Author Topic: Which RFM69 error rates do you get with weak signal?  (Read 1160 times)

curious

  • NewMember
  • *
  • Posts: 7
Which RFM69 error rates do you get with weak signal?
« on: October 24, 2019, 02:07:04 PM »
My own tests have shown me that if the RSSI falls below -95 @ 19.2 kbps bit flipping starts to occur (FSK modulation). The transmitted ~20 byte message gets corrupted.

I stumbled upon this pdf: https://forum.pjrc.com/attachment.php?s=5f386358d2be4d118c8c92235deb6bbb&attachmentid=2719&d=1411542036
In section 3 it states that for a 10 byte payload @ 19.2Kbps you need an RSSI of -81 for <3% retransmission rate.

In the LowPowerLab Forum you can find lots of range tests, but in none of them could I extract information about what kind of messages were sent or if they were received malformed.
Has anybody here done tests and can give some figures about @ what RSSI / speed, you get some kind of bit flipping (=transmission errors)?

I must assume that my tests are misleading because if transmission does not work reliably at all @ -95 dBm, what use is a reception sensitivity of -120 dBm?
« Last Edit: October 25, 2019, 10:48:36 AM by curious »

curious

  • NewMember
  • *
  • Posts: 7
Re: Which RFM69 error rates do you get with weak signal?
« Reply #1 on: October 25, 2019, 10:46:55 AM »
A further test with a different board, again using the RFM69, showed me an RSSI down to -106 dBm using the same sender.
Even @ -106 dBm the received message was uncorrupted.

So, it seems to be possible to receive weak signals and still receive a message without errors.

Felix

  • Administrator
  • Hero Member
  • *****
  • Posts: 6866
  • Country: us
    • LowPowerLab
Re: Which RFM69 error rates do you get with weak signal?
« Reply #2 on: October 25, 2019, 11:28:09 AM »
Are you planning to to keep your RSSI around -95 to -100dBm?

I would not expect reliability at such weak signal with the default RFM69 library settings. Some packets might get through but basically you're drowning in the noise floor. To get a fair chance to send packets in such noise you need to drastically reduce bitrate.

curious

  • NewMember
  • *
  • Posts: 7
Re: Which RFM69 error rates do you get with weak signal?
« Reply #3 on: October 25, 2019, 12:09:25 PM »
I'm not planning to stay @ -95 dBm all the time, but I want to figure out how far I can go and how to maximize efficiency/range.

I'm using different versions of custom boards - battery powered or powered via a step down converter and trying to understand if it is feasible to reliably receive messages @ -100 dBm with my settings.
Because if it is, then I need to optimize my design so that this goal is achieved. If it is not, then there's no use in spending more time with EMI optimization, ferrite beads, inductors and so forth.

From what I can see so far it seems to be possible - on a battery powered node at least.
The upcoming tests will show if data gets corrupted as soon as I use USB power or power via a buck converter / different boards.
« Last Edit: October 25, 2019, 12:13:44 PM by curious »

Felix

  • Administrator
  • Hero Member
  • *****
  • Posts: 6866
  • Country: us
    • LowPowerLab
Re: Which RFM69 error rates do you get with weak signal?
« Reply #4 on: October 26, 2019, 09:00:11 PM »
What boards are you testing with?
The datasheet is for the SX1231 transceiver, not for the RFM69 module. The hoperf just slapped their logo and replaced photos and called it their work in typical shameless chinese style.
But if you want to build your own board then you can probably do it better than you see on the RFM69 modules. Even though, these modules have been used in so many places, even high end and extreme stuff, and they work well, that I don't think your efforts are worthwhile unless you like the hobby and have the time and slightly better margins give you the satisfaction.
Otherwise ... use what works, don't fix what's not really broken, and move on. Life is too short.

harbinger

  • NewMember
  • *
  • Posts: 1
  • Country: ua
Re: Which RFM69 error rates do you get with weak signal?
« Reply #5 on: October 28, 2019, 03:10:50 AM »
(Sorry for probably bad English). One of the reasons may be "dirty" power supply, due to which the noise of the VCO increases and therefore the sensitivity degrades. In my practice, using LD1117 effective sensitivity was about -75 dBm (BER=0,1%) at 38400 bit rate and 62.5 kHz RBW. After replacing LD1117 by TPS7333 sensitivity has come to normal value ~ -105 dBm.
« Last Edit: October 28, 2019, 03:21:19 AM by harbinger »

curious

  • NewMember
  • *
  • Posts: 7
Re: Which RFM69 error rates do you get with weak signal?
« Reply #6 on: October 28, 2019, 10:16:28 AM »
I'm testing on custom built boards with RFM69W/HW modules soldered on them.

My last tests have shown that @ about -100 dBm errors start to show up in the received message.
Sensitivity (= weakest signal received) is
a) ~ -117 dBm for battery powered nodes
b) ~ -115 dBm for a node powered with clean power supply + LC filter + low noise regulator
c) ~ -104 dBm for a node powered with noisy power supply + LC filter + low noise regulator

The RSSI reading seems to be not so precise - sometimes I get a valid message @ -106 dBm, sometimes I get an invalid message @ -100 dBm with the first 2 options a+b.

Option c) I will investigate further...

Possible reasons for bad results with option c):
EMI from nearby board
high frequency noise picked up by power line of RFM69 (I can measure up to ~200 MHz, radio frequency is 868 MHz though)
maybe 48MHz MCU noise (8MHz oscillator) picked up by RFM69 but that would have to show up in option b) too