开云体育

Hi,

Is my setup ok ? Is there any monstrous mistakes in my montage ? I'm quite new to the nanoVNA. Well, I am totally new on VNAs themselves. I did not know their exitance 2 months ago ! Well, I want to match my antenna because so far, it is pretty horrible. But, before ordering some matching network component and then screwing everything because I am missing something obvious, I would like someone to point me to the rookie mistakes I could have made.

The project is a LoRa module (902-928MHz) connected to a small antenna mounted on a pcb. The pcb is inside a small polycarbonate box. I have followed the design rules for this antenna from the manufacturer. I have calibrated my nanoVNA-F V2 (from sysjoint) using the supplied calibration set, straight at the S11 connector. My board is inside its final box and mounted on a wood plank that mimic the final location. My board is hooked up to the vna through an RG-174 I had laying around. It is 2 meters exactly. I have added an E-Delay of 10.1ns as per a calculator I found on the Times microwave website. I don't know if there is a bug in the vna's firmware but I had to put 0.66 as the velocity factor in order to make it display 2.00 meters as the cable length. If I put 66 as in 66%, it shows 200 meters. Anyways, I'm not sure if that part is that relevant.

So, by looking at the pictures, does anyone can see some obvious mistakes ?

Thank you so much

it?s close ... but you will get much better results if you use a short cable ... and calibrate at the end of that cable

dg9bfc sigi

Am 23.02.2025 um 04:46 schrieb Nico via groups.io:

May be it's a new version in program, that you must enter the velocity factor as .66 and not in percent. Your measured data are ok, but the 2m RG174 has a loss of 2dB @ 900 Mhz. So the real mismatch at your antenna is -2.34 dB. (You must add 2 x 2 dB). That is very bad for an antenna. A better way will be to use the 2m RG174 with a connector at the end and make a calibration at this end. Thn use a very short cable from that connector to your PCB.

Am 23.02.2025 um 04:46 schrieb Nico via groups.io:

These are very interesting points, thank you very much for that.

I was questioning myself wondering if my technique was appropriate. Thanks again for confirming it can be improved.

I will follow up on that once I could have put my hands on some connectors...

Regards,

On Sun, Feb 23, 2025 at 12:36 AM, Nico wrote:

My board is hooked up to the vna through an RG-174 I had laying around. It is
2 meters exactly. I have added an E-Delay of 10.1ns as per a calculator I
found on the Times microwave website. I don't know if there is a bug in the
vna's firmware but I had to put 0.66 as the velocity factor in order to make
it display 2.00 meters as the cable length. If I put 66 as in 66%, it shows
200 meters. Anyways, I'm not sure if that part is that relevant.

When you use the edelay feature in a NanoVNA to move the reference plane from the SMA connector to a distant point you really need to have a short cable so the attenuation is minimal. This is not the case with 2M of RG174 at 900 MHz as others have pointed out. You really need to establish the reference plane at the end of the cable by "de-embedding" and doing the SOL calibration with the cable open, shorted and using a 50 ohm SMD load.

The other problem you have is what you are attempting to measure. From the pictures it looks like a helical antenna with a short connecting PCB trace that will be a loading inductor. The "other half" of the antenna looks like the ground plane of the PCB. When you connect your RG174 cable to the board the outer surface of the coax shield and the NanoVNA will now be part of the antenna, will radiate and affect your measurement results. You need to slide many ferrite beads of the correct mix over the cable near where it attaches to the board to act as an RF choke. If you shorten the RG174 and see the results change (after calibrating at the end again) then this indicates this kind of problem.

Another thing to consider is that the manufacturer does not specify an antenna with a 50 ohm feedpoint impedance. Maybe the antenna suggested is an end fed half wave and they have a matching circuit. I am only guessing here.

Roger

Nico,
You don't necessarily need more connectors.
You could calibrate at the end of your RG174 cable by detaching it from your board, then using a small 50 ohm chip capacitor, an open circuit, and a direct short as your standards. If the leads are kept very short this should work reasonably well at 900MHz.
Once re-attached to your board, ferrite cores near the antenna end of the cable would be a good idea, too.
--John Gord

as shorfter the cable as better your setup

dg9bfc sigi


Am 23.02.2025 um 20:32 schrieb Nico via groups.io:

Once again,

Thanks to all of you for your inputs. Poorly translated from french, I would say that I did not know in what type of gearbox I was going to put my finger in !

