Chip Antenna 47948--Ground plane under antenna and/or matching network?

[davethomaspilot]

[ Specified attachment is not available ]Hi Guys, long time EE, first time post.

I'm designing a pcb with a Molex 47948 chip antenna.  Here's a spec:

https://www.mouser.com/datasheet/2/276/0479480001_ANTENNAS-701294.pdf

The spec references an application note, but it mostly just details what they used for a test jig:

https://www.molex.com/pdm_docs/as/AS-47948-001-001.pdf

I can see where "same side" copper plane is spaced from the antenna and feed line, but what about backside copper?  Should there be any (they call this an "on-ground" antenna).

I'm assuming there should NOT be copper under the antenna or matching network.  Here's my first cut at component placement and bottom ground plane.  Top ground plane will be identical to bottom.  Or should it extend under everything?

[Geoff-AU]

I would vote for no backside copper, for the simple reason that Figure 6.1 of the app note shows the ground is kept back from the isolated mounting contacts on the antenna.  If it's not connected to those pads, they probably don't want it on the underside either.

[MartinL]

Usually you need to clear some space around a chip element because it's in-plane and needs some space to stick out into. But this type works by being a tall component, with a helical conductor winding upwards in the Z axis. Look at it from the side - you have it sticking up into open space. That's what's giving the necessary clearance, and that's the advantage of this style and why they call them "on-ground".

So I'd expect it's been designed with ground under the whole thing, and this promo sheet seems to confirm that - see the caption for figure 3a. It seems silly that they're not more explicit about this but I'm pretty sure that's the intent. This doc gives a dimensioned drawing for the corner layout and doesn't mention the back side at all, which it would surely have to if a cutout were expected.

Keep in mind though that when dealing with little parts like this, the component is not a complete antenna. It can't be - it's way smaller than a wavelength (13cm). It's really just an element which is driving the whole board to work as a corner-fed patch antenna. It's that big groundplane in the reference layout that's doing the work. So stay as close as you can to the reference layout, including the size and shape of the board, because otherwise you're basically redesigning that patch antenna and it won't perform the same way. And keep a solid groundplane on the bottom.

[davethomaspilot]

Yes, I made those same seemingly, contradictory observations:

1)  There is drawing that shows clearance from same side ground copper.  So, if I have to have clearance to adjacent, same side copper, doesn't that imply no back side copper?
2)  No where is there any constraint, or even mention of opposite side ground plane constraints.   You could say that means there are no constraints, or that it's just assume no back side copper
3)  It's called an "on ground" antenna.

Also, there is no mention of a minimize size ground plane, or data that shows performance degradation as a function of small ground plane size.

My pcb is no where near as big as the the test pcbs shown in the application note.  Those are huge!  If the antenna requires anywhere near that much ground plane area, it's pretty much useless (at least to me).

My pcb is for a handheld fob and less than 2 sq inches total.

The length of the helical coil in the antenna might be 1/4 wavelength--no way to know.

[Geoff-AU]

Ah I missed the "OnGround" in the name.  Yep MartinL is right.  The promo sheet says it explicitly, internal GND layers and components on reverse side are OK.

I wonder why the topside clearance to the mounting pads..  maybe detunes it too much.

As MartinL says, the ground plane is doing most of the work.  By my count there's about 3cm of helical element (3mm sizes, 9 sides traversed, plus the small vertical riser), much smaller than a wavelength.  The turn spacing is also extremely tight.  Might behave more like an inductor than a radiating element at that size.

If you want to make the PCB smaller be prepared to retune the matching or maybe find out that style of antenna isn't suitable.  A handheld fob is also going to get significant detuning effects from being held in different ways.

How much range do you need? 

[MartinL]

So, if I have to have clearance to adjacent, same side copper, doesn't that imply no back side copper?

No, I don't think it implies that at all.

Remember, the principle of all chip antennas is that you use the groundplane of your PCB as part of your antenna. Think of it like a dipole antenna where one side is the groundplane, and the other is the chip element, and RF currents are flowing back and forth between them. Normally both sides of a dipole need to be the same size. But the chip element side is made of fancy ceramics with a high dielectric constant, so it can be a lot smaller than the groundplane side.

