Author Topic: Measuring impdeance of a PCB trace  (Read 7303 times)

0 Members and 1 Guest are viewing this topic.

Offline jgalakTopic starter

  • Regular Contributor
  • *
  • Posts: 237
  • Country: us
  • KQ2Z
    • Blog, mostly about learning electronics.
Measuring impdeance of a PCB trace
« on: December 20, 2017, 03:32:32 pm »
In one of the videos on design for manufacture Dave shows a PCB panel with an impedance test strip, presumably to allow the testing system to verify that the correct impedance is present.

How does one measure this?  Is it possible to do so with simpler tools, without a VNA?
Blog, mostly about learning electronics: http://kq2z.com/
 

Offline IonizedGears

  • Regular Contributor
  • *
  • Posts: 248
  • Country: us
Re: Measuring impdeance of a PCB trace
« Reply #1 on: December 20, 2017, 10:44:39 pm »
Put a resistor in series with the trace and use a function generator to apply a signal of known amplitude and phase. Usually this means that you have to measure in circuit and the phase usually doesn't matter too much as you'll be triggering off the function generator for your oscilloscope. You should choose the resistance based on your expected line impedance such that you aren't practically shorting your function generator and that the your oscilloscope is sensitive enough to measure the voltage across the line. You should be able to measure the voltage across the resistor as well but you still need to be able to detect small changes in the voltage due to the transmission line. Keep in mind that the loading from your oscilloscope can effect these measurements. Also note that you need to choose a frequency such that effect from the impedance of the line is even noticable. In the end, you need to measure the voltage from your function generator and the voltage across the circuit element you chose to measure from and calculate it out as a voltage divider in the phasor domain.

P.S. I'm typing this on my phone and I have no idea the relative scale of the line impedance, the resistor, or the needed frequency to start seeing any transmission line effects in the trace. Nor do I know if this would work on a noticable level for your specific case. You basically need to find a frequency such that the relative short from the trace stops looking as a short (ignoring the resistivity of the line). Perhaps the resistor could be replaced with a known capacitance or inductance but to get accurate enough results you'll need a VNA. A vector network analyzer would be best for this.

See if you can find the transmission line parameters of the trace and calculate it out.

Sent from my Pixel XL using Tapatalk

I am an EE with interests in Embedded, RF, Control Systems, and Nanotech.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Measuring impdeance of a PCB trace
« Reply #2 on: December 20, 2017, 10:53:32 pm »
The trace impedance could be found indirectly by measuring the electrical length of the trace with a transmission line oscillator.  A time domain reflectometer can return the exact impedance but needs to be fast enough to handle the short trace length:


 
The following users thanked this post: rhb

Offline Kire Pûdsje

  • Regular Contributor
  • *
  • Posts: 65
  • Country: nl
Re: Measuring impdeance of a PCB trace
« Reply #3 on: December 20, 2017, 11:57:59 pm »

As for measuring: For lower frequencies, less than say a couple of GHz, the easiest way is just to place a big rectangle on the PCB (stripline or microstrip) and accurately measure its (DC) capacitance to ground, from this you can derive the dielectric constant. With a microscope, measure the actual trace widths/layer heights realized. Then put both of them in a simulator. (In your design, let some internal traces run accross the edge of the board, to get a cross section for free. Or in case of a production board, run them very close to the edge/corner, so you can use a file to get to the trace)
The mentioned method is also typically used for higher frequencies, albeit that the dielectric constant is determined with ring resonators.

I do not know what your aim is for knowing the impedance:
For digital circuits: if your board has layer heights of say 0.2 mm or less and lines that are of substantial length compared to the rise time, ask the manufacturer for controlled impedance lines. This is because etching tolerance and pressing will be major contributors to the impedance (apart from the FR4 itself), If using thicker layers, do not bother.

