Alternative PCB Dipole Designs (433, 868, and 915MHz)

[LukaQ]

@LukaQ, can you take a pic of your setup? Thanks!

Here you go

[LukaQ]

How is it with radiation patent since it has ground plane on other side of antenna. Simple wire dipole would have active upper element and lower ground element. How does this one work?

[Felix]

LukaQ,
Not meaning to hijack your current quest, but have you considered the velocity factor of your coax? what kind of coax is that?

[LukaQ]

LukaQ,
Not meaning to hijack your current quest, but have you considered the velocity factor of your coax? what kind of coax is that?

I have not, I went with length that is from this page for the 868. Coax is one from ebay sold as rg316. VF should be probably 0.7?
So what does that mean, that is has to be longer than signal would be in air?

[ChemE]

I have not, I went with length that is from this page for the 868. Coax is one from ebay sold as rg316. VF should be probably 0.7?
So what does that mean, that is has to be longer than signal would be in air?

For 868MHz, the ideal feedline length accounting for velocity factor was calculated as being 11.40 cm from the ANT pin of the RFM69 to the closest edge of the PCB dipole.  I'm not sure why you are seeing such a strong signal attenuation.  In my own testing with different feedline lengths at 915MHz, going up or down by a few centimeters didn't strongly affect average RSSI.  I use code which reports the average RSSI so it will drift towards the true mean.  I usually collect data from 100 packets sent once per second before recording a result to help account for environmental noise.

[LukaQ]

For 868MHz, the ideal feedline length accounting for velocity factor was calculated as being 11.40 cm from the ANT pin of the RFM69 to the closest edge of the PCB dipole.  I'm not sure why you are seeing such a strong signal attenuation.  In my own testing with different feedline lengths at 915MHz, going up or down by a few centimeters didn't strongly affect average RSSI.  I use code which reports the average RSSI so it will drift towards the true mean.  I usually collect data from 100 packets sent once per second before recording a result to help account for environmental noise.

What do you mean, strong signal attenuation? That is not attenuation on feedline. Those RSSI were between two nodes some distance apart, no averaging, I saw RSSI to be ~ -56 After adding first feedline on one node, RSSI became much better, adding 2nd one and RSSI went up again. All in all, just by adding feedlines (instead of having antenna soldered to AND pin of RFM), there was 6db or more gain (lower RSSI). One node is pretty much 11.4cm, other one as a bit short at ~11cm for 868MHz. And with that being said, if someone else goes for such approach, you need to add feedline or you'll be pi**ing away sensitivity for nothing

[ChemE]

Ah ha, my mistake.  I hurriedly read your private message and then the new posts on this thread and completely misread the numbers!  Glad to see that the feedlines did what they were supposed to do and you confirmed as much with RSSI numbers.

[tbalon]

I had ordered ChemE's version 3 dipole design and have some concerns. It's seems that the
wider trace on this version, or perhaps the board material has a different velocity factor. It
seems its elements (1/4 wave) should be longer (closer to 78mm than 60mm.)

I can see that capcha's design v1 pcb uses a narrow trace element. I was confused by the
60mm length as a 1/4 wave 915 Mhz element should be 78mm (about 3"). His tuning was
done presumably to deal with the velocity factor of the copper trace track with 1mm on FR4 of 0.77
which yields 78mm * 0.77 or 60mm which is what he made the trace length on his v1.0 915Mhz
PCB antenna.

I tested the Captcha v1.0 version and used it with a LORA mote. It worked very well. It worked even
better when I optimized the feed line length as discussed here to make my coax the optimum
length. Attaching Captchs dipole to my VNA I got a VSWR of about 1.6 with an impedance of 52 ohms
at 915 Mhz.  (This was using a 15cm feed line I had for testing so it was not ideal. )

Using ChemE's version 3.0 with the same feedline I got a VSWR of 5.1 with an impedance of only about
15 ohms. at 915 Mhz. However, at 1100 Mhz, the VSWR drops below 2 and the antenna's impedance is about 50
ohms. It took some time for me to believe this so I double checked the antenna connections and tried again. The
VNA showed ChemE's antenna had a nice wide bandwidth that was shifted more toward 1.1 Ghz.

So, ChemE's version 3.0 appears to be a nice antenna, but it's tuned for 1.1 Ghz not 915 Mhz. I believe this is
due to the shortness of the elements which were calculated by Captcha for 1 mil copper trace by applying the
velocity factor for that particular trace on FR4. I believe the easiest way to address this would be to increase
the length of the elements so they are closer to 1/4 wave (i.e. 78mm ).

Here are some photo's of the tests. https://photos.app.goo.gl/JmiqdGHLOQTF2m6m1

[ChemE]

Wow tbalon that is one heck of a first post!  Welcome to the forums!  That is awesome that you tested my adaption of captcha's work, thanks for the data and write up.  I'm all too happy to increase the length of the resonators (even though the boards will be a little more expensive) if it means we all get better performance.  As I said when I set out on my adaption, I don't have a VNA and have no plans to buy one and learn how to use it, so hopefully captcha's work transfers directly.  It would seem that perhaps it doesn't and the design can be further optimized.  I happen to have the 868 and 433 MHz versions of my antenna here so I can do some RSSI measurements using those longer resonators and see if they do indeed improve upon the RSSI at 915MHz.  Longer term, I can post revision 4.0 designs which have longer resonators and hopefully you'll be willing to test them as well.  If you prefer not to buy copies of those boards, I can and ship them to you if you are in the US.  I'm curious to see what others think since RF is most certainly not my forte.

[tbalon]

Wow tbalon that is one heck of a first post!  Welcome to the forums!  That is awesome that you tested my adaption of captcha's work, thanks for the data and write up.  I'm all too happy to increase the length of the resonators (even though the boards will be a little more expensive) if it means we all get better performance.  As I said when I set out on my adaption, I don't have a VNA and have no plans to buy one and learn how to use it, so hopefully captcha's work transfers directly.  It would seem that perhaps it doesn't and the design can be further optimized.  I happen to have the 868 and 433 MHz versions of my antenna here so I can do some RSSI measurements using those longer resonators and see if they do indeed improve upon the RSSI at 915MHz.  Longer term, I can post revision 4.0 designs which have longer resonators and hopefully you'll be willing to test them as well.  If you prefer not to buy copies of those boards, I can and ship them to you if you are in the US.  I'm curious to see what others think since RF is most certainly not my forte.

I'm in Toronto Ontario and I would be happy to test your v4 board.  I did another test where I scraped off some of the
laminate to expose the copper at each end of the dipole. I then attached copper foil and extended the length to about
78mm and tested again. The results looked very good .. VSWR dropped to 1.2 at 915 Mhz.

This was only a quick test, but I think its fair to say the longer elements did help. I was trying to find out the influence of
the copper trace width .. but it could also be attributable to the substrate / board material.

I've added a few pics of the antenna with copper foil soldered to the ends to extend the elements to get a better match
at 915 Mhz.  https://photos.app.goo.gl/JmiqdGHLOQTF2m6m1  What I like about your version is the increased bandwidth.
LORA transmits in the 900-905 Mhz range (uplink) the downlink is higher, up to 927 Mhz I believe. The wider bandwidth
will perform better in this case.

Happy to help...

[ChemE]

Wow a VSWR of 1.158 means that about 99.5% of the RF power is being radiated to air as far as I can tell.  I believe it was Joelucid that posted a link to a paper which claimed wider resonators are more wide band and perhaps change the optimal length shorter.  Glad to see at least one of those claims is true!  I'll post a link to rev4 later today with slightly over-sized resonators that can be trimmed at your test bench to confirm that indeed on FR4 78mm is optimal.  I'm thinking to make the resonators 82mm would you suggest a different starting point?  It is nice to see data each mm and confirm indeed that 78mm is better than 77mm and 79mm.

[ChemE]

Here is the link to the refined 915MHz design: https://PCBs.io/share/zOdQn.  I went for 80mm resonators since the math is very easy that way.  It is interesting that these may work particularly well for LoRa.  I haven't yet fooled with LoRa but it is on my list and now that the SX1261 has been announced: https://www.semtech.com/products/wireless-rf/lora-transceivers/sx1261 with its 4.2mA receive current, it has moved higher on the list.  Hopefully Hope has some transceivers soon that I can play with.

[perky]

Careful, the extensions have no dielectric under them so the velocity factor for that part is about 0.98 while the rest is lower (it seems greater than 0.77, but in theory you should be able to work it out from the data you now have), so the PCB trace should be shorted a little to take that into account.

Mark.

[tbalon]

Careful, the extensions have no dielectric under them so the velocity factor for that part is about 0.98 while the rest is lower (it seems greater than 0.77, but in theory you should be able to work it out from the data you now have), so the PCB trace should be shorted a little to take that into account.

Mark.

Yes indeed. As I stated this was a quick test to illustrate my belief that the elements should be closer to the 78mm than 60mm that
the v3.0 boards were built. The photo shows that adding length made the antenna work better for the freq of interest ( 915 Mhz )

[tbalon]

Here is the link to the refined 915MHz design: https://PCBs.io/share/zOdQn.  I went for 80mm resonators since the math is very easy that way.  It is interesting that these may work particularly well for LoRa.  I haven't yet fooled with LoRa but it is on my list and now that the SX1261 has been announced: https://www.semtech.com/products/wireless-rf/lora-transceivers/sx1261 with its 4.2mA receive current, it has moved higher on the list.  Hopefully Hope has some transceivers soon that I can play with.

I believe the VF of FR4 should be in the range of 0,84 - 0,86 so a length between say 65.5 mm and 67 mm would be my guess.

Tom