VSWR measurement of PCB antenna.

[mrpackethead]

I'm contemplating designing some 915Mhz Low Power radio widgets, and i want to create the antenna on teh pcb.. Its compact, low cost and i can live with the compromises of not having a higher gain antenna.

However its goign to be critical that the anntenna is well matched, and i want to be able to measure/calculate VSWR with a Spectrum Analyser.  The problem is that i dont' have a way to 'break' into the antenna particaully well.   

Because the housing, the pcb etc all end up impacting on how the antenna behaves, its important that it gets measured 'in situ'.

I was thinking perhaps i could put a switched connector on teh board as way of getting onto it?   Somethign like this.

https://www.murata.com/~/media/webrenewal/support/library/catalog/products/switch/switch/o33e.ashx?la=en-gb

That would let me attach to the antenna, ( in its case, ) and attach it to the SA via the directional coupler..

Thoughts?

[hendorog]

When you say critical that it is well matched, I don't think you should obsess over that as it won't necessarily guarantee you a good result:

* The matching probably doesn't really need to be _that_ good before you hit a point where you don't get much benefit for a lot of effort.
* Having good VSWR doesn't mean that you will have good range (assuming range is what you want). A 50 ohm resistor has great VSWR but terrible range. The housing and other crap close to the antenna that will affect the VSWR will also mess with the radiation pattern of the antenna.
* Your ability to measure the return loss of the antenna with an SA will be limited by the directivity of the directional coupler, and everything between that and the antenna including the Murata connector which is spec'd at 1.3 VSWR max.
* The transmitter might not be 50 ohms exactly which means making the antenna perfectly 50 ohms is a bit pointless as it won't match the transmitter impedance anyway.

You might be better served just testing the range of the unit with and without the housing, using an antenna attached to your SA in an open area/field somewhere. What would be good would be some sort of reference antenna to compare it with so you know your design is better or worse than the reference.

Edit: I wrote the above thinking you were going to use a chip antenna, but realised you are going to design it into the PCB as a trace. In that case you could make the antenna as a seperate PCB initially and test it on it's own - and compare it to an off the shelf rubber ducky perhaps?

[T3sl4co1l]

How about: put in a trio of 0 ohm jumpers, and a u.FL connector.  That way you can connect:
Transmitter to coax: external antenna (flexibility option)
Coax to antenna: testing, use as external antenna for something else (?!)
Transmitter to antenna: normal

Tim

[TMM]

Agreed with hendorog regarding VSWR vs. radiation pattern concerns. That said, given that it sounds like you only want to make the measurement during development and not measure every single assembled widget, the way I would attack the problem is to make a dummy PCB which just has the antenna going to a coaxial connector located where the output amplifier would normally be. Trying to integrate a directional coupler or jumpers into your production PCB design is just going to add unnecessary losses and degrade performance, although would be educational.

[mrpackethead]

When you say critical that it is well matched, I don't think you should obsess over that as it won't necessarily guarantee you a good result:

Absolutely agree. A well matched antenna wont' guarrentee a good result, but a badly matched one will guarantee a bad result.  In essence just one of several things that need to be done well.

* Having good VSWR doesn't mean that you will have good range (assuming range is what you want). A 50 ohm resistor has great VSWR but terrible range. The housing and other crap close to the antenna that will affect the VSWR will also mess with the radiation pattern of the antenna.

Yes, and it is the 'other crap' that is what makes measuing it all important.    Theres only so much you can do 'on screen' with models etc,  before you actually need to measure it.       The radiation pattern will be impacted of course.    And that is a set of measurements that i'll need to do as well.    But one thing at a time.

* Your ability to measure the return loss of the antenna with an SA will be limited by the directivity of the directional coupler, and everything between that and the antenna including the Murata connector which is spec'd at 1.3 VSWR max.

Well worth considering.. and your correct about the  measurements, all the bits in it impact on the final calculations,  there will be some 'error' or 'shift'.  however its a lot better than just hoping its good is'nt it.

* The transmitter might not be 50 ohms exactly which means making the antenna perfectly 50 ohms is a bit pointless as it won't match the transmitter impedance anyway.

Again, your correct.  and as part of other measurements, I'll need to be looking at that and confirming it.    There is matching to be done so that the LNA and PA ports of the radio are done.   In that regard I have some very good reference designs to start with, but i will be testing that.

Edit: I wrote the above thinking you were going to use a chip antenna, but realised you are going to design it into the PCB as a trace. In that case you could make the antenna as a seperate PCB initially and test it on it's own - and compare it to an off the shelf rubber ducky perhaps?

You could,  ( and in fact i've already done that ) but the impact of all the 'other crap' around it, changes how it behaves, so  this is why I want to do the VSWR measurements and see if there is any scope to improve things.

