Author Topic: VNA calibration for pi network match  (Read 5848 times)

0 Members and 1 Guest are viewing this topic.

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
VNA calibration for pi network match
« on: March 11, 2019, 11:13:20 pm »
Hello,

wondering if there are opinions about the following two methods for best possible match for a PCB antenna. Referring to the attached pi network schematic, I am contemplating doing the following:

Method A:
1. Calibrate the VNA with S/O/L at the VNA port
2. Attach semi-rigid SMA-connector cable (cut off at one end)
3. Set port extension (verifying both closed and open, finding the best compromise)
4. Measure antenna from cut off end of cable soldered at C23, and shielding to ground
5. Proceed with match

Method B:
1. Solder cable at L32 + ground
2. Measure S/O/L using 0402 0 ohm/open/50 ohm
3. Place 0 ohm at C71
4. Measure antenna
5. Proceed with match

TIA!

 

Offline radiolistener

  • Super Contributor
  • ***
  • Posts: 3282
  • Country: ua
Re: VNA calibration for pi network match
« Reply #1 on: March 11, 2019, 11:34:13 pm »
Does your antenna is intended for TX and RX? If so, then output impedance should be the same for RX and TX mode.
It means that you can connect antenna directly to the VNA and measure antenna impedance at target frequency.
Then you can connect VNA into antenna connector of your device and measure it's input impedance, it should match antenna impedance. If it is not, then tune your pi network. That's it.

If your antenna connected to the pi network without connector, I'm afraid there is no way to measure your antenna.
But you can make cut-off between antenna and pi network, tune it and then solder jumper to restore cut-off.

In order to approximate pi network components without cut-off, you can desolder C71 and C23.
Then calibrate your VNA with a coax cable, solder one side of calibrated coax cable on C23 pads, measure antenna impedance.
Then calculate pi network and solder components. But in such way you cannot check the real impedance match.
« Last Edit: March 11, 2019, 11:50:58 pm by radiolistener »
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #2 on: March 11, 2019, 11:46:11 pm »
Hi, sorry- should have mentioned that the antenna is both RX and TX, AND it is a PCB trace antenna, so the measurement can't be made by just connecting it to the VNA.
 

Offline radiolistener

  • Super Contributor
  • ***
  • Posts: 3282
  • Country: ua
Re: VNA calibration for pi network match
« Reply #3 on: March 11, 2019, 11:53:23 pm »
I think, then there is no way to measure antenna. But you can cut-off wire between pi network and antenna and then use VNA with a piece of coax cable. VNA should be calibrated with this piece of coax cable. After tune, just solder jumper and that's it.

Also, you can desolder C71 and C23 and solder VNA coax cable to C23 pads to measure antenna impedance.
Then calculate pi network to match antenna and solder calculated values. But in such way you cannot tune pi network for the best match.

In both case you should calibrate your VNA through piece of coax cable used to connect VNA with your device.
This is needed to remove coax cable influence, so your VNA will measure value on the end of coax cable.
« Last Edit: March 12, 2019, 12:06:22 am by radiolistener »
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #4 on: March 11, 2019, 11:59:32 pm »
Of course you can measure and match a PCB antenna! Read the steps I outline-- I am talking about using a coax cable. There is a lot of literature, but some point to calibration at the plane, others at the VNA with port extension, some measure before the PI network, others after, etc.

Looking for opinions from people who've actually had experience doing it.
 

Offline radiolistener

  • Super Contributor
  • ***
  • Posts: 3282
  • Country: ua
Re: VNA calibration for pi network match
« Reply #5 on: March 12, 2019, 12:17:32 am »
Of course you can measure and match a PCB antenna!

could you please explain theory? With no need to cut-off, you can measure just antenna.
It can be done with desolder C71 and C23 and connect VNA coax into C23.
VNA should be calibrated through coax cable, to remove it's influence.

But how you're planning to tune your pi network with no need to cut-off wires?

There is a lot of literature, but some point to calibration at the plane

Did you mean literature about tune in circuit pi network or just VNA calibration through cable?
Could you please share some example of literature about tuning in circuit pi network?

I had similar experience, but I was needed to tune 7-th Chebyshev LPF and it was not so easy task.
I'm very interesting if there are methods for in circuit tune of two port network with no need to cut-off wires.

