The LoRa Developer Forum is now in read-only mode and new content will not be added.
Semtech, in its commitment to enhance user experience and streamline content, has successfully integrated the LoRa Developer Portal content into Semtech.com. As a result of this consolidation effort, the LoRa® Developer Portal Forum will be discontinued on May 1st. After this date, you will be automatically redirected to Semtech.com.
For any technical support related to LoRa, please feel free to reach out to our experts here. If you have sales inquiries, please contact us here.

FORUM

[LLCC68] Poor Sensitivity

Semtech Transceivers
#1

Hi Semtech and Everyone,

I have been setting up the LLCC68 for ExpressLRS and run into and issue of poor sensitivity performance.

For example, using BW500 SF7 CR4/7 I can connect a LLCC68 Tx to a SX1276 Rx and packets will be received below -120dBm. But when using the LLCC68 as the receiver, packets will be dropped above -110dBm.

Below is a logic analyzer capture of the LLCC68 Rx from boot for 5s. I am hoping you are able to comment on the setup registers and if anything is incorrect or missing.

20230205_LLCC68_RSSI_Issue.sal https://drive.google.com/file/d/1EGTd6tvaXEU_gDHXnChLU96ZltmU-vpM/view?usp=sharing

Cheers
Jye

SX126x 127x interoperability at 500khz
Lora SX1261 Sensitivity Test
SX1262 sensitivity
#2

Does that translate into a real world sensitivity differance, or could it just be a difference in the calculated RSSI ?

A 10dBm difference would give an actual distance reception difference of about 3, have you been able to confirm this is the case ?

#3

Hey Stuart,

Thanks for the suggestion and we are still testing. Once we have some solid distance numbers Ill let you know.

But that still raises the question of why the rssi is so different to the calculated numbers? For the SX1276 I can easily go another 3dBm lower than its calculated -117 sensitivity for the above parameters. But we aren’t getting anywhere near the LLCC68 calculated sensitivity.

Cheers

#4

I would wait until you have been able to compare the real World sensitivity of the LLCC88 vs the parent SX126x.

Do remember that the ‘calculation’ for RSSI on the SX127x changed considerably over the years.

I dont have a LLC68, but a simple ping test in a large open field would not take long, and you would have a direct comparision between the two devices.

A far better indication of when you are close to link failure, often what is wanted, is the SNR value.

1 Like
#5

Firstly, thanks to Wezley Varty for acquiring the below data.

The below setup shows a SX1276 receiver and a second LLCC68 receiver mounted on a radio controlled wing. Both receivers connect to a single LLCC68 transmitter set to ~1.5mW, receive the same data packets with RF modulation parameters BW500 SF7 CR4/7. The max distance flown in the plots was 15km.

The first plot shows a consistent 20dBm different between the two transceivers, with the LLCC68 being higher.

The SNR plot shows while the number are similar, the LLCC68 has a wider spread of values.

The final plot is a Link Quality metric and shows the percentage of the last 100 successfully received and processed (passes CRC) packets. Its easy to see the LLCC68 is under performing and dropping a lot more packets.

image
image
image
image
image

1 Like
#6

Very good.

The RSSI shows a gap of circa 20dBm, but the SNR gap is only 5 to 7dBm.

So maybe the RSSI calc is a bit off, but there is also a difference in actual sensitivity but its smaller.

One for Semtech to answer me thinks.

#7

Final comment after talking to Semtech and some more testing.

Swapping out the LLCC68 for the SX1262 gave similar results and neither performed as well as the SX1276.

Testing the SX1262 with jgromes/RadioLib and the most basic of settings gave the same result as the ExpressLRS code I have been working on. This is giving me confidence that the reported rssi and SNR values are actually different to the sx1276.

Semtech were also unable to provide a comment on if the LLCC68 performed similarly to the SX1276.

So I think its currently a no go as a substitute for the SX1276 and my use case. Disappointing after all of the time (months) put into it and I was pretty excited about the additional functionality it offered.

Jye

#8

At the reception failure point, I do see a a significant difference between the reported RSSIs, that from the SX126X is maybe 20dBm different to the SX127X, but that could just be an issue with the calculation.

Running decending power tests, SX127X versus SX126X, I do see a small difference in sensitivity of maybe 2dBm (the SX126X being lower) but that could be variations in the modules and their board layouts re the antenna.

I dont have an LLCC68 to test.

#9

Maybe the rssi is just a calculation issue. But the SNR is also a concern as it’s directly related to packet loss. Checkout the example plot below that is taken from a different data set.

