Author Topic: LoRa Library experiences / range  (Read 135077 times)

syrinxtech

  • Sr. Member
  • ****
  • Posts: 347
  • Country: us
    • Syrinx Technologies
Re: LoRa Library experiences / range
« Reply #30 on: June 15, 2017, 07:29:23 AM »
There's some inferential evidence that the xtal in the RFM69 is 10ppm (https://lowpowerlab.com/forum/index.php/topic,1434.msg9957.html#msg9957), so maybe HopeRF uses the same crystal in its RFM95.  Admittedly, that's purely just a WAG.  Wouldn't you be surprised, though, if HopeRF put a worse crystal in the LoRa than it puts into the RFM69?

Here's the rub:  using the Semtech Lora calculator,  it would appear that the required ppm is a function not of the spreading factor but rather of bandwidth.  According to the calculator, at a bandwidth of 31.25kHz (which is what I tested with a pair of Moteino LoRa's), the max crystal offset is 8.5ppm.  In comparison, at a bandwidth of 20.8kHz, the max crystal offset is 5.7ppm, and at a bandwidth of 7.1kHz the max crystal offset is 2.1ppm.  So, maybe that explains why 31.25kHz is the narrowest of the RadioHead preconfigured settings.

So, assuming that's true, is there a way to compensate in software if it turns out that the HopeRF module uses 10ppm or even 20ppm crystals?  Maybe.  Here's an idea, and maybe you all can help me vet it.  According to the datasheet, if you're running the Tx at maximum power, you're limited to a 1% duty cycle.  So, suppose you run the Tx at low enough power that you can instead run it at 100% duty cycle.  [What Tx power would that be?  Simply anything that's not the maximum?]  The idea is that the module would warm up but eventually reach a steady state where the temperature is constant (let's assume ambient temperature is constant, or else  that won't be true and this gets more difficult).  Likewise, run the receiver in constant Rx mode, so that it too reaches a steady state temperature (same ambient temperature assumption).  Let's assume voltages are held rigidly  constant, so that we can ignore that as a possible cause of variances in frequency offset.   Let's assume both sender and receiver are stationary, so g-forces are not a factor.  With those assumptions, is there anything else that might cause variations in the frequency offset?   If not, then the crux of the idea is that the Tx module could programmatically (and purposefully) send packets at a range of frequencies above and below its set frequency with the aim that some of them will closely align with the receiver's Rx band.

If that holds water, then
1. the above method suggests redundant packet transmissions at different frequencies is a shotgun way of getting some packets through, and
2. the receiver could take note of which packets get through the best and adjust its frequency accordingly so as to better align with the Tx frequency.

Through progressive refinement from further iterations, perhaps the receiver could be well tuned to the transmitters frequency, despite using less than ideal crystals.  If the above possibly/probably unrealistic assumptions of duty cycle and ambient temperature needed to support method #2 can't be met, then at least there's still the shotgun technique of method #1. 

At least on its face, does that idea hold water?  Or did I fail to account for something?

Could someone post a specific URL for the Semtech LoRa calculator?  I tried downloading it yesterday and was told I had to create an account first.  After filling in a bunch of information I received an email stating that I wasn't allowed to join their club because my work email indicated I was involved in computer/network security and not IoT.  Therefore I was deemed unworthy to join.  What?  I thought the goal was to encourage as many people as possible to use this new technology?  The reason I wanted the calculator is because I have to LoRa radios on my desk that I want to experiment with.....they said if I could prove I had an interest in IoT I could reapply.  No thanks.  Sorry, stepping down off the soapbox.  Apologies to Felix and everyone else.

WhiteHare

  • Hero Member
  • *****
  • Posts: 1300
  • Country: us

syrinxtech

  • Sr. Member
  • ****
  • Posts: 347
  • Country: us
    • Syrinx Technologies
Re: LoRa Library experiences / range
« Reply #32 on: June 15, 2017, 10:57:03 PM »
Many thanks @WhiteHare!!!


syrinxtech

  • Sr. Member
  • ****
  • Posts: 347
  • Country: us
    • Syrinx Technologies
Re: LoRa Library experiences / range
« Reply #33 on: April 01, 2018, 09:47:30 AM »
Funny.....I joined the Semtech group the other day with no problem and they provided the link so I could download the calculator.

