13

Say I have a trace running across a board. It's 50 mils the majority of its length, but in one short place it narrows to 25 mils to make it through a tight area. As best I can tell, this will be preferable to a 25 mil trace the same length, and only slightly inferior to a 50 mil trace without the few percent of its length narrowed to 25 mils.

Is there any disadvantage to the narrowing? Odd high-frequency effects? EMI? Obviously traces have many possible uses, including delivering power, carrying signals of different frequencies, grounding... so under what circumstances will this matter?

Stephen Collings
  • 17,373
  • 17
  • 92
  • 180
  • It's the same as if you were going from a smaller wire gauge to a larger one. The entire wire, or trace in your case, will only be able to handle the amount of current that the smallest section can, the "bottleneck" if you will. So your trace would be limited to the current capacity of the 25 mil – krb686 Jan 31 '14 at 15:07
  • What kind of signal is on this trace? If it is a 5GHz signal, yes the discontinuity is significant. In the end it depends on how long the run is. If the run is only 4 or 5 inches and you're only carrying a few mA then the whole trace can probably be 25 mils – HL-SDK Jan 31 '14 at 15:08
  • 1
    @krb686 True. But the whole trace will still have a lower impedance than a 25 mil trace would, meaning it would still have advantages for (say) signal grounding. Right? So as HL-SDK points out, it very much depends on what you're doing with the trace. Updated question to reflect this. – Stephen Collings Jan 31 '14 at 15:09

3 Answers3

10

Yes, but these disadvantages may be negligible.

Disadvantage 1: High frequency signals encounter a discontinuity.

I would start worrying at a few hundred megahertz because the change in trace width changes the characteristic impedance (not just dc resistance) of that line. The discontinuity changes scattering parameters, creates harmonics, reflections, and other headache-inducing problems.

Disadvantage 2: Voltage drop (and increased power dissipation) due to higher trace resistance.

If the decreased-width percent of the trace is less than 10%, I wouldn't worry. All of these effects can be calculated for your potential design, however.

Here's an online tool that helps estimate trace resistance

Here's a downloadable tool that has a lot of built-in equations

HL-SDK
  • 2,539
  • 1
  • 19
  • 26
9

For one thing, many PCB layout programs will allow, or automatically incorporate, "necking" of traces due to unconnected pads or keep-out areas. This is a reduction of trace width for a portion of the trace.

There are some concerns with such trace width reduction:

  1. If the reduced trace width is over an extended distance, then the increased resistance of the narrower trace will give rise to more heat, and will dissipate generated heat less easily, than the wider trace. For brief neck sections, this is not so much of a concern, as the heat gets conducted out to the wider traces on both sides of the neck.

  2. The narrowest trace width is the one determining how much current can be borne by the trace. If the narrow trace is still wide enough, then for moderate signal frequencies it isn't a major problem to have the trace equally narrow throughout, instead of having wider sections at all.

  3. Signal impedance and signal reflection issues as pointed out in comments and other answers - specifically for higher frequency signals.

Anindo Ghosh
  • 50,188
  • 8
  • 103
  • 200
3

If your are dealing with high frequencies ( around 100 MHz and above ), it will definitely matter. The change in trace width will be seen as a discontinuity, causing mismatch and ultimately resulting in unwanted reflections. You will see its effect on the timing edges and hence the digital I/O levels.

EMI would depend upon the layout routing and the isolation ( or rather lack of adequate isolation ) between adjacent traces. Stripline Vs Microstrip.

For low frequency operations, the major factor to take cognizance is the amount of current carried by the trace and heat. The safe current carrying capacity of the trace is determined by the narrowest section of the trace.

From the data you provided, using 50 mils trace...it seems like you are planning for a high current application. For a standard FR4 1 oz copper, 20 mils is good for 1A ...stripline routing . Others sometimes use thick traces for robustness in production.

Roshan Shanbhag
  • 81
  • 1