Layout curved tracks/traces


#1

Hi Community,

has someone tried to do ‘curved traces’ ?
i have a round board (a 90° part of circle) and would like to use round traces / tracks for the powerlines…

i only found the preference option to allow more than HV45… but that does not make it round :wink:

sunny greetings
stefan


#2

Can’t be done.

Best alternative: approximate with enough line segments.


#3

I would love to get this feature though! A shame that it is not implemented. In my opinion the user should be able to “paint” tracks however wanted (just like it is possible with the board outline). Tracks (the main thing on a PCB!) that just allow 4 different angles are pretty limiting…

Airic


#4

@madworm: yes i have painted on the comments layer an arc and than tried to trace this with enough segments. it worked ok… -
only thing that did not look nice is the only 32 segments of the copperfill…
i think at some point in time i have to get into (learn) coding :wink:

@Airic: you can paint in every angle (ther is a option in the preference to switch between H/V/45 and arbitrary)

only arcs are not supported till now… but comes time comes nice features…

sunny greetings


#5

@s_light: thanks for the hint! I need to check that - but curved traces are still placed up high on my feature wish list.


#6

Interesting that the differential pair routing update uses automatically radiused traces when meandering is used for trace length matching. At 5:57 in this intro video it shows tuning of the meandering settings which made me immediately think of this forum question.

So curved traces are not excluded by the underlying KiCad codebase, at least in the bleeding edge nightlies. Bodes well for the feature being possible in future I think.

Geoff


#7

curved tracks:
+1 for this feature


#8

It’s been discussed a few times. First we need a good enough 2D geometry kernel but I don’t know what’s happening. I was told some time ago that work is being done on the geometry code but I never saw a plan.


#9

This feature could help in better RF circuits actually AFAIK: https://youtu.be/cQ7p7n1gp3E?t=11m.

Unfortunately I cannot see it clearly stated in the roadmap for KiCAD v5 nor the one for v6. :confused:


#10

Oh youtube video… Well, to put it as politely as I can, I think he is just repeating engineering “folklore” :slight_smile: His explanation about energy sounds pretty bogus to me.

People who have researched the topic more scientifically conclude that sharp corners are insignificant overall.
https://edeca.net/post/2012/06/right-angle-corners-for-pcb-traces/


#11


http://www.ti.com/lit/an/scaa082/scaa082.pdf
pg. 14 ti doc
High-Speed Layout Guidelines


#12

Those papers are basically repeating the folklore, but in a more formal setting. Consider the phrase

A right angle in a trace can cause more radiation

“Can” means “there is a possibility of”. No number specified.
“more” is unqualified. 1% more is more.

So it could mean “there is a 1% chance of 1% more radiation”, in engineering terms it is useless. The paper then goes on to explain

The capacitance increases in the region of the corner, and the characteristic impedance changes. This impedance change causes reflections

Those are plausible reasons, but without any numbers, pretty useless. The paper then dives into recommendations

Avoid right-angle bends in a trace and try to route them at least with two 45° corners.

Oh, so no EMC test report from anechoic chamber in controlled conditions, for a range of different frequencies?

If I gave that to my boss, he would just toss it back and say “get me some numbers”.

There guides are usually written by FAEs, who are often good engineers, but ultimately are just repeating stuff they have heard others say. Without some numbers, I really wouldn’t trust them any more than the guy on youtube.


#13

Texas Instruments folklore … not IMO


#14

Having worked for some manufacturers, let me tell you a little secret: they hire from the same pool of engineers as every one else

In my experience, customers often know more about a companies products than the manufacturer does. We analysed some modules which were giving faults. There are companies that specialise in analysis, they pot the components and take photos at microscopic level, etc. Really detailed stuff. They found micro fractures in the bond wires, which caused failures.

We took this to the manufacturer, who initially denied any problems. Eventually we got to talk to their technical guy, and he said “this is really interesting stuff!”. We asked what their analyses showed, he said “oh, we don’t do anything like this!”

Turned out they had recently changed manufacturing to a “cheaper location”, sourcing cheaper materials, with people who lacked training.

Folklore is rife at manufacturers too. Basically they are folklore plus added sales pitch.


#15

no comment
(these are characters to let the comment to be published :smiley: )


#16

Good engineering should be based on data, not “appeal to authority”. If there is a good case, then it should be easily made with data to back it up. Not just “Texas Instruments said so”.

I’ll bet you a crate of beer the guy writing that guide had never done or seen a test report on the effects of different shaped traces. There are only two charts in the TI guide that are based on real data, the rest is theoretical.


#17

sorry not being as good engineer as you are …
I’m just trusting an official “High-Speed Layout Guidelines” from a leading manufacturer of products I use…
Not trying to do myself measurements nor build chips or whatever…

Anyway everyone can fortunately behave as he/she desires in his/hers own business


#18

Right, that’s exactly how the folklore spreads, engineers trust other engineers who are no more competent than themselves. We do our own EMC testing, so we can verify these things, but I guess a lot of companies outsource such testing.

I’m not saying you are a bad engineer… but right angled corners have been proven to not make any difference to EMC. Making traces curved makes no improvement, but it doesn’t make it worse either. I guess it is like throwing coins into a fountain for good luck, it doesn’t work, but does no harm either.


#19

this conversation is just too tedious for me…
you can stay with your great opinion of Ti engineers writing down what they heard, may be drinking a beer instead of elaborating what they learned :smiley:
I’m just quitting
M


#20

Did you read the paper that was linked earlier? It is quite short and easy to read. Even if you don’t read the whole thing it is worth reading the conclusions.

There may be other good reasons for implementing rounded corners, but we can be sure that better EMC is not one of them.