Author Topic: Weird antenna problem, low RSSI [SOLVED - with proper grounding]  (Read 16098 times)

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Weird antenna problem, low RSSI [SOLVED - with proper grounding]
« on: February 06, 2016, 05:40:34 AM »
Hi again,
I've managed to build a few nodes using Mega2560, level converter, AM1117 LDO, using pretty long breadboard cables and lousy RF wiring, antenna measured by eye, all working good and stable. Excellent RSSI. (Maybe from the perfume chip box ;D )




Then I've crammed all up on some piece of perfboard, keeping all wires short, neaty, tidy.




And problems start to arise:
-only big packets going through, no ACK gets across
-replacing chip roles (send/receive) the same happens
-it only works when "cuddling" the chip, touching GND or the antenna
-the antennas were exactly 86mm long, as calculated for 868Mhz, tried trimming them to 85...84...83..., nothing
-added decoupling caps 10nF, 100nF, 10uF, 100uF, nothing

What i tried:
CUT the antenna in half and it started working but with really low rssi, like -70 at a few centimeters apart, but I get ACKs.
REPLACED antennas with 12cm antenna, works but still low rssi
CURLED 12cm antenna as in the picture, the signal is -93 at 5 meters through a door, direct sight, radio set at full power.

I've read somebody else having the same "chip needs cuddling" problem and it was due to some missmarked chips. So I did try the 433Mhz setting with the 86mm antennas but only got worse rssi and working seldomly when touching the antenna. Felix maybe you can shed some light on how to identify the correct freq of the RFM69HW.
And maybe somebody can explain how come the chip refuses to work when using the right antenna length but starts working when using a much shorter or longer antenna.

Thank you.


« Last Edit: February 07, 2016, 06:15:34 PM by Felix »

emjay

  • Full Member
  • ***
  • Posts: 119
  • Country: nl
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #1 on: February 06, 2016, 11:26:38 AM »
The frequency band of the RF Module is conveniently marked on the underside (small white silkscreen table) - once soldered in place, IIRC there is no way easy way to determine this from the topside.


john4444

  • Jr. Member
  • **
  • Posts: 71
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #2 on: February 06, 2016, 11:52:15 AM »
Hi Execu,
A reasonable guess for the poor performance would be that the ground wiring is long and thin.
You might experience much better performance with a ground-plane under the radio.
John AE5HQ

WhiteHare

  • Hero Member
  • *****
  • Posts: 1300
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #3 on: February 06, 2016, 12:54:48 PM »
Probably EMI/RFI are also factors:   https://lowpowerlab.com/forum/index.php/topic,1538.0.html

Also, your RSSI numbers are quite high, no doubt from having them so close together.  That could mean more RFI through those long dupont wires from the close range.  Try putting them on opposite sides of the room and running them directly from 3v battery power.  That would narrow the list of suspects.  Any difference?
« Last Edit: February 06, 2016, 01:14:59 PM by WhiteHare »

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #4 on: February 06, 2016, 02:27:10 PM »
Thanks for the support guys.

@emjay The modules are marked as 868mhz HW variant, bought from tme.eu (pretty decent retailer). I've read somewhere around here about bad marked modules and Felix mentioned something about different BOM, if I understood right a cap is missing for the 433 variant, and another small piece is different between 868 and 915 Mhz. Maybe he can say which one so I can measure it just to be sure mine are the right ones.

@john I'm sure you are right, but for now it's the best I can do, I'm planning on building some Gigainos (2560+rfm) after I move in the new house and have decent space for electronics lab. For now stock Arduino Mega 2560 + a few miles of wrap wire must work :)

@White: that's the problem, first pictures show extremely high rssi even though there are a lot of wires to create interference and capture emi, getting the chips close together and shorter wires has worse rssi.

For the moment I use straight wires trimmed for max rssi on the other side, the length varies a little between the modules I've built, or maybe my trimming is pretty coarse, about 90-95mm seems to be the sweetspot for about -70rssi in the other room.

WhiteHare

  • Hero Member
  • *****
  • Posts: 1300
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #5 on: February 06, 2016, 02:56:41 PM »
You  really should try running them on  separate batteries.  Also, try keeping your datalines as far away from your power lines as you possibly can.

If you ever do figure it out, I hope you post what the cause was, as this is more entertaining than most mystery novels.  So far the best clue is that the "cuddling" has an effect.  When you say that long packets are received, but not ACKs, I assume you're referring to two different receivers, or the same receiver?  That would be another significant clue, and I'm guessing the former.  Try reversing the sketches on each node and see if that still holds true.

BTW, if it matters to you, good luck doing wireless programming with a 2560 "Gigaino".  The usual bootloader won't work on a 2560, at least not without modification.  You may not care about that now, but later on you probably will.  It really is a pity.
« Last Edit: February 06, 2016, 03:44:24 PM by WhiteHare »

john4444

  • Jr. Member
  • **
  • Posts: 71
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #6 on: February 06, 2016, 07:44:08 PM »
What you describe sounds to me like poor RF grounding.
As for the ground-plane:
A piece of circuit board works great as a ground plane but, you could also use a small piece of sheet-metal (soup can).
It does not need to be perfectly sized. You could probably slip it under the radio with an insulator
such as cardboard or electrical tape to prevent shorting. I'd bend the edges up to make soldering to
the radio's ground connections easier.
I would be interested to see if it was helpful or not.

Good Luck, John

John AE5HQ

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #7 on: February 07, 2016, 05:46:37 AM »
As for clues:
-sender says: sending...nothing
-receiver says: packet received. Pinging...nothing
Reversing send/receive sketch same thing happens. So the "main packages" get trough but acks do not.
Ruled out SPI problems since receiving works and printing registers spits the same values for different modules.

