EEVblog Electronics Community Forum

Electronics => RF, Microwave, Ham Radio => Topic started by: alexxs88 on June 15, 2021, 06:47:52 am

Title: Antenna matching circuit - weird behaviour
Post by: alexxs88 on June 15, 2021, 06:47:52 am
Hi,

I have a 2.4GHz device with room for a Pi matching circuit. I cut J5 (to disconnect the MCU - brown line in layout picture), soldered the ground of the coaxial cable in the white region and the feedline from the coax going to the VNA where it's marked in light green.

[attach=1]

[attach=2]

Calibrated the VNA using C24 to install a short and the 50R as load calibration. With that done, I soldered 100pF caps as shorts in L1 and C23 and started measuring to get a baseline. With the help of a series capacitor (0.5pF) and a shunt inductor (6.8nH) I went from the green trace in the Smith chart to the red one and from a VSWR of of ~7 to 1.17. In theory the gains should've been massive. However, and this is what I'm puzzled about, doing back to back tests (transmission) with a device that only has the 100pF caps in series, the wireless performance is nearly identical.

[attach=3]

I racked my brain over this and can't find any decent explanation. The only theory (and it's a pretty weak one) is that angling the feedline of the coax back towards the MCU somehow influenced my readings. I've used the same VNA and tools to match another device a few weeks back and it all worked perfectly, so I'm quite puzzled.

Thanks all!

Title: Re: Antenna matching circuit - weird behaviour
Post by: evb149 on June 15, 2021, 08:01:25 am
Yeah it could be the influence of your measurement apparatus / fixture / cabling.

It could be other things.

Were you testing "in the far field" i.e. at a fairly long distance e.g. 10 wavelengths distance from the antenna?

Was your test area relatively free of things that could detune / reflect the signal?

Were you measuring the RSSI strength from a paired IC receiver, or were you using a spectrum analyzer or what to measure the received signal strength?  Or were you doing a receive test from an unmodified distant transmitter and looking at the incoming RSSI?

Obviously small antennas particularly in apparatus which affects the pattern will have very irregular spatial gain patterns.
Are you sure your physical angular orientation of the transmitter and receiver were consistent between the red / green trace tests?

Did the spatial angular pattern shift appreciably after the retuning of the antenna?

Edit: nevremind, I see the chip antenna model in the schematic.

What about those VNA sweeps, is the frequency you tested the link loss at near the frequency where the feed was closely matched to 50 ohms?  I see the red trace that goes in sweeping all the way from a match not much better than the green trace to a pretty good match near the end of the frequency sweep but the test frequency / frequency range is not indicated with markers or comment.

Did you calibrate the VNA at the end of the coax cable over the sweep range before connecting the coax to the PCB?




Title: Re: Antenna matching circuit - weird behaviour
Post by: alexxs88 on June 15, 2021, 10:08:58 am
Quote
Were you testing "in the far field" i.e. at a fairly long distance e.g. 10 wavelengths distance from the antenna?

Yes.

Quote
Were you measuring the RSSI strength from a paired IC receiver, or were you using a spectrum analyzer or what to measure the received signal strength?  Or were you doing a receive test from an unmodified distant transmitter and looking at the incoming RSSI?

This is how I tested. Had two identical devices acting as the transmitter, one had shorts in place of the Pi network and the other one was tuned. They were both ~2m away from the receiver, in exactly the same orientation. The signal strength is called "link quality" and derived from RSSI on the receiver. I also did various other tests, with one of the transmitters disabled,  swapped positions, just to make sure it wasn't some freak occurrence.

Quote
What about those VNA sweeps, is the frequency you tested the link loss at near the frequency where the feed was closely matched to 50 ohms?  I see the red trace that goes in sweeping all the way from a match not much better than the green trace to a pretty good match near the end of the frequency sweep but the test frequency / frequency range is not indicated with markers or comment.

The receiver is the one who controls the frequency, the network was formed on channel 11 on 802.15.4 . Below is the VNA sweep with the specific frequency markers before/after the matching circuit was implemented.

