Why you don’t make right-angle traces and why lightning rods are pointy

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
CJ said:
nice read, thanks!

that stuff gets real important when you get up into the higher frequencies,

Up until recently I worked in the film industry as a senior graphics programmer.  A colleague of mine who was also a film graphics programmer recently moved full-time into electronics manufacture of GPS systems for the same film company I used to work for.  He has been working on a high-quality high-speed GPS receiver.  The first version simply didn't work, and after some debugging it was down to a single right angle trace leading from the GPS surface mount chip to the antenna connection.  He changed it in his revision 2 board to a curve and it worked perfectly.  He was the person who sent me the original link.

Yep, this stuff is fascinating.
 
etheory said:

It's total BS, especially if you consider how copper on a PCB is etched.

The article (and the rest of the site) read like they've been written by an enthusiastic college kid. There has been plenty of research on the (non)effect of right angle bends, like this paper and this paper.

While sharp corners can have some manufacturing issues (acid traps), the myth about 90 degree corners being bad needs to die.

JD 'pet peeve' B.
[RF can sometimes be a special case, but if one single right angle trace adds enough capacitance to break a design then it was pretty marginal to begin with. And from my experience with designing GPS receivers, if you're seeing anything it's more likely to be bad quality control of the dielectric constant of your PCB substrate]
 
jdbakker said:
It's total BS, especially if you consider how copper on a PCB is etched.

The one thing that those papers you link to mention that I only just thought of then, is that the junctures between the tracks and the legs of the components would create far more of a theoretical issue than the track geometry would.  BS or not ( and you are quite likely right, it could well be ), I like to think it's always good to get info on stuff from all angles.  Thanks for the response!  Gonna give it all a good read through.
 
I just finished a prototype layout with the poor electrons making a few tens of 90' turns (north-south top traces jumping to east-west bottom traces at vias) to get around. But I am not making a GPS radio receiver or clocking at GHz so I expect it will work.

There is an all too familiar pattern in audiophllia of latching on to some real physical phenomenon, then extrapolating this down into the audio bandwidth where it is insignificant. Look at all the funny wire designs based on esoteric out of band effects. 

For very fast digital clocked circuitry layout can surely matter.... If you're worried about the electrons falling off at sharp right hand turns, coat the traces with an extra layer of conformal coating (just kidding guys).

JR 
 
What about all the billions of sharp right turns in nm-pm (and lower) scale from the etching process? Think shore line and sand. Why would the electrons care of an absolutely massive (0.1mm) right turn of your PCB design?

Perhaps the writer of that paper should step down to the factory floor every once in a while?
 
etheory said:
jdbakker said:
It's total BS, especially if you consider how copper on a PCB is etched.

The one thing that those papers you link to mention that I only just thought of then, is that the junctures between the tracks and the legs of the components would create far more of a theoretical issue than the track geometry would.  BS or not ( and you are quite likely right, it could well be ), I like to think it's always good to get info on stuff from all angles.  Thanks for the response!  Gonna give it all a good read through.

Another consideration is that many multilayer boards are routed with one layer biased towards traces going horizontally, and an adjacent layer routed vertically. The jumps through the vias between the layers are 90-degrees.

Also, nothing in his article mentioned routing traces over planes.

-a
 
there are machines that measure propagation time of curcuit borad traces for hi-freq apps, and there is an issue with corners.

also issues with inductance.  trace size and length, amount of copper, microwave freqs are sensitive to board layout.

cell phones and wi-fi gear use traces as antennas,

some clever usage of traces can be found on the Langevin AM-16 circuit board, top and bottom traces run over each other to form balanced input capacitance of about 10 pf. 
 
CJ said:
there are machines that measure propagation time of curcuit borad traces for hi-freq apps, and there is an issue with corners.
but not important for audio frequency bandpass...
also issues with inductance.  trace size and length, amount of copper, microwave freqs are sensitive to board layout.
this can also matter for high current audio, like in power amps.
cell phones and wi-fi gear use traces as antennas,
yup
some clever usage of traces can be found on the Langevin AM-16 circuit board, top and bottom traces run over each other to form balanced input capacitance of about 10 pf.
In the margin, everything about traces matters... There are also tricks associated with audio frequency trace pairs to reduce effective loop area, but this is a veer from the topic of traces that veer.

JR
 
CJ said:
there are machines that measure propagation time of curcuit borad traces for hi-freq apps, and there is an issue with corners.

The papers I linked to (this paper and this paper) show the results of both Time Domain Reflectometer and frequency domain measurements on traces with and without corners. They all show that there is no measurable effect for corners, at least at speeds/frequencies where FR4 makes sense.

