a nothing burger.