[mrpackethead]

How about: put in a trio of 0 ohm jumpers, and a u.FL connector.  That way you can connect:
Transmitter to coax: external antenna (flexibility option)
Coax to antenna: testing, use as external antenna for something else (?!)
Transmitter to antenna: normal
Tim

This is'tn a bad idea, and one i'll think about, i could find a place for the connector.

[mrpackethead]

Agreed with hendorog regarding VSWR vs. radiation pattern concerns. That said, given that it sounds like you only want to make the measurement during development and not measure every single assembled widget, the way I would attack the problem is to make a dummy PCB which just has the antenna going to a coaxial connector located where the output amplifier would normally be. Trying to integrate a directional coupler or jumpers into your production PCB design is just going to add unnecessary losses and degrade performance, although would be educational.

Yes, just during development.   Doing it for every board would be a big time waste!     However the units will get tested obviously, and one of the tests will be radio performance to a known base station.   If the device falls outside of an acceptable 'window' of operation we'll know its not good, and we and go looking for a reason.   If its just one out of 1000, then the reality is that it will probalby just get scrapped. Its not worth the time to look for a fault, if you have 999 good ones.

[mrpackethead]

This is the first draft of what i'm looking at.   it all fits onto a 2 layer PCB.  The underside forms the gnd plane,  theres one trace on teh bottom that is travelling NS, which is cutting the  gnd plane up, badly, which i need to look, but i think i've seen a way to improve that.

[cdev]

is your antenna a monopole? If so, you likely could model it in free software and avoid complicated procedures...

[Someone]

Just remove the first series element from the matching network (from the chip towards the antenna) and use the pad as your access point for attaching the VNA, easy enough to do an open-short cal against it and then return the components toward the antenna.

[Koen]

Add an u.FL footprint before the matching network. Take three empty boards, solder an u.FL connector to them along with nothing, a short to ground and 50ohm to ground. They'll be used to calibrate the VNA. Make a normal board, add the u.FL, measure, tweak, check out Atyune, remove the u.FL, move on. Using u.FL gives you less uncertainties about cable length, soldering, measurement repeatability.

[IconicPCB]

Minicircuit  have a reasonable selection of SMD directional couplers, Include one in the feed line and Bob's your uncle.

[nctnico]

How about taking a piece of coax connected to the directional coupler at one end and create an attenuator (which is connected to the antenna feed point) at the other end. That way the coax sees 50 Ohm (ish) and you can still measure the effect of the antenna.

[mrpackethead]

Add an u.FL footprint before the matching network. Take three empty boards, solder an u.FL connector to them along with nothing, a short to ground and 50ohm to ground. They'll be used to calibrate the VNA. Make a normal board, add the u.FL, measure, tweak, check out Atyune, remove the u.FL, move on. Using u.FL gives you less uncertainties about cable length, soldering, measurement repeatability.

I was intending on using a SA to do the VSWR rather than a VNA.  Which ( and why ) is better..  or should i be doing both.

[nctnico]

A VNA is much better because it will tell what you need to change (whether you antenna is too inductive or too capacitive). A VNA will generally also allow to calibrate a piece of coax out of the equation by doing an open-shorted-terminated calibration cycle.

[hendorog]

Add an u.FL footprint before the matching network. Take three empty boards, solder an u.FL connector to them along with nothing, a short to ground and 50ohm to ground. They'll be used to calibrate the VNA. Make a normal board, add the u.FL, measure, tweak, check out Atyune, remove the u.FL, move on. Using u.FL gives you less uncertainties about cable length, soldering, measurement repeatability.

I was intending on using a SA to do the VSWR rather than a VNA.  Which ( and why ) is better..  or should i be doing both.

Further to what ntnico said, it also allows you to remove the effect of just about much of your measurement path - including the directivity of the directional coupler.

What I was alluding to earlier about the effect of this stuff on your measurement. The accuracy of your measurement can be degraded to the point that it may just confuse the issue. A quick google found this document with some graphs in it: http://www.markimicrowave.com/assets/appnotes/directivity_and_vswr_measurements.pdf

You can see the measurement errors can be massive when the directivity approaches the return loss of the DUT.

Now my understanding (I could well be wrong!) is that the return loss of any subsequent adapters - like the Murata connection - and any cables you use after the directional coupler will also also degrade the directivity, and hence the accuracy of a scalar measurement. With a VSWR of 1.3 of the connector, you are at -17dB return loss and so can measure a return loss of -12dB +/- 5dB. So your true return loss could actually be anywhere from -7dB to -17dB! This is assuming you have a decent coupler that won't degrade this much more.

So (if I am correct) your measurement might not be worth doing at all. Which is why I think the measurements in a field are more important, as they are easy to do, are reasonably reliable and should reflect the real world performance of your device.

With a VNA, you can calibrate out the effect of directivity, and so make a more precise measurement.  All of these measurements are a heck of a lot easier if the antenna has a plug on it.