Microwave circuitry. Do not let the manufacturer control the impedance, unless you are only using 50 ohm lines. If any fancy stuff is happening, do net let them touch it. Just ask the manufacturer for the specific board stackup, including prepregs etc. Also ask for etching tolerances. When knowing the FR4-manufacturer, the dielectric constants will have a tighter tolerance than just stating FR-4. Use these values to determine the production tolerances.
When using multilayer boards, skipping ground planes for the RF part is the easiest way to keep the effect of etching tolerance under control.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3012
  • Country: gb
Re: Measuring impdeance of a PCB trace
« Reply #4 on: December 21, 2017, 06:00:04 pm »
The other way to do it is to use E and H field probes and a spectrum analyser with a tracking generator to characterise the transmission line based on a couple of measurements using the probes. This works for a variety of unknown transmission lines, including twisted pair wire, coax and also microstrip traces on a PCB. It works well if the transmission line is quite short, eg 15cm or so. From this you can predict the L/m and pF/m and also the Zo and the velocity factor (and the effective dielectric constant)
 

Offline jgalakTopic starter

  • Regular Contributor
  • *
  • Posts: 237
  • Country: us
  • KQ2Z
    • Blog, mostly about learning electronics.
Re: Measuring impdeance of a PCB trace
« Reply #5 on: December 21, 2017, 06:09:15 pm »
I understand that I can calculate the impedances.  What I'd like to do, however, is measure them.  I've used calculators to design them, but I don't know how well the PCB manufacturers stick to tolerances, or how accurate the assumptions I put into the calculator are.  My goal here is to design one from calculations, have it made by the same PCB house I plan to use for my boards, and verify that I'm getting the right impedance.

Right now, I'm working on things in the 144 MHz range (amateur 2M band), but I really don't have equipment that goes that high.  My function generator goes to about 14 MHz, my scope is 100MHz, and I don't have a spectrum analyzer (I can set one up on an RTL-SDR dongle, but that won't have a tracking generator and is not exactly well calibrated).

The technique from that video is one of the options I'm considering, but will it work with a short trace (a couple of inches at most)?  Will that be enough to make it work?

The resistor idea sounds great as it requires minimal hardware, but I have no idea how to do the calculations described.  Is there a good explanation of this technique somewhere (preferably on-line)?
Blog, mostly about learning electronics: http://kq2z.com/
 

Offline Kire Pûdsje

  • Regular Contributor
  • *
  • Posts: 65
  • Country: nl
Re: Measuring impdeance of a PCB trace
« Reply #6 on: December 21, 2017, 08:04:14 pm »
Unfortunately you need a trace length that is a substantial part of the wavelength/rise time. Otherwise you cannot easily differentiate between phase velocity and impedance. For short line lengths, where a transmission line can be approximated by its lumped equivalent, these properties can be traded. As an example the input impedance of a shorted line is Z=j*Z0*tan(beta x). where beta = 2*pi/lambda. For lower frequencies, the small angle approximation can be used, so Z=j*Z0*beta*x. As you can see, it is just a product. No (easy) way to separate the wavelength in the medium from Z0. (theoretically you could use the oc/sc method, but for short line lengths, you would need to be able to accurately measure the capacitance/inductance, furthermore it will probably be dominated by parasitics).
The fact that the transmission line can be seen as the lumped equivalent for short lengths, also gives freedom for layout. You can make narrower lines where it is convenient, and add the lack of capacitance somewhere else.

Purely practically speaking, say you wanted to use the TDR method with a step excitation, if the line is less then the rise time of the excitation, the responses of the source step and the reflection will start to overlap. You cannot separate the effects.
« Last Edit: December 21, 2017, 08:20:19 pm by Kire Pûdsje »
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3012
  • Country: gb
Re: Measuring impdeance of a PCB trace
« Reply #7 on: December 21, 2017, 09:09:36 pm »
Quote
(I can set one up on an RTL-SDR dongle, but that won't have a tracking generator and is not exactly well calibrated).
If you use E and H field probes the analyser doesn't need to be calibrated in amplitude, but it does have to be reasonably accurate in terms of frequency so you could use a sig gen and an SDR to replace the spectrum analyser and tracking generator. But it will be much slower and more prone to operator mistakes.

If you use E and H field probes you can also measure balanced or unbalanced lines. I guess a lot depends on what accuracy you expect. I'd expect the E and H field method to often outclass most alternative methods apart from a decent VNA based method of course.
 

