The Use of 90 Degree Traces in Power Distribution Routing

[ChanceMcCrance]

Hi All,

I have a question for all the PCB design experts and enthusiasts on the forum. Over the past month I've been looking to step up my skills/knowledge in the realm of power distribution in PCB layouts. One thing that I've seen in some examples I've found online is this routing technique in which a straight trace is drawn for a power rail (often wider than other traces on the board) and a number of traces branching off this one trace at a 90 degree angle distributing power to various parts of the circuit. It king of resembles a comb.

The notion of a 90 degree trace bend on a PCB being some evil thing that should be avoided at all costs has been called into question quite a bit over the recent years. I understand there may be instances where it might be an acceptable solution or even the best solution.

The following questions come to mind:


- When is this power distribution routing technique acceptable to use?

- What are the circumstances under which this technique would be the best option for routing power? 

- When this technique is used, what are the best practices for its implementation?

Any insight would be helpful as I've struggled to find articles online explaining this.

Thanks,
Chance

[thm_w]

Right angle is always acceptable functionally unless you are dealing with RF. 100MHZ+ stuff. Its been a myth for decades.
https://www.nwengineeringllc.com/article/right-angle-pcb-traces-its-time-to-kill-the-myths.php

You can read some more myths here: https://www.circuits-central.com/blog/6-common-myths-about-pcb-design/

[Alex Eisenhut]

Every via is a 90 degree trace in the Z axis. I don't know why these odd PCB myths persist.

[redkitedesign]

90 degree angles can be a bit harder to manufacture. However, most PCB shops can do them without problem nowadays.

HF signals don't like the impedance variation in corners, so round traces can be better. And indeed, via's are bad for HF too.
But depending on the board size (trace length) and actual signal, you might be fine until way past 1GHz.

[james_s]

I avoid 90 degree traces unless doing so requires a significant effort or compromises something else only because I think it looks nicer and I take pride in my work and try to do nice tidy layouts. If you're doing something where a sharp 90 degree bend is easier or has some other advantage then go ahead and do it, there's no law against it and as has been said, it isn't a problem for modern day manufacturing.

[BravoV]

HF signals don't like the impedance variation in corners, so round traces can be better.

Nope, just nope.

Supporting facts ...

From Texas Instrument's datasheet.

Click image to enlarge.



If thats not enough, watch EEVBlog #569, on Dave's site visit at RF transmission room, they use "right angle" instead of just "curved" pipes to transmit the so called RF signal in there, in high power density.

[redkitedesign]

HF signals don't like the impedance variation in corners, so round traces can be better.

Nope, just nope.

Supporting facts ...

Better be carefull with your facts. That TI datascheet is about microcontroller circuits, and <1GHz/>100ps risetime signals.  Thats not where impedance matching is a big issue, let alone trace angles.

And you cannot compare the outside of a waveguide with a PCB trace. Especially not a waveguide thats that big; 10cm in freespace is only 3 GHz.

That being said, by the time you should worry about trace angles, you should worry about a lot of other details too. If you can still work with normal PCB material, I  wouldn't worry.

[CatalinaWOW]

I suspect the 90 deg 'myth' started back in the days when PWBs were literally taped out with Rubylith.  Making a 90 degree turn would require either a fold in the tape, or a cut and overlap joint.  Either technique leads to a gap under the tape and risk of light leaks that could result in thin spots or undercutting.  Those days are long gone, even in homebrew boards.

[alm]

I don't think that much has changed since Howard Johnson wrote Who's Afraid of the Big Bad Bend?.

[TheUnnamedNewbie]

HF signals don't like the impedance variation in corners, so round traces can be better.

Nope, just nope.

Supporting facts ...

If thats not enough, watch EEVBlog #569, on Dave's site visit at RF transmission room, they use "right angle" instead of just "curved" pipes to transmit the so called RF signal in there, in high power density.

This is just demonstrably false at high speeds. There is a reason why people spend time simulating corners with miters in EM simulations for RF applications.

[2N3055]

Hi All,

I have a question for all the PCB design experts and enthusiasts on the forum. Over the past month I've been looking to step up my skills/knowledge in the realm of power distribution in PCB layouts. One thing that I've seen in some examples I've found online is this routing technique in which a straight trace is drawn for a power rail (often wider than other traces on the board) and a number of traces branching off this one trace at a 90 degree angle distributing power to various parts of the circuit. It king of resembles a comb.

The notion of a 90 degree trace bend on a PCB being some evil thing that should be avoided at all costs has been called into question quite a bit over the recent years. I understand there may be instances where it might be an acceptable solution or even the best solution.

The following questions come to mind:


- When is this power distribution routing technique acceptable to use?

- What are the circumstances under which this technique would be the best option for routing power? 

- When this technique is used, what are the best practices for its implementation?

Any insight would be helpful as I've struggled to find articles online explaining this.

Thanks,
Chance

Not fully defined question and bunch of loosely related and unrelated answers ...

Please, explain what PDN network is for (the actual application?).
For a retro computer with few TTL chips in PDIP it is going to be vastly different answer than for a very fast CPU or FPGA with 100W of power and 100ps edges...
And everything in between....

What you are describing would be something like TTL power net in the olden days..

Today with very fast logic and highly integrated chips, you need to start with proper PCB stack (many PCB layers, dedicated to power or data or combined and their layout) and then go from there.