Shall try a groundplane or dipole later today. Computed the length to be 82mm for 868Mhz and that's what I'm aiming for. I'll keep you posted.

About the bootloader I've tried asking Joelucid for his bl as it doesn't require an external flash chip but had no luck. So I shall make my own starting either from stkv2 or the modified 2560 optiboot. That's the next thing on the todo list.

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #8 on: February 07, 2016, 11:50:28 AM »
As I've promised an update:


Node 20 is the node without ground plane/shielding, uncurled antenna, 82mm vertical
Node 21 is the node WITH a piece of grounded tin can under the rfm chip, uncurled 82mm vertical antenna

The gateway is an RPi type B with a pro mini 5V + level converter + ldo + rfm 69 with an uncurled vertical 120 mm antenna, placed in another room, about 3m away, a concrete wall separates the rooms, there is a doorway between.

Conclusions:
1. there is no significant difference between node 20 and node 21, so the ground plane / shield is not any improvement. the rssi influenced more by the position of the node's antenna (more/less vertical).
2. touching the 82mm node antenna yields a rssi drop of 5-6db, so for the nodes the 82mm antenna seems to be in tune! Cuddling not allowed!  ;D

3. the rpi hat does not work with an 82mm antenna, can see packages received but can not send. The nodes get no ack. The server pings get no ack back. Touching the 82mm antenna (cuddling) makes the system work.
4. the rpi hat can send and receive with an 120mm antenna as seen in the picture.

Maybe I should join forces on the other thread about emi on the rpi.
« Last Edit: February 07, 2016, 11:52:50 AM by executivul »

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #9 on: February 07, 2016, 12:53:01 PM »
THE LAST PIECE OF THE PUZZLE!



I thought to myself: I must try a dipole on the RPi hat, maybe that works... [Tweety voice]It did! It did! It did work![/Tweety voice]
Just adding an 82mm piece of wire to the GND pin next to the ANA pin, oriented the other way made the RPi hat work perfectly with the 82mm antenna.

Next thing: revert to the 82mm monopole + ground plane. Scavenged some more metal sheet and it works like magic! The signal received now is -62 with node power set to 0 (zero) and with 3 concrete walls between (2 rooms), a position in which my phone does not get wifi from the router placed in the same spot as the RPi.

Final conclusion: ground plane or shield the metal back does it's thing in a noisy environment! Thanks John!
The nodes on the table do not seem to need it but they just might need when crammed in wall boxes.

PS. Now crying over some old computer case I've thrown away with a solderable metal sheet the size of a letter paper  >:(

LE. please excuse the multi posting, I posted as I've worked, no post hunting intention whatsoever.
« Last Edit: February 07, 2016, 12:55:08 PM by executivul »

WhiteHare

  • Hero Member
  • *****
  • Posts: 1300
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #10 on: February 07, 2016, 01:52:52 PM »
Great job solving it.   Would please post a photo of the other side also so that we may fully admire what you did?

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #11 on: February 07, 2016, 02:43:36 PM »
Of course, there it is.
Only one of the corners connects to the GND pin next to the ANA pin, the rest of wires are used just to keep things in place.
Sorry for the phone camera quality.

WhiteHare

  • Hero Member
  • *****
  • Posts: 1300
  • Country: us
Re: Help identify RFM69HW variant. Weird antenna problem.
« Reply #12 on: February 07, 2016, 03:04:09 PM »
Thanks!   It's interesting that such a small ground plane would do the business.  I had thought it would require at least a 3 inch diameter, based on k2ox's antenna  posting.  I wonder why it works?

john4444

  • Jr. Member
  • **
  • Posts: 71
  • Country: us
Re: Weird antenna problem, low RSSI [SOLVED - with proper grounding]
« Reply #13 on: February 07, 2016, 11:01:31 PM »
WhiteHare,

RF can be pretty mysterious.
However, I suspect that the ground-plane may be more for the matching network than the antenna.
A good matching network can make almost anything radiate.

Regarding antennas:
I have experimented with a single radial, multiple radials, a large ground-plate, a counterpoise and a J-pole antenna.
None of those 'grounds" or antennas were significantly better than the straight-wire antenna soldered to the radio.
John AE5HQ

executivul

  • NewMember
  • *
  • Posts: 48
  • Country: ro
Re: Weird antenna problem, low RSSI [SOLVED - with proper grounding]
« Reply #14 on: February 08, 2016, 03:21:02 AM »
"Suppose you have a large container you can’t lift.  If it is setting in your driveway you can push against it and it will move in the direction it’s being pushed.  You are delivering power to the load.  Now suppose the container is located on a frozen pond.  When you push, your feet slip and the box doesn’t move.  You can’t deliver any power to the load.  The same goes for your radio.  It needs something to push against." john k2ox https://lowpowerlab.com/forum/index.php/topic,181.0.html

-Using the modules with separate chips tied together by dupont wires I have the GND dupont which is about the same length and diameter as the antenna and such it formes a dipole, malformed and not 180 deg rotated, but still a dipole. It can "push against" the negative wire even though it's also used for the dc power.
-Using the modules soldered on perfboard still uses dupont cables to the Arduino micro, so the same thing applies, the kynar between rf chip and the dupont acts almost the same as a pcb track between the chip and a sma connector, just adds some mismatch.
-The Rpi hat has no such long thick wires, the 30 awg kynar is not enough, it's too thin and too short.
-Using the modules with short kynar GND instead of dupont still works for now, but with degraded signal and lost packets!

So the nodes shall need ground planes too, or a dipole, when finally deployed since then I will use only kynar to connect them to the main board.