So your antenna is designed as a combination of the chip element and the solid groundplane. There may be ground fill on the top layer as well, but the shape of that top layer ground fill is going to vary a lot with the top side layout. So to get consistent results regardless of board layout, you want to keep the RF currents flowing primarily between the chip element and the solid groundplane. That's why you want to clear the topside copper around the chip. Once it's cleared around by a half millimetre or so, there'll be practically no coupling from the top ground directly to the chip element, so things should behave consistently regardless of top layout.

Also, there is no mention of a minimize size ground plane, or data that shows performance degradation as a function of small ground plane size.

That's because the reference design is the antenna. You can try using that chip element with a different PCB size, but when you do, you're building a different antenna, and you'll get different results.

My pcb is no where near as big as the the test pcbs shown in the application note.  Those are huge!  If the antenna requires anywhere near that much ground plane area, it's pretty much useless (at least to me).

My pcb is for a handheld fob and less than 2 sq inches total.

Then don't expect to get the same results. You may still get something usable, or you may not, but once you deviate from the reference layout you're on your own. That's the reality of chip antennas. Either look for one with a reference layout closer to your board size, or be prepared to do your own design and testing.

Molex do publish an HFSS model for this part, so you could experiment with the board shape in simulation. But HFSS isn't cheap!

[davethomaspilot]

I planned to use my VNA to impedance match the  PA output to the chip antenna using the pi network.   That's why there's a spot to populate a U.fl antenna.  So, if it's just a matching issue, I'm going planning on that hassle.

But, if the intended application is for such big pcbs (doubt it), then I need to find something else.

I read that you need 1/4 wavelength "radius" on the ground plane. 

All 1/4 wave antennas work best if they are installed in the center of a metal ground plane with at least 1/4 wave length radius (1/2 wave length diameter: ~6 inches for 900Mhz and ~3 inches for 2.4Ghz); larger is better.


Given a roughly rectangular pcb, how do I get "radius" or diameter?  1/4 wavelength would be about 3 cm.  My board is longer than that, should it suffice?

Would an antenna like this require less ground plane area/"radius"?

https://www.digikey.com/en/products/detail/linx-technologies-inc/ANT-2-45-CHP-T/523151

I don't need anywhere near the range that I get with a 1/4 wavelength whip antenna.   I need about 100' and I get over 1000' feet using the same NRF24L01 + PA hardware. 

Thanks for your attention!

[davethomaspilot]

Also,  I want to orient the antenna so the maximum radiated power is optima.  The radiation diagram confuses me:



So, I'd think the color of the 3d plots represent power intensity.  But then, what are the lines that I'd think would be "iso-power".  But, clearly they are not, since following those lines goes through different colors.

I'm inclined to just use the color and assume the lines are bogus?

[MartinL]

That rule of thumb about needing at least a 1/4 wave radius of ground plane, is one that's used for a 1/4 wave monopole antenna.

At 2.4GHz what that would look like is a 6cm x 6cm groundplane, with a 3cm long wire element sticking vertically upwards from the middle of it.

If you can arrange that, great. It should radiate reasonably well upwards and outwards in all directions, with a null vertically upwards and little gain towards the backside of the board. More groundplane than that is fine  - an ideal monopole sits on an infinite groundplane, after all. You get some variation in gain and takeoff angle depending on the groundplane size. Make the groundplane too small and the antenna will no longer resonate naturally at the operating frequency, so you'll find it harder to match to and it will have poor efficiency.

But your corner-mounted chip antenna bears very little resemblance to that configuration. Yes, there's about a 1/4 wave of electrical length in that helix, but it's not going to behave in at all the same way as a straight 1/4 wave wire element, so pretending it's a monopole isn't going to bring you to meaningful conclusions. It's going to do something weird, and your options if you want to know how it behaves in a given configuration are to either simulate it, test it, or stick to the known design the manufacturer provides. You're not going to get far with throwing rules of thumb at it from a very different, and much simpler, form of antenna.

At the end of the day, most of the time you can stick any old chip element on any old board and it will radiate to some extent. If you only need 100', outdoors, and you have a PA on there then it may well be fine. But don't expect that 3D plot in the datasheet to even vaguely resemble what the radiation pattern will look like in your own configuration.

(The colours are gain, and correspond to distance from the origin. The lines are just the polygon mesh that it's using to display the 3D shape, they don't mean anything as lines)

[davethomaspilot]

Thanks!