Well, I ran some other experiments tonight. Here is what I've found.

The antenna is an helical monopole where the pcb copper pour serves as the antenna ground plane. I really made sure to follow the manufacturer's layout recommendations. The feed line (as pictured in my original post) shows the feedline also, designed as suggested by LINX. I've included the PI matching network at the very end of the ground plane as also recommended.

This is the datasheet link:


I did what you guys suggested. I've cut a 6in piece of RG-174 and the used a 30cm RG316 extension I have. I did the SOL at the end of the extension and added a delay of a few pico seconds for the remaining piece that attach to the pcb.

@Roger Need: Thanks, Oh, I forgot to mention some board details. I attached pictures of my pcb. My cable is attached where the LoRa radio module will go, as seen on the bottom side. The feedline to the antenna should be a 50 ohms impedance feedline, it goes to a matching network (as recommended by the manufacturer) and then to the antenna. The ground plane dimensions are specified and the distance of the antenna from the ground plane border also, which I've respected. It is a 4 layer PCB. Layer #2 has a ground strip line under the trace that goes along with it that should form the 50 ohm line right down to the antenna itself.

@John, Thanks for your input. could you point me out to some ferrite cores ? I would not know what to look for exactly. If you know some specs and a good source for it, that would be awesome.

Well, I get some interesting result but I'm not sure if I understand what is happening. As you can see with the pictures, if I hold the board in my hands, I get what I would call a "proper" response. But once I lay the boar on the top of the pole, the signal looks like crap. Putting the pcb back into its enclose does not make any significant differences, good or bad.

By holding it and then read something better, does that I mean that my body becomes the ground plane it needs ? Would that mean then that the pcb ground plane is not sufficient in size ? That would be weird as the pcb ground plane is slightly larger than what is recommended by the manufacturer.

Once again, thanks to all of you for lighting up lantern, I truly appreciate your advices.

RF has a steep learning curve ;)

Nicolas

Nicolas,
Your results actually look pretty good. An SWR of 2.4:1 represents a loss of only 17%.
Small changes in the spacing of the turns on your antenna should enable you to put the minimum SWR at your desired frequency. If that does not quite do it, make the antenna wire a little shorter to raise the frequency of best match.
As for the ferrites, I usually save ones cut off of old computer accessory cables (monitors, power supplies). They are often large enough to allow an SMA connector to pass through. You can also get split ferrites which can clamp over cables with connectors that are too large to pass through. There are various ferrite types, but the ones used for interference suppression on cables seem to work fairly well for this kind of use.
--John Gord

John,

Thanks once again for the precious advices. As I’ve put the antenna exactly as I t has been manufactured, could I simply populate the matching network instead of stretching the spirals ? C-L-C network is planned on the pcb, it only misses the parts. I’ve installed simsmith software. Once I’ve put the values from the vna, I will be ready to order the correct caps and inductor to match.

I measured my ground plane yesterday night after posting and it does not have exactly the same size as provided in the data sheet. Mine is a little bit wider and shorter. I would suspect that does not help.

If I want to give it a try and stretch or squeeze the spacing, which direction will increase or decrease the résonnant frequency ?

Thanks,

Nicolas

you should albsoutely calibrate on cable end and use one as short as it is practical (just a few inches would be best!!)

only THEN you have exact values to type into simsmith .. then calculate parts .. solder them in and THEN measure AGAIN ...

(does the transmitter have exctly 50 ohms or is the matching also to transform the trx outputs impedance to antenna impedance???)

just thinking

dg9bfc sigi


Am 24.02.2025 um 17:32 schrieb Nico via groups.io:

Yes, you're on the right track. However, I'd strongly recomment you place
several appropriate clamp-on ferrites on your feedline right where it exits
the assembly. Also place a few at the VNA, itself.