Update: have some thoughts about your case, since there is no transmission line between pi network and antenna, there is no loss. So you can tune your pi network together with antenna as a whole antenna system.
But it still needs to cut-off wire before L32 in order to connect VNA instead of in circuit source.
So, I'm wonder if there is a way to tune it with no need to cut-off wire.
« Last Edit: March 12, 2019, 01:46:28 am by radiolistener »
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #6 on: March 12, 2019, 09:47:02 am »
Hi, if you read my question, you will find point #2:

Quote
2. Attach semi-rigid SMA-connector cable (cut off at one end)

That is, in fact, what some references suggest (others add a U.FL or similar connector). Here are two examples:

http://www.ti.com/lit/an/swra161b/swra161b.pdf (section 6.2.1 - connecting at the feedpoint)
https://www.cypress.com/file/136236/download (section 15.1.1 - connecting before the pi network)

But reading your reply, I think you may be talking about the radio side of the network that needs cutting off? If so, that side is not populated yet, and I would in fact solder the coax cable where an ipc balun will be in the final product. So, effectively, the network is cut off from the rest of the board for these measurements.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: VNA calibration for pi network match
« Reply #7 on: March 12, 2019, 10:20:23 am »
In cypress document 2.4GHz IC (it's RF pin) is 50-ohm matched - that's why you can connect 50Ohm VNA in place of RF chip. This is not always the case, especially with sub-1GHz chips. In such case (when chip is not 50ohm) only place you can connect 50ohm VNA - point where you expect 50ohm impedance which is antenna "port".

[edit]  When IC is 50ohms you just tune matching network to be 50ohms on RF IC side. When you do not have luxury of 50ohms RF chip, first you tune matching network for 50ohms, then tune antenna for 50ohms.
« Last Edit: March 12, 2019, 10:31:57 am by ogden »
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #8 on: March 12, 2019, 10:40:48 am »
@ogden -- aah, perfect! In my case, the chip is not 50-ohm, but a balun immediately follows it. I guess this gives me the luxury of taking measurements at the balun's output, rather than directly at the feedline of the antenna (I might do both to compare anyway).

Now back to the question: would you calibrate the VNA at the reference plane, or use port extensions?
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: VNA calibration for pi network match
« Reply #9 on: March 12, 2019, 10:46:13 am »
@ogden -- aah, perfect! In my case, the chip is not 50-ohm, but a balun immediately follows it. I guess this gives me the luxury of taking measurements at the balun's output, rather than directly at the feedline of the antenna (I might do both to compare anyway).

Now back to the question: would you calibrate the VNA at the reference plane, or use port extensions?

TI document you mentioned, 6.2.2 Calibration  ;)

Quote
To measure the impedance of the antenna it is necessary to move the reference plane from the SMA connector to the feed point of the antenna. This must be done to adjust for the phase change caused by the semi rigid coax cable.
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #10 on: March 12, 2019, 04:23:31 pm »
Oh, I know what TI says in that doc :-)

But there are other sources that suggest that same approach OR calibrating at the reference plane (don't have a link handy now, but it was from a manufacturer), with no clear explanation as to which might be better under what circumstances.
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: VNA calibration for pi network match
« Reply #11 on: March 12, 2019, 05:09:13 pm »
i've done it both ways. I'd look what is more practical in a given case. Port extensions assume zero loss in the extension so you keep the extension line as short as possible, that may require doing calibration not at the vna terminal but closer to where you intend to measure. Also solder the extension to the board ground plane connection point before dialing the extension in the  vna.

Using a in-place OSL cal is more ok for sub-1GHz  frequencies, as you unlikely to know the makeshift 'cal kit' coefficients. At low enough frequencies they won't matter much so you can get away with just a 0603 50 Ohm resistor for Load and the center wire shorted to the shield for the Short.
Facebook-free life and Rigol-free shack.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: VNA calibration for pi network match
« Reply #12 on: March 12, 2019, 05:43:36 pm »
Oh, I know what TI says in that doc :-)

But there are other sources that suggest that same approach OR calibrating at the reference plane (don't have a link handy now, but it was from a manufacturer), with no clear explanation as to which might be better under what circumstances.