Note that corner mounting is the recommended configuration.  Sure, it's not an ideal monopole.  I guess we are just supposed to guess on how much ground plane area is needed.

Board size is about 2.2" x 1".  Too small?

Do you have a suggestion for something that might work better? 

Unrouted pcb picture attached.  Top side has pushbuttons and battery.  NRF24L01 + PA/LNA on back side along with matching network, processor, decap, programming pads, and chip antenna.

 

[Geoff-AU]

There's no absolute, black and white, "X amount of groundplane is needed".  You can go smaller, and attempt to match the antenna to suit, but at some point your bandwidth will become unworkably narrow, the efficiency will be poor (short radiators are inefficient, and lumped tuning elements especially inductors are quite lossy).  The result is that your usable range decreases.  Hence the earlier question about "how much range do you need".

Regarding transceivers, bandwidth and sensitivity are inversely proportional (wider bandwidth = more noise).  NRF24L01 is 250kbps to 2Mbps, but your board only has 4 buttons on it.  250kbps seems like too much system bandwidth for 4 buttons, you'd get better RF performance with a narrowband transmitter.  Any reason it has to be 2.4GHz?

I've worked on keyfob transmitters in a previous life that used sub-GHz ISM (433MHz, 868MHz, 915MHz) and those could easily be >20m of range because narrowband and simple modulation trumps the lower operating frequency.  Your 2.4GHz bluetooth-esque radios seldom go that far.  LoRA is optimised for distance at the expensive of bandwith, they have a huge link budget.  Most are sub-GHz but you can get 2.4GHz too.

[MartinL]

Do you have a suggestion for something that might work better? 

Yes: make it big enough to build a planar yagi into the layout in the direction you want the signal to go and it will have fantastic directional performance.

The product requirements may say it has to be the size you want, but physics isn't required to agree.

[davethomaspilot]

Changing the frequency is not an option.  The remotes have to be compatible with existing equipment that's been in the field 10+ years. 

The rubber membranes in those FOBs are wearing out and replacements are needed.   The original ones no longer available.

Their range was adequate but marginal--about 100' but less when lots of interference was present.  Those original FOBS used an NRF24L01 without a PA/LNA and a pcb trace antenna. 

The pcb size for a new design is driven by availability of an enclosure with the correct button orientation.  There are not many choices for a handheld.

But the need for a large ground plane is inconsistent with my experience.   Over the past decade I've used hundreds of these in different applications with and without PA/LNA .  The picture shows a shielded one that is by far the best of several alternatives.

On the end of a 4" ribbon cable (no ground plane), line of sight range for the shielded module is greater than 1/4 mile.   I don't know how much more than a 1/4 mile range--that's way more than I need and  just as far as I can conveniently get line of sight in my neighborhood.

In fact, I did a replacement FOB about 8 years ago  using the same key button enclosure.  It used a different NRF24L01 + PA/LNA module  that had its own, onboard chip antenna (see picture).  It worked ok, but I had to use aluminum foil between the pcb and the RF module to prevent the pushbutton traces picking up so much RF that the digital inputs couldn't be read reliably by the processor.

I forgot to do the ground pouring on that pcb.  I have another pass in process that does have the pours AND places to populate caps on the wire traces.  That might be THE solution, but I decided to also try a design that crammed the better transceiver module into the case.

Since that design had more than adequate range (and has very little ground plane area) I'm skeptical about a large ground plane area is required.

A picture is attached that shows the that prototype replacement  design I did in 2013 beside the best NRF24L01 +PA/LNA (shielded) I've found.

I've found the shielded version to have FAR more range in several applications.  Maybe the aluminum shield acts as a "counterpoise" so additional ground plane area is not required.
 
That range is with 1/4 wave whip or homemade 1/4 wave wire antenna.  And, using them at 2.51 Ghz (out of the wifi range to avoid interference) when the OTS whips were at 2.4 Ghz.  The module with the integrated antenna has no where near that range, probably 200-300' IIRC

Is there any reason the ground plane requirements for the chip antenna would differ from a 1/4 wave wire?  Or, that the chip antenna with clearance (as used in the RF module with onboard antenna) would need less ground plane area?

[MartinL]

Nobody is telling you that you need a giant ground plane for this to work.

