EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: woody on February 18, 2024, 12:54:25 pm
-
I'm working on a Lorawan device. The digital aspect thereof is more or less in my comfort zone but the RF part of it is so far out of said zone that I cannot even see its couch from where I am.
The first revision of my device used a small external antenna with 15 cm of coax and an on-board U-FL connector to connect it to. That performed quite well. Emboldened by that I started to simplify (cheapify) the antenna stuff. So I selected an on-board antenna, a Linx ANT-868-HETH. They have a nice datasheet that details how to use this antenna. Using this I placed the antenna on the board, moved the groundplane back 16.7mm from it, connected it to the RF chip via a 50 ohm microstrip and tested. Only to find that this worked not so well.
I got a LiteVNA in to be able to see at least something. I calibrated the device, connected a length of coax to it, soldered the other end to the PCB, configured the e-delay for the length of the coax and measured (1). Indeed, the SWR and impedance were all over the place.
I then removed the Linx antenna and replaced it with a piece of wire that I shortened until I got a decent result (2). Now the SWR was more or less acceptable. And even the impedance was in the neighborhood of where I wanted it. The performance was slightly better in terms of distance covered.
What I do not understand about this is why the Linx antenna performs so bad, while I (think) I met the requirements for it. Does it mean I always have to tinker with the Pi filter (that atm only contains a 0 ohm resistor)? How do people that are fluent in RF go about this?
For now I think the easy way out for me is to just go back to the external antenna but it leaves me wondering how one would get this right with an on-board antenna.
Maybe someone can point me in the right direction.
-
My experience (as a hobbyist) tuning antennas has been that it is necessary to match them using the Pi filter. Even then, you'll find that antennas easily get detuned by the environment they're in.
-
I recommend you to do the calibration by placing a 50 ohm SMD resistor, open and short (small wire or solder) on C20 pads. Preferably, you can solder the tip of the coax around L2?. And you can place many ferrite beads around the coax connected to the VNA. You don't need to use e-delay with this method.
For example, this is the load I use for calibration before PCB antenna measurements:
https://www.digikey.com/en/products/detail/vishay-dale-thin-film/FC0402E50R0BST1/1769756 (https://www.digikey.com/en/products/detail/vishay-dale-thin-film/FC0402E50R0BST1/1769756)
Is the track between the antenna and C20 50 ohms? What is the height between the track and ground plane below it?
What is the output impedance of the chip? Make sure it is 50 ohms. If not, you are going to use C20, L2? and C19 for impedance matching. You can use a smith chart software to calculate the values.
-
That antenna only quotes an VSWR of 2.2 or better - not a good sign. A simple dipole can get closer to 1 when tuned as you found out, and a full dipole will have better radiation pattern I suspect. But as a few dB here or there is swamped by multipath variation, in urban environments the antenna performance isn't so critical - small antenna size wins.
-
Is the track between the antenna and C20 50 ohms? What is the height between the track and ground plane below it?
It's an Aisler 4-layer PCB: https://community.aisler.net/t/4-layer-hd-pcb-stackup/86 (https://community.aisler.net/t/4-layer-hd-pcb-stackup/86) The antenna track is 245um wide on the top layer. The 1st inner layer is 136um below that. According to Aisler (and the calculator in KiCad seems to corroborate that) this should give me an impedance of 50 ohm. This also gave me a problem: according to the antenna specs no ground plane may be under the antenna. But I need some ground track on the first inner layer to act as part of the 50 ohm track to the antenna. So I had to choose a width for the ground track under the antenna track. I made this 600um. No idea if that is too wide, not wide enough or just OK.
What is the output impedance of the chip? Make sure it is 50 ohms. If not, you are going to use C20, L2? and C19 for impedance matching. You can use a smith chart software to calculate the values.
That is a good question. I reckon it is 50 ohm but could not find anything to back that up in the datasheet. No pointing to other impedances either.
Allow me a question that immediately shows my level of RF knowledge: does a bad impedance matching influence the SWR or are those two different things?
-
does a bad impedance matching influence the SWR or are those two different things?