I guess time does heal all wounds and patience is really a virtue.

fgomes

  • Jr. Member
  • **
  • Posts: 65
Re: LoRa Library experiences / range
« Reply #34 on: July 16, 2018, 07:04:44 AM »
Hi, I'm currently testing the RFM95W 868MHz radios (have used the RFM69WH in previous projects), and in the first tests I have noticed a strange behaviour. I'm using the Radiohead lib, and configured the following parameters on both nodes:

    rf95.setFrequency(868.0);
    rf95.setTxPower(23, false); // Also tried with 20
    rf95.setModemConfig(RH_RF95::Bw125Cr48Sf4096);

I'm observing a low RSSI even with nodes close (rf95.lastRssi() reads between -50 and -70 with nodes 2 m apart), and goes down to -100dBm with nodes in the same floor but in separate rooms. I have a setup with the RFM69HW radios where I read about -25 or -30dBm for nodes side by side, and for the nodes in the same floor but in different rooms (exactly in the same places where I tested the RFM95W) I read around -60dBm instead of -100dBm. Do you know any configuration problem that can lead to this behaviour? Or could the transceivers be damaged? I'm using a wire as a monopole 1/4 wave antenna without a ground plane, soldered directly in the transceiver antenna pin (similar situation with the RFM69HW radios).

Didn't have yet the opportunity to do a range test (this would be the next step), but until I can clarify this fast RSSI drop the range test will be meaningless.

Best regards

Fernando


Felix

  • Administrator
  • Hero Member
  • *****
  • Posts: 6866
  • Country: us
    • LowPowerLab
Re: LoRa Library experiences / range
« Reply #35 on: July 16, 2018, 11:54:14 AM »
Can you compare against a known working set of RFM95W nodes?
Are you testing settings randomly to see what works?
How are the RFM95 vs the RFM69HW soldered? You still have a bit of GND plane on the PCB of the module. But you have to ensure you compare apples to apples and not make assumptions that if one works then the other also has to work.

fgomes

  • Jr. Member
  • **
  • Posts: 65
Re: LoRa Library experiences / range
« Reply #36 on: July 16, 2018, 08:37:22 PM »
Hi Felix,

Thanks for your comment, the units are similar with the first experiences I made with RFM69, the transceivers are soldered in a breakout board, connected by jumper wires to ATMEGA boards.  I currently don't have other RFM95W nodes, and rotating the ones I have I get similar results. These were just some quick tests, I have some situations where a bit extra range could be extremely useful (I have to use nodes as repeaters in some situations when I use RFM69 radios), so this was just a simple first evaluation of the RFM95W radio. After your message I've added a ground plane below the radio/antenna (and a decoupling capacitor on the RFM95W power supply pins), the results are now better, but the RSSI is still lower when compared with the RFM69 (and the RFM69 don't have the ground plane below the 1/4 wave monopole). But for the RFM95 it had a positive impact, now I have about -38 / -40 dBm in nodes close, and in the same scenario  I had -100dBm now I'm reading -73dBm, so the antenna ground plane / decoupling capacitor on the radio power supply did have a very positive impact on the RSSI level, all the rest being equal.
But I think that the RSSI difference could also be related to the different technologies and radio configuration used, this was the reason for my question in the first place, is it expected or not that the RSSI level would be different when using RFM95 versus RFM69, both at 20dBm output power?
I've noticed some posts referred to the use of the maximum output power at 20dBm and others at 23dBm (setTxPower), besides the legal aspects, is the RFM95 really capable of generating 23dBm? I've only saw 20dBm on its datasheet.

Best regards

Fernando

Felix

  • Administrator
  • Hero Member
  • *****
  • Posts: 6866
  • Country: us
    • LowPowerLab
Re: LoRa Library experiences / range
« Reply #37 on: July 17, 2018, 09:21:26 AM »
The RFM95/96 can only output 100mW (20dBm).
I would tweak the settings and compare differences there.
Also you can get extended RFM69 range by lowering the bitrates and narrowing the bandwidth. There are various threads in the forum about this, some code shared etc.

fgomes

  • Jr. Member
  • **
  • Posts: 65
Re: LoRa Library experiences / range
« Reply #38 on: July 17, 2018, 11:30:07 AM »
Hi Felix,

Thanks once again, I was just testing the RFM95W to see if they are an easy alternative to longer range scenarios (up to 2km without buildings but with some valleys / mountains between nodes, I know about the RFM69 low bit rate settings and have done some tests as well and I have participated in these discussions, the problem when going that way is the stability with temperature (temperature compensation), frequency offset compensation, etc., but going the LoRa way seems to be also tricky since the long-range configuration settings might also need a TCXO to use high SF and low bandwidth. I'll keep testing both scenarios

Best regards

Fernando