As a test, with the assembly laying on your wooden structure, run your hand
up and down your feedline between the assembly and the VNA. If you note
changes on the VNA while doing so, you require more decoupling - ferrites.

And while grabbing the assembly: when grabbed and the VNA traces change,
something requires decoupling. Certainly it's an indication your hand and
arm are becoming part of the radiating structure - not desirable. A
free-space full wavelength at 902 MHz is 13-inches (33-cm). That PCB is
nowhere near even a 1/4-wavelength. Therefore, a good portion of a "real
antenna" is missing with the assembly.

In your first and third photos, I note what appears to be a metalized
membrane laying below and to the right of your wooden structure. It has a
circle of what appears to be a conductive circle. When the assembly is put
completely together, that may serve as part of the radiating structure?

And remember, SWR isn't a sole indicator of a good antenna!!!. Rely on the
Smith Chart or the complex reflection coefficient, not just SWR. Those two
parameters will tell you a WHOLE BUNCH MORE than just SWR. Besides, we're
not attempting to protect any final amplifier that "requires" a low SWR to
survive. It's such low power that even an infinite SWR likely would not
damage anything.

Dave - W?LEV

<>
Virus-free.www.avg.com
<>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On Mon, Feb 24, 2025 at 2:36?AM Nico via groups.io <nicolassimard=
[email protected]> wrote:

Once again,

Thanks to all of you for your inputs. Poorly translated from french, I
would say that I did not know in what type of gearbox I was going to put my
finger in !

Well, I ran some other experiments tonight. Here is what I've found.

The antenna is an helical monopole where the pcb copper pour serves as the
antenna ground plane. I really made sure to follow the manufacturer's
layout recommendations. The feed line (as pictured in my original post)
shows the feedline also, designed as suggested by LINX. I've included the
PI matching network at the very end of the ground plane as also
recommended.

This is the datasheet link:



I did what you guys suggested. I've cut a 6in piece of RG-174 and the used
a 30cm RG316 extension I have. I did the SOL at the end of the extension
and added a delay of a few pico seconds for the remaining piece that attach
to the pcb.

@Roger Need: Thanks, Oh, I forgot to mention some board details. I
attached pictures of my pcb. My cable is attached where the LoRa radio
module will go, as seen on the bottom side. The feedline to the antenna
should be a 50 ohms impedance feedline, it goes to a matching network (as
recommended by the manufacturer) and then to the antenna. The ground plane
dimensions are specified and the distance of the antenna from the ground
plane border also, which I've respected. It is a 4 layer PCB. Layer #2 has
a ground strip line under the trace that goes along with it that should
form the 50 ohm line right down to the antenna itself.

@John, Thanks for your input. could you point me out to some ferrite cores
? I would not know what to look for exactly. If you know some specs and a
good source for it, that would be awesome.

Well, I get some interesting result but I'm not sure if I understand what
is happening. As you can see with the pictures, if I hold the board in my
hands, I get what I would call a "proper" response. But once I lay the boar
on the top of the pole, the signal looks like crap. Putting the pcb back
into its enclose does not make any significant differences, good or bad.

By holding it and then read something better, does that I mean that my
body becomes the ground plane it needs ? Would that mean then that the pcb
ground plane is not sufficient in size ? That would be weird as the pcb
ground plane is slightly larger than what is recommended by the
manufacturer.

Once again, thanks to all of you for lighting up lantern, I truly
appreciate your advices.

RF has a steep learning curve ;)

Nicolas

--

*Dave - W?LEV*


--
Dave - W?LEV

1) Do you have the eval board? (Always useful to look at a "known good" implementation)
2) I'm not sure it's expecting copper under the coil. The figure in the data sheet sort of looks like only the "shaded" part has ground plane.
3) As others have commented, Make your connector to trace *really* short. (as in <1 cm). Then calibrate at the end of your longer test cable. Bear in mind that the cable *will* have an effect on the pattern and Z, so try moving it to different positions and seeing how much difference it makes.
4) ferrite chokes *might* help, but might not. Choosing an appropriate mix is tricky - you need something that "works" at 900 MHz, and random cores intended for suppression of VHF noise might not do much good at 900 (people run into this when testing GPS antennas at 1.2 and 1.5GHz). Look at the FairRite catalog - you want large loss (R) more than large X at your frequency of interest. A quick check shows that you might be looking for a 61 (200-1000 MHz). 61 peaks at about 250 and rolls off gradually as frequency goes up (and the plot only goes up to 900)