Yes.
I mean, in most cases VSWR is literally defined by the impedance mismatch ratio.
-
They are mathematically related.
-
I've calculated the same 50 ohm according to your specs. Calculations are correct but as far as I know that formula needs an infinite ground plane to be valid. There were some calculations about fringing waves to ground under the track. I couldn't find them now but I think ground width of 2 times the track width is not enough. You may measure again by calibrating the VNA on antenna entry point. Bypassing the microstrip. Or just try calibration on C20 first.
You should match the impedance of the RF amplifier inside the chip, transmission line and the antenna. You cannot get a good SWR without matching. Since you don't use the chip right now, the problem should be related to the calibration or transmission line. But when you populate the chip on the PCB you will need matching to the chip.
-
You don't need to worry too much about the transmission line impedance when the transmission line is shorter than 1/10 of the wavelength. Also, any efforts at perfecting the transmission line impedance would be killed by the larger solder pads of your PI network, creating a huge mismatch. You would either need to choose much smaller components, or use another layer of your stackup as your ground plane, so that you can make the pads match the microstrip width. But this is not necessary here.
Just
0. keep a distance to any large conductive objects like your ESD-mat, lay the board on top of an empty cardboard box
1. calibrate your VNA using C19 (leave L2 unpopulated during calibration)
2. leave C19 open and populate L2 with a 0 ohm resistor
3. measure impedance at desired frequency
4. calculate PI matching network to reach 50 ohms.
Edit: Ideally, you would also need to match to the output impedance of the power amplifier. For that, you need to do a load pull analysis using an impedance tuner.
-
You should match the impedance of the RF amplifier inside the chip, transmission line and the antenna. You cannot get a good SWR without matching. Since you don't use the chip right now, the problem should be related to the calibration or transmission line. But when you populate the chip on the PCB you will need matching to the chip.
So how is that done then? I started out with placing various L's and C's at L2 and C19 and came upon much better readings using 15nH and 8.2pF, but you tell me I have to do that again when the RF chip is placed. Which makes sense, but how do I measure that? Solder the coax to the contacts the chip is on?
-
0. keep a distance to any large conductive objects like your ESD-mat, lay the board on top of an empty cardboard box
That was my first mistake then. Of course I did all of this on an ESD mat :)
3. measure impedance at desired frequency
How? Do I guestimate this using the green line on the VNA and where it is in relation to the '50 ohm' reference point I get after calibrating using the load?
4. calculate PI matching network to reach 50 ohms.
How? I need some help there :)
And to come back to something @selcuk mentioned, do I install the RF chip while I perform this or can this be done with an unpopulated (apart from antenna and PI network) PCB?
-
How? Do I guestimate this using the green line on the VNA and where it is in relation to the '50 ohm' reference point I get after calibrating using the load?
First you do the regular open/short/50ohm cal just placing nothing, a 0 ohm resistor, a 50 ohm resistor on the C19 pads. After that, placing the cursor at the desired frequency, you should get some reasonable complex impedance value at the top right of the screen.
How? I need some help there :)
https://www.eeweb.com/tools/pi-match/ (https://www.eeweb.com/tools/pi-match/)
And to come back to something @selcuk mentioned, do I install the RF chip while I perform this or can this be done with an unpopulated (apart from antenna and PI network) PCB?
No, don't install the chip for this. What you will also see in many designs, you can just use one matching network to match the chip to 50 ohms (this will usually already be found on a module or a SiP), and another to match the antenna to 50 ohms.
-
I'm a bit confused by your feed line from antenna to the pi matching network. The way it looks to me from datasheet is that it's supposed to NOT be a 50 Ohm transmission line, but rather part of the antenna radiating element. If you put a GND plane under that (and if my understanding is correct) then you've detuned the antenna greatly.
-
I'm a bit confused by your feed line from antenna to the pi matching network. The way it looks to me from datasheet is that it's supposed to NOT be a 50 Ohm transmission line, but rather part of the antenna radiating element. If you put a GND plane under that (and if my understanding is correct) then you've detuned the antenna greatly.