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 7934
  • Country: us
  • Retired, now restoring antique test equipment
Re: Measuring impdeance of a PCB trace
« Reply #8 on: December 21, 2017, 10:01:07 pm »
If you have an impedance meter that measures complex impedance Z = R + jX, the following method works:
1.  Guess the electrical length of the line in nanoseconds.  Accuracy not important.
2.  Choose a frequency where the length of the line (in ns) corresponds to maybe 0.01 to 0.1 of the period (= 1/f) of the test frequency.
3.  Measure the impedance of the line at one end, with the other end shorted.  Zshort.
4.  Measure the impedance with the other end open.  Zopen.
5.  The characteristic impedance of the line is then given by (Zo)^2  =  Zshort X Zopen.
Use proper complex algebra to do the multiplication in step 5.  If it is good measurement, the produce will be almost totally "real", with a small imaginary component.
 

Offline Kire Pûdsje

  • Regular Contributor
  • *
  • Posts: 65
  • Country: nl
Re: Measuring impdeance of a PCB trace
« Reply #9 on: December 21, 2017, 10:42:25 pm »
The sc/oc method for say a 2 inch line would require you to measure 16 nH and 7 pF with high accuracy. For a say 5% accuracy, the uncertainties of the measurement should be o.o.m 1 nH, resp 0.5 pF. This will also be in the range of the parasitics.

Anyway, when starting to play with RF frequencies, besides an RTLSDR, might I recommend at least one of those adf4351 modules, which can be bought for say 15-20 euros's or 50-70 euros including control display, or 80-120 euros as a fully built instrument.
And an ad8318  log detector module as a power meter, prices around 10-15 euros
Furthermore a VSWR bridge, for about 5-10 euros.
With this combined set, you could generate a cheap manually controlled scalar network analyzer.
 

Offline jgalakTopic starter

  • Regular Contributor
  • *
  • Posts: 237
  • Country: us
  • KQ2Z
    • Blog, mostly about learning electronics.
Re: Measuring impdeance of a PCB trace
« Reply #10 on: December 21, 2017, 11:46:19 pm »
Oh, hey!  Impedance meter!  Can I use an antenna analyzer for that?  I have a good one that goes to 600MHz.  Need to play with the numbers you described....

I absolutely plan on a spec an, with tracking generator, soonish.  My next major purchase.  But not there yet.
Blog, mostly about learning electronics: http://kq2z.com/
 

Offline jgalakTopic starter

  • Regular Contributor
  • *
  • Posts: 237
  • Country: us
  • KQ2Z
    • Blog, mostly about learning electronics.
Re: Measuring impdeance of a PCB trace
« Reply #11 on: December 21, 2017, 11:55:42 pm »
Anyway, when starting to play with RF frequencies, besides an RTLSDR, might I recommend at least one of those adf4351 modules, which can be bought for say 15-20 euros's or 50-70 euros including control display, or 80-120 euros as a fully built instrument.
And an ad8318  log detector module as a power meter, prices around 10-15 euros
Furthermore a VSWR bridge, for about 5-10 euros.
With this combined set, you could generate a cheap manually controlled scalar network analyzer.

Are those modules actually good?  I was looking for an RF source and a power meter that goes to UHF (the power meters I have are all for under 50MHz).
Blog, mostly about learning electronics: http://kq2z.com/
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3012
  • Country: gb
Re: Measuring impdeance of a PCB trace
« Reply #12 on: December 22, 2017, 12:36:30 am »
I've not used an AD8318 log detector module as a 'power meter' but I'd be a bit concerned to use it with a ADF4351 module because of all the odd order harmonics the ADF4351 spits out. I'd expect this harmonic energy to introduce a fair bit of measurement uncertainty in the log detector. So you would typically have to filter the signal to get rid of the harmonics. You would probably want to filter the signal to get the harmonics at -30dBc or better. But I guess a lot depends on what levels of uncertainty are acceptable to you.
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2758
  • Country: us
Re: Measuring impdeance of a PCB trace
« Reply #13 on: December 22, 2017, 03:09:52 am »
I understand that I can calculate the impedances.  What I'd like to do, however, is measure them.
"Measuiring" is hard, but matching is easier.  You need a step pulse source, where the step rise time is significantly less than the electrical length of the strip.  Set up a T on the scope input, one branch of the T goes to the pulser, the other is connected to the strip.  First, if you feed the strip with a foot or so of known-impedance cable, the discontinuity between the cable and strip can be seen.  Then, to figure out the impedance of the strip, terminate the far end of the strip with short-legged resistors near the value expected for the strip.  When you match it exactly, there will be no reflection off the far end of the strip.

Jon
 
The following users thanked this post: jgalak

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Measuring impdeance of a PCB trace
« Reply #14 on: December 22, 2017, 04:00:30 am »
... my scope is 100MHz ...