As for stretching or shrinking? (or dropping blobs of wax on it - something that's also done). Most people optimize for received field strength in the assembled unit, rather than adjusting the impedances. What is the *real* output Z of your LoRa module? (yeah, nominally 50 ohms for design, but really? who knows). Then, after you've tuned it up, you can dismantle it and measure it.


The matching CLC matching network might be to accommodate some other modules with different Z.

This kind of thing has a LOT of empiricism in it, unless you've got access to fairly fancy design tools like HFSS that can model it. Note the problems Apple had with the iPhone's antenna (For which the solution was, "don't hold it like that, because it shorts out the feedpoint")

Putting your hand near the antenna (or, really, anything) will also "load" it with a lossy medium, improving the match.

As noted before, though, most people do this by using a tool to design it (e.g. HFSS), build it, check the performance in a "radiated field strength" way, and call it done.
If you need to verify the design - you simulate the entire thing (fixture, connector, etc.) and then compare.
Getting rid of the feedline effects is more art than science. Absorber around the feedline (be it ferrite beads, absorbing tape, or whatever) also is putting absorber in the near field of the antenna.
For what it's worth, there's some interesting tape from Laird that is basically a flexible sticky material loaded with ferrite of some kind or another. We use it for emi suppression at work on equipment that's sensitive.

Let me apologize if my yesterday's post was not clear. The pictures I posted are actually the result of the recommendation you gave me. I took a 30cm RG-316 cable I have and calibrated at the very end. Then I made a 6in RG-174 from that end that connect on the pcb. I added a 760ps delay to compensate for the RG-174. Tonight I will try to shorten it as much as I can and some measurements.

The radio transceiver I intend to use has a 50ohm output. The feedline I created on the pcb should also be a 50 ohm line. I made the pcb fabricated with controlled impedance. This way, the fab house provides the dielectric constant of the pcb stack up and provide trace width recommandation to achieve the desired impedance, which I followed. The antenna I chose is also 50ohm.

I will follow up.

Thanks again

I don't know that shortening it even more makes all that much difference. You can guestimate the loss and to a first order, you're looking for reasonably flat S11 across your band, right? So if 174 it's <1 dB round trip, and you get -10 dB S11, then it's probably really something like -9.

How picky is your source about reflected power?(most low power sources aren't very picky)
-10 dB is only 10% reduction in radiated power (assuming it doesn't reflect back to the antenna and get radiated <grin>)
Those antennas aren't super efficient anyway.

Nico,

A couple of points to consider.

1. I can't tell from your PCB photo if you had tracks on both sides of the board in order to implemented a transmission line from the matching network to the helical antenna. This is required because a single track will make for an additional radiating element. Attached is the manufacturers layout for your antenna.

2. The edelay on most of the NanoVNA products is the round trip delay. You used 760 ps which is only one way. On my NanoVNA I have to enter 1.52 ns. To verify your edelay is correct short the end of 6 inches of RG174 and see if you are on the far left of the horizontal line on the Smith Chart. Adjust your edelay until you are.

3. The edelay method will be OK for Return Loss and SWR measurements. Your impedance measurement (R & X) results will be MUCH better if you calibrate right at the end of the cable with an open, short and 50 ohm SMD chip. Then solder to your board. The edelay at 900 MHz. will not give you good estimates of R and X because of slight loss in the cable and impedance bump at the SMA connection. Try this yourself by measuring a 100 ohm SMD resistor using both methods.

4. You need to use ferrites on the cable which are appropriate for the frequency of operation. I suggest a number of Fair Rite Mix 61 as shown in the attachment. A binocular core with holes big enough for a few turns would help reduce common mode current.

And also remember line lengths on/in the dielectric are shortened by the
reciprocal of the SQRT of the dielectric constant (I assume
FR-4?....nominally 4.2).