It’s calculation might also be offset. In the above SNR plot the LLCC68 has higher SNR in some sections and the packet loss (LQ) is still greater.

163480086-741d52f9-fe38-4b52-99a9-073224a188ec (2)

#10

I just did some checks on SF11, BW125000.

At the failure point the SX127X never reported a SNR lower than -17dB. The SX126X did report some SNRs at -18, -19, -20.

However, ignoring the reported RSSI and SNR, the SX127X was able to receive packets that were between 2dBm to 3dBm weaker than the SX126X could.

However, testing just a couple of modules is maybe a bit too selective, the real world variations in sensitivity appear to be small, so to get a balanced view you would probably want to check a good few examples of each type, different manufacturers modules too.

#11

Annoying as the difference in rssi is, we can work around that. Ultimately what matters is the ability to get packets through. I assume your 2 to 3dB estimate is based on transmitter output setting. It would be nice to correlate Wezley’s LQ data to a dB delta, but we’d need LQ vs txPower relationships to do that, and LQ is horribly noisy data to try and gather. Agreed on the need to test with a few more examples from different manufacturers. If it turns out to be 2 to 3dB across the board then that will be manageable for some use cases and a show stopper for others.

#12

Indeed so.

I had an SX1262 setup as the transmitter, sending packets in this case from 15dBm to -9dbm. The TX power used is in the packet, so the receiver can check and record what power packets it is getting. As long as the transmitter and receiver stay in exactly the same positions and using the exact same antennas, the comparison is valid.

I have checked previously the linearity of the power settings, with an RF power meter, and they seemed reasonably good.

#13

So far I’ve tested the a couple of llcc28, sx1262, Wez the llcc68, and Stuart the sx1262, all with similar results of greater packet loss compared to the sx1276. Coupled with Semtechs responses, I’m not inclined to spend more time tracking down hardware to test.

We could eventually do a multi sx1276 test with different hardware and look at the variation compared to the llcc68/sx1262. But I’m not expecting anything different since there have already been 3 independent comparisons.

#14

Maybe when the weather (here in the UK) is a bit warmer I will compare the modules I have. A walking ping test would only take about 5 mins per module and you would fairly quickly have a dBm relationship between a large range of modules.

#15

Another set of results but this time 433MHz and with a different module. RSSI looks to hit a similar limit as our 900M testing and no packets appear to be received beyond -110.

https://www.lcsc.com/product-detail/LoRa-Modules_DreamLNK-DL-LLCC68-S-433_C2894789.html

image
image1604×778 82.3 KB

#16

Hows the weather going over there :smiley:

About the Ideas category
#17

Another test but this time with the Semtech LLCC68MB2CAS to rule out any question of bad rf front ends on the 3rd party dev boards.

Here is the llcc68 setup and next to it a BETAFPV sx1276 receiver. The test was run twice with the antenna swapped so that any differences in the antenna would be seen. Both are connected to a esp8255 (RF modulation parameters BW500 SF7 CR4/7).

image
image1164×944 466 KB

RSSIdbm is still very different
image

SNR appears to be better at close range but becomes worse than the sx1276 as the signal strength decreases.
image

Relay quality (percentage of the last 100 packets received and passed CRC). The LLCC68 is still performing worse.
image

#18

  1. Your SX1276 receiver has a u.FL cable connected. This might change things. A u.FL cable adds capacitance afaik. It might also be that it dampens the signal and also dampens noise.

  2. Do both receivers have the same power supply? Same ripple voltage etc.? If no, then you should provide a clean power supply to the LLCC68. To me it looks like the LLCC68 is powered by the 3V line of some ESP like thing that might be quite instable.

#19

  1. Adding a pigtail if anything should make the reception worse. But the sx1276 still much better.

  2. Maybe but the results are the same as the previous tests where common power supplies were used.

#20
  1. Yes and no. I have had a similar setup with totally different transceivers, and the one with the pigtail was able to receive weaker signals, while the one with the direct SMA connection was overwhelmed by noise so it would not receive those weak signals. From my own experience I can say: it DOES make a difference.
    You can try it with an SMA to SMA cable - put that in between the antenna and the Semtech SMA connector.

If you’re interested in finding out the reason, you can also try it with a SAW filter. They are available as SMA to SMA adapters, and if you use one for your desired frequency then it will dampen noise by ~40 dBm outside your target frequency. If the SAW filter helps, then you know at least: the reason is noise.