Trace separation - when and why does it matter?

[cprobertson1]

Good Morning folks, I'm afraid I've got a bit of a noob question here!

I've just finished laying out a board (a small multipurpose rate/counter module - just an RTC/uC with some settable filters and debounce depending on the frequency of the input pulses - might add SD card support for logging at some point as well)

Now my noobie question is simple: when and why does track spacing matter?

Off the top of my head I came up with a few reasons where it might matter: but what I'm worried about are the reasons that I'm unaware of - have I missed anything? Are there any issues with my existing points? While I'm by no means a professional - I thought it might still be a useful learning experience ;

-PCB manufacturing limitations - even the best manufacturer can only reliably provide separation down to a certain size!

-Isolation of high voltages - don't want a high voltage rail right next to my logic rail!

-Avoiding/reducing RFI - don't want a high-frequency input/output right next to my DC reference!


Thanks in anticipation folks! Hopefully the question wasn't too daft

[paulca]

Noobie to Noobie I was reading in Small signal audio design book about track capacitance.

If you think about it, putting two peices of metal close together and putting a charge onto one of them creates a potential between them, this is the same thing as a capacitor.  It is actually measurable, but very small.

Also the electromagnetic field is three dimensional, so if it's actually an issue for you the advice was to put a ground plain either side, requiring a 4 layer PCB!

GND plain
Signal plain
GND plain
Power plain

I'm not sure how much of that would apply in your use case, but remember that capacitors pass AC signals.

[Vic20]

-PCB manufacturing limitations - even the best manufacturer can only reliably provide separation down to a certain size!

And reduced distances can have cost implications.
Sometimes the lower the track space / width, the more the manufacturer can charge for your circuit.

Also, track distances can give reliability issues. It's better no to go to the manufacturer limits if you can.

-Isolation of high voltages - don't want a high voltage rail right next to my logic rail!

Important for clearance (direct track distance) and creepage (distance over the surface).
Creepage can be incresead using slots.
For anything connected to mains there can be obligatory regulations about those distances.

-Avoiding/reducing RFI - don't want a high-frequency input/output right next to my DC reference!

I would call it crosstalk.
It can be modelled by a coupling capacitance.
For digital signals it mainly depends on the rising and falling times, not on frequency.
You need a complete model to see the efect as it not only depends on track distances.

I will also add controlled impedance tracks to the subject although it may not apply for simple PCBs.
For high frequency content lines you usually lay controlled impedance tracks and having a track too close can change the impedance and give bad signal transmission or generate EMI.

[forrestc]

Now my noobie question is simple: when and why does track spacing matter?

The most common place that track spacing really matters in my work is where you care about controlled impedance.   For instance, if you are running an ethernet circuit across the board you need to run certain wires next to each other to create a balanced pair.  The size of the conductors and the spacing between them and the spacing between them an the ground plane all play into the impedance - for ethernet in particular you need 100 ohms of impedance, and you use that to determine what will work based on your needs (type of board, etc).

There are similar issues with passing rf around a board, but since that is an unbalanced (Single signal w/ground) line, the engineering is a bit different.

One resource you probably should be aware of is the saturn pcb toolkit which is available for free.   It has calculators and other information for various things such as the spacing between conductors at certain voltages.  Calculators for impedance of various traces and balanced lines.   Calculators for sizing of traces which carry current, etc.   The url for this tool is at http://www.saturnpcb.com/pcb_toolkit.htm

[cprobertson1]

I would call it crosstalk.
It can be modelled by a coupling capacitance.
For digital signals it mainly depends on the rising and falling times, not on frequency.
You need a complete model to see the effect as it not only depends on track distances.

Ah! Crosstalk! That was the word I was looking for!

I was on the right track with what I was thinking at least (and that pun actually wasn't intentional btw) - I should really have termed it "noise" rather than "RFI" - as I was meaning any sort of noise being passed between traces (capacitively or radiatively... which is a real word apparently! Cool!)

I had actually forgotten about the involvement of the rise/fall time in crosstalk - I would have just termed it "noise" and tried to deal with it from there: rather than worrying about the trace capacitance and resistance acting as a differentiator/integrator and whatnot.

I will also add controlled impedance tracks to the subject although it may not apply for simple PCBs.
For high frequency content lines you usually lay controlled impedance tracks and having a track too close can change the impedance and give bad signal transmission or generate EMI.

Ah, yes - luckily I haven't had to deal with that yet - I might do a radio-based project in the near future (as in within the 12 months or so) just so I can try out all this RF voodoo I've heard so much about - depending on what I do I can also work in some practice with controlled impedance inputs. That should also give me a chance to work on impedance matching and related topics.

Of course, the real problem I've found with impedance is trying to say it in such a way that the person I'm talking with is unambiguously aware that did NOT just say I was having problems with impotence in my latest project...


Anyhoo - thanks for the info folks! This is one of the problems with being self-taught and only working in a hobbyist capacity - lots of things that are easy to glance over (if you're only looking at the practical side and not the theory that is!)

I should probably go about getting a formal education in EE at some point - the problem is finding the time and money to take a course

[Nitrousoxide]

Yes, a small amount of capacitance will form between two traces depending on their length, spacing and PCB dielectric. However, the question is whether the capacitance will affect the signals present. To answer this, the best solution would be to perform a simulation for reliable results, it is also possible to do some ballpark calculations. It is of note that trace capacitance varies if power planes are used in the design. (It would be in the picofarads range).

Another thing that most people gloss over or simply don't consider is, in fact, the negative space between tracks.

Regardless of if you have a mixed system of analog, digital, high or low dI/dt and dV/dt. You must consider the ground/return paths. When designers don't consider this they may accidentally form a current loop and for fast, high (relative) currents (such as the PWM of a motor), this may cause a radiated EMI issue and magnetically couple itself into other traces that are present.


tl;dr: Capacitance is probably insignificant, simulate if worried. Don't forget about well-defined ground return paths.

[ocset]

This is a good ref on this subject...

http://www.alternatezone.com/electronics/files/PCBDesignTutorialRevA.pdf

[cprobertson1]

This is a good ref on this subject...

http://www.alternatezone.com/electronics/files/PCBDesignTutorialRevA.pdf

Bedtime reading: sorted 

[paulca]

This is a good ref on this subject...

http://www.alternatezone.com/electronics/files/PCBDesignTutorialRevA.pdf

Thanks   Bookmarked.  Shame it uses an old and silly measurement system based on the size of kings fingers or something.