Dave - W?LEV

<>
Virus-free.www.avg.com
<>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On Mon, Feb 24, 2025 at 9:57?PM Jim Lux via groups.io <jimlux=
[email protected]> wrote:

I don't know that shortening it even more makes all that much difference.
You can guestimate the loss and to a first order, you're looking for
reasonably flat S11 across your band, right? So if 174 it's <1 dB round
trip, and you get -10 dB S11, then it's probably really something like -9.

How picky is your source about reflected power?(most low power sources
aren't very picky)
-10 dB is only 10% reduction in radiated power (assuming it doesn't
reflect back to the antenna and get radiated <grin>)
Those antennas aren't super efficient anyway.

-----Original Message-----
From: <[email protected]>
Sent: Feb 24, 2025 10:21 AM
To: <[email protected]>
Subject: Re: [nanovna-users] Am I in the right track ?

Let me apologize if my yesterday's post was not clear. The pictures I
posted are actually the result of the recommendation you gave me. I took a
30cm RG-316 cable I have and calibrated at the very end. Then I made a 6in
RG-174 from that end that connect on the pcb. I added a 760ps delay to
compensate for the RG-174. Tonight I will try to shorten it as much as I
can and some measurements.

The radio transceiver I intend to use has a 50ohm output. The feedline I
created on the pcb should also be a 50 ohm line. I made the pcb fabricated
with controlled impedance. This way, the fab house provides the dielectric
constant of the pcb stack up and provide trace width recommandation to
achieve the desired impedance, which I followed. The antenna I chose is
also 50ohm.

I will follow up.

Thanks again

--

*Dave - W?LEV*


--
Dave - W?LEV

Wow ! Thank you so much to every one for taking your time in helping me. I truly appreciate it.

@Siegfried @John
I finally did it. I've cut down my RG174 down to the shortest length I could (35mm). I'v calibrated at the end of an RG316 I have, then connected the "board-to-pcb adapter". I added a 180ps Edelay. I tried to double the value (for round trip) but with no differences.

@WOLEV : The Whole board measures 100mm and the ground plane measure 74mm in length. I'm realizing at writing this this could be a part of the problem ? If this is a 1/4 wave monopole, then it should needs the equivalent as a ground plane, that could mean my ground plane is not long enough ? it should be 84 according to the datasheet. mmm I may have to reorder some pcbs !

@Jim
1- No, unfortunately I don't have the evaluation board. I'm on my own, I know almost nothing in RF !
2- This antenna is designed for high volume applications. In my case the "high volume" will be 2 units !, Anyways, this antenna is sold in either through hole or surface mount. There are squares as pads for the surface mount version. I may have to remove them as I use the through hole one.
Unfortunately, I won't have access to any design tools in that field. Even if I had one, I would not know ho to operate it to extract valuable data...
If my LoRa (20mW of extreme power !) module is not 50Ohms on the output, I will never know and there won't be any ways I'll be able to measure it.

@Roger
1)My pcb stack is a standard 4 layer from JLC PBC with controlled impedance option selected. They offer the calculator for the appropriate trace width. As per the datasheet of the antenna, the transmission line from the RF module to the antenna itself has a ground plane on layers 2 and 3 to offer the return path for the feedline. This is what is recommended in the datasheet. This is what I did and know I understand the reason. So, in theory, from the LoRa module to the antenna, this should be a transmission line and not a part of the radiating element. Thanks for pointing out ;)
2)With my short 35mm adapter tonight, I tried 180ps and 360ps as Edelay with no differences.
3) I tried yesterday to calibrate with the cable shorted, opened and then loaded with a combination of a 47 + 3 ohms of 0805 chip resistor I had on hand. I don't remeber exactly and I don't have any pictures but by memory it gave me negative values as the R. Assembling two chip resistor to make one, not the greatest idea. I will have to order som 50Ohms chip resistor .
Question : Can I use 0805 it should definitely be 0603 ? I've somewhere there could have capacitance effect on a too big of a chip.
4) Thanks for the info on ferrites. I'll give them a look.