Yea, that will not be fast enough for such a short trace for the reason Kire Pûdsje identifies and you would still need a fast transition time pulse generator.  Time domain reflectometers for this sort of thing typically operate at 50 picoseconds or better giving them a bandwidth of at least 8 GHz.  The last time I measured circuit board impedances, I used a 25 picosecond sampling oscilloscope which was good enough to start distinguishing different areas of a BNC connector including the crimp.  For FR-4 with a velocity factor of about 0.5, it comes out to about 170 picoseconds per inch.
 

Offline rx8pilot

  • Super Contributor
  • ***
  • Posts: 3634
  • Country: us
  • If you want more money, be more valuable.
Re: Measuring impdeance of a PCB trace
« Reply #15 on: December 22, 2017, 04:17:41 am »
I just purchased a 20Ghz TDR for this. Guessing there will be a learning curve and some pricey cables / adapters in my future.

Ancient HP behemoth - not even sure if it works.

HP 54120B with 54121A. If it works, it will be a bargain- it was cheap. Fingers crossed. Super curious if it will allow measurements of relatively short traces.



« Last Edit: December 22, 2017, 05:37:59 am by rx8pilot »
Factory400 - the worlds smallest factory. https://www.youtube.com/c/Factory400
 

Offline Kire Pûdsje

  • Regular Contributor
  • *
  • Posts: 65
  • Country: nl
Re: Measuring impdeance of a PCB trace
« Reply #16 on: December 22, 2017, 05:16:12 pm »
Are those modules actually good?  I was looking for an RF source and a power meter that goes to UHF (the power meters I have are all for under 50MHz).
I've not used an AD8318 log detector module as a 'power meter' but I'd be a bit concerned to use it with a ADF4351 module because of all the odd order harmonics the ADF4351 spits out. I'd expect this harmonic energy to introduce a fair bit of measurement uncertainty in the log detector. So you would typically have to filter the signal to get rid of the harmonics. You would probably want to filter the signal to get the harmonics at -30dBc or better. But I guess a lot depends on what levels of uncertainty are acceptable to you.
These modules are a good start, to be had for cheap. It is certainly better than not having any RF equipment. As for the log detector, you will need some reference source to determine the offset. Also most of the time relative measurements can be made. One should definitely be aware of the harmonic content of the signal, but LPF in octave steps can be easily created/bought.
I actually created a pretty clean 0-2 GHz sweep generator by mixing two of them together (one from 2.4-4.4 GHz, the other at 4.4 GHz) and adding a 2 GHz LPF.
 

Offline Kire Pûdsje

  • Regular Contributor
  • *
  • Posts: 65
  • Country: nl
Re: Measuring impdeance of a PCB trace
« Reply #17 on: December 22, 2017, 05:29:23 pm »
I just purchased a 20Ghz TDR for this.
Try to resist the urge to always test at the fastest rise-time/largest bandwidth. To make sure you do not excite higher order modes, check your measurements with various rise-time/bandwidth settings
 

Offline rx8pilot

  • Super Contributor
  • ***
  • Posts: 3634
  • Country: us
  • If you want more money, be more valuable.
Re: Measuring impdeance of a PCB trace
« Reply #18 on: December 22, 2017, 05:52:08 pm »
Thank you for the tip. The learning curve is has now officially started.

While I do have some reason to do these measurements, the primary driver is curiosity and general skill building.

Short and misplld from my mobile......

Factory400 - the worlds smallest factory. https://www.youtube.com/c/Factory400
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3012
  • Country: gb
Re: Measuring impdeance of a PCB trace
« Reply #19 on: December 22, 2017, 10:46:11 pm »
Quote
As for the log detector, you will need some reference source to determine the offset.

I'm not sure what you mean but for general use I think the only workable solution would be to filter the signal. Otherwise the harmonics can cause significant problems in a typical log detector. I think the harmonics from the ADF4351 can be -10dBc to -13dBc so this could easily introduce an error of 1dB or maybe 2dB in a log detector. But I've not used the AD8318 so I can't be certain how much it could be affected.  I don't think you can simply add an offset because the harmonics can make a typical log detector read high or low depending on the relative phase. So the uncertainty can't be managed easily. Best to use a suite of filters or your mixing scheme or maybe use a different type of detector.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf