6

I'm trying to design the matching network between an antenna and its transmitter, running at 868MHz, as shown below. There's probably about 20mm between the transmitter output and the Pi network.

PCB layout

When connecting the VNA to the PCB in order to measure the impedance, where should the VNA be connected? At the transmitter end, or at the start of the Pi network?

My understanding was that in order for the VNA to "see" the load how the transmitter sees it, the VNA should ideally be connected close to the output of the transmitter (in my case, I have a blank PCB, so I've soldered a UFL connector directly on the transmitter output pad). However, after reading this article by Tektronix, it states the following:

it’s important to keep in mind that the impedance to be matched must be measured at the point where the matching network will be placed.

I understood this to mean that I should solder the UFL connector close to C13 in the image above (and possible cut the trace between the transmitter and the UFL connector).

Amr Bekhit
  • 1,278
  • 10
  • 24
  • how do you propose connecting at start of PI network? If your using sufficient bandwidth, you can connect at the pcb end and time-gate out the board launch as well as the device on the other end – johnnymopo Mar 26 '19 at 01:49
  • @johnnymopo - I've edited my question to add further clarification. You mentioned the following: `you can connect at the pcb end and time-gate out the board launch as well as the device on the other end`. Can you explain what you mean by this? – Amr Bekhit Mar 26 '19 at 05:38
  • @johnnymopo Yes, could you please explain that? :) – pfabri Nov 05 '20 at 15:12
  • look up VNA de-embedding - essentially using the time-domain transform on the VNA you can "gate" the time-response to de-embed to your DUT – johnnymopo Nov 16 '20 at 22:15
  • @AmrBekhit I've found a [great explanation of the theory behind this technique](https://www.slideshare.net/JohnMower/time-gating) and a [practical approach with python](https://scikit-rf.readthedocs.io/en/latest/examples/networktheory/Time%20Domain.html#Intro) – pfabri Nov 19 '20 at 22:33
  • 1
    @johnnymopo Wow, it's really not very helpful to just say _look up VNA de-embedding_ so smugly and tersely when you yourself have already given [an, in fact, really nice answer](https://electronics.stackexchange.com/a/198992/128715) to this elsewhere a whole year before this question came up. I'm torn: I'm both grateful for the answer you gave there and slightly mad that you didn't deign to link to it... An odd one. – pfabri Nov 19 '20 at 22:53
  • In truth, I forgot that I posted that answer and I was too brusque in the response. Thanks for linking the answer and I appreciate your opinion. – johnnymopo Nov 20 '20 at 04:07

1 Answers1

4

The VNA should be connected in one of two ways:

  1. Connect the VNA directly to the start of the matching network and cut any trace before the connection point.
  2. Connect the VNA at the output of the transmitter (like in my case), but compensate for the length of the transmission line between the connection point and the start of the matching network.

Basically, when you calibrate the VNA up to a particular calibration plane, any extra cables/transmission lines that are added after the calibration plane result in a signal delay that rotates the impedance measurement around the Smith chart, giving an incorrect reading. Ideally the VNA would be calibrated right at the measurement point, but if this is not possible, then the length of any test cables/transmission lines between the calibration plane and the measurement plane need to be entered into the VNA in order to compensate.

If designed properly, the trace between the transmitter output and the matching network should be at 50 ohms, and as such is equivalent to simply having a longer test lead between the VNA and the matching network. If the VNA is calibrated only up to the transmitter output, but the transmission line length between the transmitter output and matching network is not compensated for, the measured impedance result will be rotated and as such, any component calculations based on that will be incorrect.

Amr Bekhit
  • 1,278
  • 10
  • 24
  • Do I understand correctly that _if the VNA is calibrated only up to the transmitter output_ but the onwards-trace/transmission line from this point has a characteristic impedance of 50R, then this transmission line length does NOT require compensating for? – pfabri Nov 06 '20 at 16:40
  • 1
    @pfabri If you connect the VNA to the transmitter output and there is a 50R transmission between between the VNA connection and the matching network, you **must** input the transmission line length into the VNA, otherwise you will end up with incorrect results. – Amr Bekhit Nov 06 '20 at 17:35
  • Thank you for the speedy response! Would this also be considered _deembedding_? If so I can just go right ahead and add in the time parameter (in ps) in my VNA. Is that correct? What happens if by error the trace I have isn't 50R but say, less than that? – pfabri Nov 06 '20 at 19:33