So here is my progress so far:
- Tonight I'v cut the cable to 35mm but calibrated at the end of my 30cm RG316. Look at the pictures to see the results.
- I've realized that my ground plane may be too short and have to modify the boards and order new ones.
- I've ordered an RF inductor and capacitor kit for the CLC matching network. I'll give it a try and see from there where it goes.

To be continued...

The feedline is still part of the antenna. The best you could do with this
setup (the first image) is to decouple immediately after the SMA connector
(ideally before the connector). 61 or 52 material ferrite clamp-ons? A
small diameter (1/4-inch or less) single layer coil of possibly three turns
right after the SMA might also do the job. You need to get the outer
surface of the coax out of the picture. As it is, it forms part of the
antenna.

With the setup in the first image, lightly grab the coax and run your hand
up and down the coax between the PCB and the VNA with your fingers around
and touching the coax. Does anything on the VNA change? If not, you're
sufficiently decoupled. If so, and things change on the VNA while doing
so, you need to decouple the outer surface of the coax shield from being
part of the antenna.

Dave - W?LEV

<>
Virus-free.www.avg.com
<>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On Wed, Feb 26, 2025 at 3:57?AM Nico via groups.io <nicolassimard=
[email protected]> wrote:

Wow ! Thank you so much to every one for taking your time in helping me. I
truly appreciate it.

@Siegfried @John
I finally did it. I've cut down my RG174 down to the shortest length I
could (35mm). I'v calibrated at the end of an RG316 I have, then connected
the "board-to-pcb adapter". I added a 180ps Edelay. I tried to double the
value (for round trip) but with no differences.

@WOLEV : The Whole board measures 100mm and the ground plane measure 74mm
in length. I'm realizing at writing this this could be a part of the
problem ? If this is a 1/4 wave monopole, then it should needs the
equivalent as a ground plane, that could mean my ground plane is not long
enough ? it should be 84 according to the datasheet. mmm I may have to
reorder some pcbs !

@Jim
1- No, unfortunately I don't have the evaluation board. I'm on my own, I
know almost nothing in RF !
2- This antenna is designed for high volume applications. In my case the
"high volume" will be 2 units !, Anyways, this antenna is sold in either
through hole or surface mount. There are squares as pads for the surface
mount version. I may have to remove them as I use the through hole one.
Unfortunately, I won't have access to any design tools in that field. Even
if I had one, I would not know ho to operate it to extract valuable data...
If my LoRa (20mW of extreme power !) module is not 50Ohms on the output, I
will never know and there won't be any ways I'll be able to measure it.

@Roger
1)My pcb stack is a standard 4 layer from JLC PBC with controlled
impedance option selected. They offer the calculator for the appropriate
trace width. As per the datasheet of the antenna, the transmission line
from the RF module to the antenna itself has a ground plane on layers 2 and
3 to offer the return path for the feedline. This is what is recommended in
the datasheet. This is what I did and know I understand the reason. So, in
theory, from the LoRa module to the antenna, this should be a transmission
line and not a part of the radiating element. Thanks for pointing out ;)
2)With my short 35mm adapter tonight, I tried 180ps and 360ps as Edelay
with no differences.
3) I tried yesterday to calibrate with the cable shorted, opened and then
loaded with a combination of a 47 + 3 ohms of 0805 chip resistor I had on
hand. I don't remeber exactly and I don't have any pictures but by memory
it gave me negative values as the R. Assembling two chip resistor to make
one, not the greatest idea. I will have to order som 50Ohms chip resistor .
Question : Can I use 0805 it should definitely be 0603 ? I've somewhere
there could have capacitance effect on a too big of a chip.
4) Thanks for the info on ferrites. I'll give them a look.

So here is my progress so far:
- Tonight I'v cut the cable to 35mm but calibrated at the end of my 30cm
RG316. Look at the pictures to see the results.
- I've realized that my ground plane may be too short and have to modify
the boards and order new ones.
- I've ordered an RF inductor and capacitor kit for the CLC matching
network. I'll give it a try and see from there where it goes.

To be continued...

--

*Dave - W?LEV*


--
Dave - W?LEV