You know, I had the same confusion. If I read page 7 of the DS, the drawing shows both the track before and after the Pi circuit as 'Feedline trace'. As the track before the Pi network is definitely a 50 ohm transmission line I reasoned that the part to the antenna would likely be too. I tried to contact Linx about this but that did not lead to anything.
But, seeing the results, I wonder, maybe you are right and the (very precisely defined, lengthwise) track to the actual antenna is part of the radiating element. It would surprise me, being perpendicular to the actual antenna, but who knows.
Tomorrow I will try @m98 's suggestions and see where that leads me. But I will certainly also look at this suggestion, as it might well explain why this antenna under performs on my PCB.
As an aside, another thing not clear to me was how far I was supposed to let the antenna sink into its mounting holes. The difference between mounting the antenna coils flush on the PCB (and cutting off the excess wire underneath) or use all of the antenna and mount it higher is only 4 mm, but at 868 Mhz that might be important. I could not find this in the DS, except the pictures on page 6, where the antenna is sunk all the way into the PCB.
-
There definitely shouldn't be a ground plane in the exclusion area around the antenna. If there is, that's the problem.
-
I'm a bit confused by your feed line from antenna to the pi matching network. The way it looks to me from datasheet is that it's supposed to NOT be a 50 Ohm transmission line, but rather part of the antenna radiating element. If you put a GND plane under that (and if my understanding is correct) then you've detuned the antenna greatly.
You know, I had the same confusion. If I read page 7 of the DS, the drawing shows both the track before and after the Pi circuit as 'Feedline trace'. As the track before the Pi network is definitely a 50 ohm transmission line I reasoned that the part to the antenna would likely be too. I tried to contact Linx about this but that did not lead to anything.
But, seeing the results, I wonder, maybe you are right and the (very precisely defined, lengthwise) track to the actual antenna is part of the radiating element. It would surprise me, being perpendicular to the actual antenna, but who knows.
Tomorrow I will try @m98 's suggestions and see where that leads me. But I will certainly also look at this suggestion, as it might well explain why this antenna under performs on my PCB.
As an aside, another thing not clear to me was how far I was supposed to let the antenna sink into its mounting holes. The difference between mounting the antenna coils flush on the PCB (and cutting off the excess wire underneath) or use all of the antenna and mount it higher is only 4 mm, but at 868 Mhz that might be important. I could not find this in the DS, except the pictures on page 6, where the antenna is sunk all the way into the PCB.
Definitely sink all the way until the kinks and/or until the coil touches the PCB, the length would be defined by those. There's no other options since it's the only constraining features.
Another aside: a 50Ohm transmission line is a 50Ohm transmission line no matter the length. The fact that they define the length of the feedline trace precisely, is a good hint that it's part of antenna structure.
-
Another aside: a 50Ohm transmission line is a 50Ohm transmission line no matter the length. The fact that they define the length of the feedline trace precisely, is a good hint that it's part of antenna structure.
My fault, I need to revise that a bit; Linx define the distance between the center of the antenna mounting hole and the ground plane as being 16.7mm. That means automatically that the last part of the transmission line is 16.7mm minus half of the solder pad length, 2.55mm, or 14.15mm. But they did not specify the length of the transmission line separately.
I am going to try and contact them again.
-
I set the e-delay to 0 then calibrated the LiteVNA as per the instructions of @m98 (on top of a cardboard box). Set the frequency from 750M to 950M. This gave me a 'flat' SWR graph with a little dent in it. Then toyed a bit with various L and C values for L2 and C20. A zero ohm resistor in L2 and a 1.5pF cap in C20 gave me the result in graph 3.
My question: is this what I am looking for?
-
Yes, this seems better. You can also enable S11 trace for a different view. I recommend you to measure the antenna without matching components and save it to a s1p file on VNA. Then load this file with the below software or similar one and make it calculate matching component values automatically. So that you can confirm your experimental component values.
http://www.ae6ty.com/Smith_Charts.html (http://www.ae6ty.com/Smith_Charts.html)