Quote
Did you calibrate the VNA at the end of the coax cable over the sweep range before connecting the coax to the PCB?

Yes, I have. Soldered the coax cable to the PCB at the points marked on the layout, then used a short and a 50ohm resistor to calibrate it.
Title: Re: Antenna matching circuit - weird behaviour
Post by: evb149 on June 15, 2021, 12:00:45 pm
You might want to check if you're using reasonably high Q passives in the circuit.
It could be fairly lossy / low Q due to the passives (e.g. 6.8nH inductor vs. a good high Q high SRF maybe wire-wound one) which would be especially relevant for the electrically small chip antenna.


Wait.. you said you're using a series 0.5pF coupling cap.  That's a bit small of a value compared to, say, the layout which doesn't look to be that "tight" relative to capacitive parasitics.

Maybe revisit your smith chart matching network solution and see if there's another network which would match it that would not use such a small capacitor value.

Title: Re: Antenna matching circuit - weird behaviour
Post by: alexxs88 on June 15, 2021, 12:34:44 pm
Thanks for all the suggestions! I'll try a different "path" on the Smith chart. Just so I understand things a bit better:
1. When you say "not tight relative to capacitive parasitics", do you mean the unplanted stubs in the Pi circuit or something else?
2. If the problem were the cap, wouldn't it show on the VNA sweep as well?
Title: Re: Antenna matching circuit - weird behaviour
Post by: David Hess on June 15, 2021, 05:28:04 pm
Is the transmitter output actually 50 ohms?  Usually it is intended to match to 50 ohms but is itself much lower than 50 ohms so a VNA measurement in place of the transmitter would produce the wrong result.
Title: Re: Antenna matching circuit - weird behaviour
Post by: alexxs88 on June 16, 2021, 11:18:18 am
From everything I could find online and in their documentation it seems to be 50 Ohms, the balun is on-chip
Title: Re: Antenna matching circuit - weird behaviour
Post by: MartinL on June 16, 2021, 10:55:58 pm
As evb149 pointed out, 0.5pF is a really tiny capacitance.

When I was first learning to play with 0402 SMD stuff some 15 years ago, I was told that if you wanted to add 0.5pF to a capacitor, just put a bigger than normal blob of solder on one end of it.

I've never actually verified that figure, but my experience suggests it's plausible - and that puts in perspective just how vulnerable your match may be. Just the act of removing the solder holding the end of the coax might change the parasitics by more than 0.5pF. And that's even before thinking about what effect the additional counterpoise provided by the coax will have on a tiny chip antenna.

Unfortunately, with things this small it's sometimes just not practical to insert a VNA without disturbing the setup enough that the measurements don't reflect normal behaviour.

At lower frequencies, my advice would be to find a path round the Smith chart that lets you use values of at least several pF or nH, where the parasitics will be smaller by comparison. But at 2.4GHz you're pretty much forced to small values.

In practice your best bet may be to tune by trial and error: changing components and then measuring far field performance. But you need to take care that your setup is consistent; with these little chip antennas, small differences in orientation or surroundings can have large effects. Ideally you would be changing parts and then slotting the device back into a precision fixture in an anechoic chamber with a calibrated receive setup at the other end of the room. Obviously that's probably not practical for you, but you get the point: you need to try to be as consistent as you can to get meaningful results.

All this is why it's usually very strongly advised to use a module or a reference layout where the work has been done for you. I never liked that answer either, but I've come to understand why it's so often given.
Title: Re: Antenna matching circuit - weird behaviour
Post by: zach05 on August 16, 2021, 12:11:22 pm
the testing cable much likely changed your RF ground configuration.