Other sources also suggest to calibrate reference plane which is *in* the actual circuit, not at the end of the cable which definitely is not ideal (nothing is). On the other hand few centimeters of good semirigid and quality connector, especially at < 1GHz will not do any harm for S11 return loss measurements. I would definitely try and compare difference myself, especially after consulting some random guy in public forum ;)
« Last Edit: March 12, 2019, 06:43:35 pm by ogden »
 

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #13 on: March 12, 2019, 07:53:25 pm »
Quote
Using a in-place OSL cal is more ok for sub-1GHz  frequencies, as you unlikely to know the makeshift 'cal kit' coefficients. At low enough frequencies they won't matter much so you can get away with just a 0603 50 Ohm resistor for Load and the center wire shorted to the shield for the Short.

Yeah, I have to do 915 MHz and 2.4 GHz. I'll have to give an 0402 resistor a try, since that's the footprints I have. And as for coefficients for the cal kit, well, even the kit I have is rudimentary and uncharacterized -- probably making this whole thread moot!

Quote
I would definitely try and compare difference myself, especially after consulting some random guy in public forum ;)

So when you calibrated straight from the board, I assume you looked for one of your better 50 ohm resistors?

I'll give both approaches a try and match two boards according to the two sets of results, then measure RSSI with a known good receiver for both.

Indeed!
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: VNA calibration for pi network match
« Reply #14 on: March 12, 2019, 09:50:40 pm »
So when you calibrated straight from the board, I assume you looked for one of your better 50 ohm resistors?

Use resistors you would use in RF circuits. Cable and resistor shall be soldered on to solid "ground plane", you can't just stick resistor to the end of the pigtail. First calibrate VNA using your cal kit, then S11-measure terminated semirigid pigtail to see how good termination is. If good - then recalibrate reference plane to be end of the pigtail. Could be so that difference is insignificant, especially knowing (in)tolerances of components used in matching network.

Quote
I'll give both approaches a try and match two boards according to the two sets of results, then measure RSSI with a known good receiver for both.

Do RSSI comparison only if you find big difference between two calibration planes. Perhaps due to spread of matching network characteristics, VNA calibration specifics we are talking about falls into "do not bother" category :)
 

Offline LukeW

  • Frequent Contributor
  • **
  • Posts: 686
Re: VNA calibration for pi network match
« Reply #15 on: March 12, 2019, 11:04:06 pm »
There's also an Atmel application note here that covers this kind of antenna matching process.

http://ww1.microchip.com/downloads/en/AppNotes/Atmel-42732-Antenna-Tune-Up_ApplicationNote_AT16802.pdf
 
The following users thanked this post: ogden

Offline mayorTopic starter

  • Regular Contributor
  • *
  • Posts: 234
  • Country: ca
Re: VNA calibration for pi network match
« Reply #16 on: March 13, 2019, 07:48:07 pm »
Do RSSI comparison only if you find big difference between two calibration planes. Perhaps due to spread of matching network characteristics, VNA calibration specifics we are talking about falls into "do not bother" category :)

Hah! That's the easy part. I'll do it as a learning experience!
 

Offline E Kafeman

  • Regular Contributor
  • *
  • Posts: 107
  • Country: se
    • AnTune VNA software
Re: VNA calibration for pi network match
« Reply #17 on: March 17, 2019, 12:48:29 pm »
Assuming RF above 500MHz and a VNA becomes a very important tool to be able to do many needed measurements in an effective and reliable way.
Then measured 50+j0 Ohm, it never happens for an embedded antenna or radio.
Both antenna and radio impedance are with freq. irregular complex impedance curves, never a straight line.
Data sheet may say radio interface is 50 or 100 Ohm. Only thing to be sure about, impedance value is different for each PCB design and is seldom close to assumed ideal value.
PCB is affecting radio impedance. Vcc decoupling, Vcc trace lengths, ground trace lengths, PCB ground impedance, length and width of all RF-traces are a few of the impedance factors.

Match both radio and antenna  to 50+j0 Ohm and then connect? Forget it. It only create losses by adding unnecessary impedance poles making a in total less good matching network..

I do design embedded antennas and impedance matching network for living. In average do I create a matching network once every day. Mostly wideband matching, GPRS 1-3 GHz, with components in size 0201-0402 and I do always also design the antenna as well.
As both input and output impedances are irregular over frequency must designed matching network also be that to create best conjugate match.