What we're telling you is that the only way to know what will and won't work is to test it.

Remember, you're building something that has to resonate, so it's not just a question of needing some minimum size. You're building something that will behave somewhat like a patch antenna. The ways that wavefronts propagate and bounce around it will matter. There are sizes and shapes that will work well and sizes and shapes that won't. It won't be a linear relationship between size and performance. The only way to know is to model it, or build it and test it. You can do that yourself, or you can go with a design that someone's already tested.

Nobody here can give you a shortcut. For very simple antennas like monopoles and dipoles there are rules of thumb that can be applied from experience, but you're working with far more complicated geometry, at a scale where all sorts of stuff is going to have an effect.

Yes, a chip antenna behaves differently to a 1/4 wire. Yes, an aluminium shield can act as part of the counterpoise. So will the hand of the person holding the fob, for that matter. I've designed antennas for animal trackers that only match and resonate when they're on the animal. Take the collar off and they behave completely differently. Same thing happens when you hold a fob in your hand.

95% of products, nobody cares about any of this, they stick a chip element on and it works well enough at short indoor ranges which is all anyone expects to work. You can stretch that a bit by throwing more power at it. If you need to go beyond that, you can't dodge the complexity any more and you have to actually take the physics seriously.

[davethomaspilot]

What we're telling you is that the only way to know what will and won't work is to test it

.

Well sure, that's the nature of RF design.  But first you have to design something to test! 

There's a big difference between what I need versus an optimal design.  I don't care if its optimal, it just has to be good enough.

Takeaways:

1)  Extend opposite side copper under antenna.  Same for under matching network?
2)  Bigger ground plane area would be better
3)  No comments on whether the 47948 style of antenna is better or worse compared to "flat" chip antenna like shown in the picture.

I've decided to submit gerber for three designs.  One with the 47948, and two with inverted F and "meandering line inverted F" (if I can get them to fit with required ground clearances).

Thanks for the attention and thoughtful comments.

[davethomaspilot]

Been doing a lot of surfing.  I found this app note to be excellent:

https://www.eeweb.com/wp-content/uploads/articles-app-notes-files-antenna-design-guide-1435555317.pdf

Looks like a MIFA antenna is a better choice for my application.  Not so dependent on ground plane area.

[davethomaspilot]

Just want to say thanks again for inspiring me to taking the trouble to simulate pcb antenna alternatives!

I was able to get reasonable correlation between simulation results that are published app notes versus my simulations using openEMS.  But as you warned, everything matters!

For example, I found that changing nothing but the board and ground plane size makes a significant difference,

19x67 shows s11 for a 18.67 x 61.47 board that's nothing but  the pcb, ground plane and the MIFA antenna.

25x57 shows s11 for the same antenna--everything identical except for the size of the pcb (25.4x57.15) and the ground plane.

My pcb is smaller--simulating that next.

But, if I have to tweak the matching network with a VNA on the actual pcb (in its plastic case) , does it matter? 

[Geoff-AU]

But, if I have to tweak the matching network with a VNA on the actual pcb (in its plastic case) , does it matter?

Antennas are a case of "size, bandwidth, efficiency.  Pick two"

If you reduce the size of an antenna then to make it resonate at the right frequency again you have to add loading to it.  You might be able to tweak the impedance match a little bit but it's impractical to impedance match something that's way off resonant.  Both tuning tricks will act to reduce the efficiency and reduce the bandwidth.  If it's a small tweak then who cares, if it's a big tweak then you may find the performance rapidly drops or detuning from holding it becomes very sensitive on your exact hand position (ie just like a recent iphone version, you might be "holding it wrong").

Another impact is manufacturing spread, you may build a prototype and have it work well but then you go to make 100pcs and they're all over the place.  If you're prepared to hand tune each one, fine, but if you want them to fly off the production line and be performant then you're in trouble.

But if your design choices work and get the result you need - no.. it doesn't matter.  Engineering is all about picking your compromises to make things fit for purpose.

[davethomaspilot]

you may find the performance rapidly drops or detuning from holding it

Yes, seems it is recommended that simulations be done with an enclosure model, but my gut tells me the human body will have a larger impact on a hand held device.  Especially when the case is plastic.

Is there a model that would provide a reasonable approximation of the impact of a human hand for a hand held device (in a plastic case)?