your antenna may be well tuned with the testing cable soldered on PCB, but completed de-tuned when the cable is removed.
Title: Re: Antenna matching circuit - weird behaviour
Post by: E Kafeman on August 20, 2021, 08:07:34 am
Yes it usually becomes a bit trial and error to find best matching option but for serial capacitors with small values can it be impossible to get a good match due to stray-capacitances not only between SMD pads but also from pads and component body to surrounding ground.
See below picture. It is often several matching network arrangements that can be used to get similar results.
With small capacitor value connected to ground will stray capacitance not be as harmful and finding correct matching value do seldom need more then one try.
(https://www.eevblog.com/forum/rf-microwave/antenna-matching-circuit-weird-behaviour/?action=dlattach;attach=1248490;image)
Unmatched curve is just made up but at 2440MHz does it have same impedance as unmatched value given by alexxs88.At 2.4GHz must selected type of inductor be taken in account. They are not ideal components and its losses must be included in calculated matching network.

Agree with previous comments that measurement cable easily becomes a extension of local ground plane and can screw-up  tuning result when tuning cable not any longer is connected. Measurement cable should leave PCB ground at a non hot location and using ferrit tubes to increase braid impedance are standard methods to avoid cable influence at tuning result.
I am showing both how to avoid that cable leaves PCB near antenna or at a hot area, mainly PCB corners, as well as how ferrit tubes are used to reduce effect of measurement cable here at 3:22 : https://youtu.be/RyMFun_KhAc (https://youtu.be/RyMFun_KhAc)
In video shown PCB is rather small, a hearing aid, and the smaller PCB ground size, the more will measurement cable affect measured antenna impedance if antenna is of type depending on ground for its resulting radiation.
Title: Re: Antenna matching circuit - weird behaviour
Post by: Joel_Dunsmore on August 20, 2021, 04:30:30 pm
Don't forget any real parts will have parasitic capacitance (for inductors) or inductance (for capacitors) that tend to make their values appear much larger than listed due to approaching the self-resonant freq.  This also includes foot-print effects.  0.3 pF to ground is not much more than the pad footprint to ground, and 8.3 nH is about what you would get with 1/3 of inch of narrow transmission line.  In the Berkeley EE 142 there's experiments where students design a BPF using SMT parts in ADS, and find they are all centered at about 2/3 to 1/2 the desired frequency. Then they measure the components, reveal their parasitic models, and redesign taking these into account.  Here's a link to the lab pages (I think they are open to all): http://rfic.eecs.berkeley.edu/142/labs.html (http://rfic.eecs.berkeley.edu/142/labs.html)
 
Title: Re: Antenna matching circuit - weird behaviour
Post by: E Kafeman on August 20, 2021, 06:51:03 pm
Yes both inductors and capacitors have non ideal and frequency depending impedance that must be handled, at least for GHz-frequencies. That is why I included brand and type of inductor for above matching network.S-parameters for components should always be used when calculating discrete matching networks for best result, both as virtual and real network, as can bee seen in linked video above. Using real impedance for discreet components is an integrated part of my VNA software, AnTune, which automatically finds optimized component values for any wideband antenna tuning, using real component impedance.Using real impedances are faster an simpler way get expected result at first try.
Remaining problems are mainly losses related to non ideal pads and lossy PCB ground impedance which may need one or two trial and error to find optimal network values, good enough for most antenna tuning jobs.
However, these PCB  shown at linked Berkeley page, I am not surprised if RF performance not is at top. PCB looks single side as well. The good thing is that it is from the problems knowledge is added. Also I have done similar PCB with trace routing done in MSPaint.

Certain tuning typologies are more tricky to tune for expected result, component value tolerances are more critical and can cost much higher resistive losses compared to alternative topology and component type, even if both alternatives results in a similar tuning result.
Above example shows two alternative solutions to get almost same final impedance.
What topology to prefer, what types of components to avoid have no absolute rules. It is something which is best learnt by experience but assume it also can be easy evaluated in ADS if using S-parameters in similar way that AnTune can present result based on alternative components types, such as say 0402 film relative wire inductor for a certain brand and value.

Using AnTune gives the advantage that you get these results live presented based on VNA measurements for both for real and virtual networks and a even as a mix of both, mixing virtual and real components for same tuning circuit.ps. HP and Agilent is supported by AnTune out of box, no configuration needed  ;) .