In below description how of a possible work flow when doing this kind of job is a software used which I have written. It is a specific kind of software that adds live functions to a VNA.
This kind of software, no need to use same as me! There are a number of more or less similar software such as Optenni and Betamatch but in this case is AnTune used as something must be used to describe the work flow of PC aided VNA impedance matching.
It is often a bit expensive software but pays for it self within few weeks as it at least doubles an RF engineer productivity and quality of job when doing embedded antenna and matching work. It also reduced lab time and even reduces time needed in anechoic chamber.
Few software have that short time of ROI.
This kinds of software which works as extension to VNA are simple to use because if you is familiar with how to handle a VNA is also this software familiar, no learning curve. you is productive within minutes after first seen this software..

Following steps can be done in several ways, with aid of software of above described type as a helping tool.
It is possible to use any software that can do complex math calculations but maybe not as practical.
Using ruler an a lead pencil directly at printed Smith chart is also possible. I have done it several times for a single frequency. For wideband irregular signals, is it way too time consuming.

The helper software reads data directly from VNA and present it in a PC in similar way as it is shown at VNA but with some new math functions added.
A interface between PC and VNA must exist if live data should be able to be shown at PC screen. Typical exist interfaces over GPIB/USB/LAN but some VNA interfaces are too slow to be usable or do not provide any standard protocol.

A typical impedance matching scenario is that a prototype antenna since before is designed at a PCB with aid of CU tape. This design is transferred to Gerber format and added to a radio PCB.
PCB have after few weeks now been printed and radio chip is implemented active and working but not yet matched to antenna. Space is reserved as a PI or T network.
Antenna impedance for printed PCB is maybe not exactly same as when cu-tape was used to create antenna pattern but that is seldom critical, as long as antenna impedance is reasonable in same range as the prototype antenna. Minor changes will be taken care of by matching circuit.

Typical tuning/matching job is done by me as following:
1. Solder 10 cm semi-rigid cable at DUT near where location is for an intended matching network. Other end of cable have an SMA connector. Cable braid have got ferrit tubes to avoid extending ground plane of DUT, which else can affect measured antenna impedance.
2. At VNA is a 1 meter long high quality coaxial cable connected. At end of this cable is 10 cm long cable connected of same type as was soldered at DUT, but this one have SMA connectors in both ends.
3. Whole cable length is calibrated, 1m+10 cm. Idea to include last 10 cm is to include same type of reactive losses as for soldered pigtail cable. Port extension part is not important at the moment.
4. Remove 10 cm cable extension and connect 1meter cable to coaxial cable soldered to DUT.
5. Cable at DUT should have its center wire 2-3 mm extended above a soldering pad at intended location for matching network but it is not yet soldered to anything, it open circuit.

Now startup AnTune or corresponding software.

6. in software, press button "auto port forward"  for automatic calibration of amplitude and phase for OPEN. It is automatically selected OPEN as center wire is unsoldered.
7. Usually with an exacto, shortcut center wire to braid and press same button again for automatic calibration of SHORT. Those both steps makes current setup calibrated with high precision. both in phase and amplitude.
8. Now solder center wire at matching network such that RX and TX circuit can be measured. If needed in software, can measured impedance curves be weighted for Tx and Rx, as it never is same impedance for TX and RX.
It is also possible to tell software if harmonics should be taken in account as part of filtering/matching network. Especially as in many cases do it exist several solutions that provide same matching impedance but suppress harmonics rather different.

8. Next step press button in software that saves current Tx and/or Rx impedance.
9. In software, set this saved impedance as reference impedance instead of assumed 50 Ohm. Software will automatically assume that conjugate matching of a network is next step.
10 Move coaxial center pin from direction radio to instead direction antenna, including empty places for a matching network (typical a pi or T network.)

11. Select in software actual filter topology and size/brand of tuning components.
12. Software will from now on calculate optimal network. If surrounding near antenna is changed due to hand effect or load of enclosure is matching network dynamically updated..

13a. Assume that software present a VSWR that seems to meet requirements for a specific setup. If PCB RF-ground is very low loss and distance between each component in matching network only add minor phase error can now matching network be soldered at PCB.
 In software verify that real matching network results in same VSWR as predicted by VSWR calculated using virtual components.

13b, this step is when 13a fails. 13a do mostly work well but if optimal result is important or PCB not is ideal, especially in combination with frequencies above 1-2 GHz must we modify a bit how to perform a high quality matching job.
Modification is to implement matching minor network in steps. Solder just matching component closest to antenna, and in software check that soldered value is optimal. If not optimal adjust value a step up or down, according to software directive.
When satisfied, do same procedure with next two components, one at the time.