There are plenty of nasties at RF, including pins/pads, vias, slots in reference planes, variations in substrate dielectric constant and even the direction of the weave of the FR4 material. 90 degree corners simply isn't one of them.

Sorry to keep harping on this, but this myth needs busting.

JDB.
[of course, if anyone has any links to peer-reviewed material saying otherwise, I'm all ears]
 
Okay... having made/designed PCBs and microchips for cell phone radio section power amplifiers (6 Watts for just under 2GHz or so including DC bias with leaky RF issues on the same traces)...

yes, right angles make a difference at these RF frequencies, and since we were pushing the envelope, the design was *VERY* touchy to the point where the division went out of business because we could not get it to go in 2 years time with some wacky technology that we were exploiting... These amplifiers must meet a spec for 10:1 VSWR (reflections and standing wave ratio stuff), because even with a fractal or patch antenna, the user is going to throw their cell phone into the metallic ash tray in the car or in their bag with a mylarised spent potato crisp bag or under their tin-foil hat ;) ....

At audio frequencies maybe not so much...  Although I always, when feasible, expect to carry audio and low frequency RF as necessary, such that stray or errant RF can be handled gracefully (whether emphasized on purpose, say with high-speed analog portions of a discrete A/D design before quantization with a latching high-speed comparator or for proper attenuation of any RF effects - depending on the circuit needs)...  It really does not cost much more to be RF conscious during analog audio design.... 

Remember that if you want to faithfully reproduce a 20kHz square wave, the circuit needs to reproduce all the higher harmonics... Although I submit that this is not ideal or practical as folks might not be listening to music on high bandwidth reproduction equipment that has 20kHz square waves (even with digital stuff these days having the inability to reproduce harmonics for 11kHz square waves; although there may be other effects happening here in phase and reflection back into the audio domain)...

However, this is more important for digital circuits in "trying" to maintain a square wave at whatever high-speed digital frequencies and in fact it becomes an analog problem at high frequencies where the square wave is rounded over and might be ringing with all the stray inductances and capacitances on a microstrip style of trace....  And of course the FCC or other radio governing body might get involved with square wave harmonic hash being generated....

However, routing PCBs with angled traces is also ideal for routing topology and shorter traces in theory....  By definition a^2 + b^2 = c^2 and thusly a+b >> c where if the trace was diagonal rather than manhattan style (right angles) then it would be shorter....

For high voltage we were concerned about sharp corners in the microchip world, especially with VSWR stuff reflecting energy back into final stage transistors on GaAs and creating higher voltage standing waves... The same is true in the PCB world, as I have had a Fender Hot Rod Deville arc over on the PCB traces....  Copper threads can electrolytically "grow" to more sensitive high gain traces nearby....

Anyway.... sorry for the tome.....



 
A 20khz square wave is not audio...

nor are the rising and falling edges of any square wave.

JR

Ps yes a little ironic that at very high clock frequency it all becomes analog again...  8)
 
JohnRoberts said:
A 20khz square wave is not audio...

nor are the rising and falling edges of any square wave.

Ha!  Then you never listened to my college radio station ....  Or maybe we were off the air... can't tell ;)

JohnRoberts said:
Ps yes a little ironic that at very high clock frequency it all becomes analog again...  8)

Irony... it's more than just how the well water tastes around here....
 
I do RF design, mostly <1GHz but we work a lot with GPS receivers.  We just did a layout recently where all the microprocessor and digital circuits were laid out exclusively with right angles, including things like serial and I2C traces.  When it came to the analog RF sections we went with 45 degrees but 90 degrees wasn't off limits either.

I went into the design thinking 90 degree traces was blasphemy and would ruin everything, based on vague memory from school, but the old scientist types assured me it was no big deal... 
 
Hi,
thanks, JDB. and others for demystifying the right-angle-myth.
It's all about scale issues if things like the gauss' law's 'interpretation' of the first post's link do affect functionality even in much higher than audio frequencies or not. It was just applied to a much too generalized context - not useful at all. So - in most cases we may be interested in here at this place, it does absolutely not affect the functionality of a pcb,
but maybe the asthetics...
8)
Best regards,
Martin

 
smallbutfine said:
but maybe the asthetics...
8)
Best regards,
Martin

Hey even that's debatable -- the MPU/digital section of the board we just laid out looked pretty sharp (ha!), and it actually helped us fit more traces into a tighter space.  I thought it looked more organized and planned out than the 45 degree stuff.
 
+1  I just did a design with all east-west traffic on one layer, and north-south on the other. Most effective way to get from point X to point Y with only 2 layers  (and without having to think  ;D ).

This was just for a development platform, so for a production design I would have invested more thought and effort into the routing.

JR
 
Back
Top