14 Done.


A bit long description, it takes longer time to read then practically perform. Time nedded is usually 10-15 minutes to implement a matching network of PI or T type and when job is is finished are all curves saved for documentation and matching network will be saved as a BOM-list.
Report, graphical curves and BOM is saved as a html file and curves can be saved as touchstone files for further calculations and reference if something changes in production of further batches of PCB.

This kind of software main advantage is else when designing prototype antenna in combination with matching network. For every minor adjustment of antenna shape will total resulting VSWR including for the moment optimized network be shown. This part is also of value when tuning for widest range of handling different surroundings near antenna, held in free space, hand-held..

The software I use provide a bit unique function which is of good help when designing antennas.
A very efficient antenna is not necessarily something 50 Ohm, which anyway not is of interest if optimal matching radio circuit neither provide a 50 Ohm interface.
But how to know if a particular antenna design is a good radiator and not just more or less a resistor? Rough estimation can be done by doing practical RSSI measurements. Long experience what usually work is also a good help.
For reliable results with somewhat precision is an anechoic chamber needed to measure antenna efficiency. It is not for every RF-lab, a such chamber is expensive, big and it takes time to do measurements.
What makes this software a bit unique is  that it can measure antenna efficiency directly at lab bench, dynamically and without delays with precision similar to an anechoic chamber, a few % +/- but now 10 times/second or depending on sweep speed of the VNA.
One sweep in anechoic chamber needs you to stop development process at lab bench and waiting for 30 minutes for this single sweep and then process data to find antenna efficiency.

All software functions together makes it a very complete tool if working with embedded antennas or RF filter matching in general.
However most functions are rather useless if you not have a VNA with GPIB compatible interface (USB/LAN/GPIB) or an VNA such as from Copper Mountain Technologies which provide a VISA compatible interface directly in PC or over LAN.

All above described functions can be done in Excel or many RF simulator tools as well, but not as live data at PC screen.
Non live data, manually handled, makes job process much slower. Instead of 10-15 minutes to create same level of optimizing is a more manual way maybe jobs for weeks if same optimized goal is needed.

Some measurement curves from my latest impedance matching for a BT 2.4 GHz unit in order how I did the job:

Measured RX impedance is not exactly same  as TX impedance but both curves are unusual similar. From these two curves will I create a third curve which is average of these both. Function to do this exist in software.


Average radio impedance and now is also antenna impedance measured. Antenna is measured with same coaxial cable as it is only center wire that is moved between two pads in reserved space for matching network.
Table refers to yellow curve.

Matching component family is selected in AnTune as S-parameters differs between brand and  physical size of component for same value. In this case did I select Murata 0402 with wire-based inductors (type LQW15).

Conjugate of average impedance is now set as reference impedance instead of 50+j0 Ohm (the cross in center of Smithchart becomes a curve).
Software create automatically an optimized matching network and resulting impedance (antenna+network) is shown as yellow curve.

Within actual frequency band is yellow curve now very close to impedance goal represented by grey curve.
Table refers to yellow curve.

So far, these curves are measured at same location , It is only center wire that is moved between two 0402-pads

Next step is to verify by software proposed network. Network is presented both as component values and as a simple schema.
If there are no problem do it remains to measure resulting VSWR and save all documentation.


Measured original and final matched result. Notify table values such as resistance and reactance for low VSWR is far away from 50 Ohm.
Antenna efficiency is not shown, It was measured as a part of antenna shape design, previous done.

If you not have a VNA that is GPIB compatible will not this kind of software working as VNA extension be of much help.
VNA extending software, adds functionality which VNA not had before. Even old HP VNA will get auto port forwarding, FFT, a lot of memory and clear big color screen just to mention a few things.
This kind of software is useless as simulation tool. There exist a lot of better alternatives for doing RF simulations.

I mention this several times as it is a common mistake to think about this kind of software as a simulation tool.
It is a measurement tool, using real data delivered from an VNA in real time. For virtual network is real measured S-parameters used for each L and C component. Each virtual component is interchangeable with real components. Measured result or virtual result are interchangeable as non ideal circumstances are taken in account in every step.
Both radio and antenna impedance are measured values, anything else is not possible when doing serious matching job.
« Last Edit: March 17, 2019, 07:13:48 pm by E Kafeman »
No signature
 
The following users thanked this post: 2N3055


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf