This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Semtech LoRa vs Performance line TI

Other Parts Discussed in Thread: CC1125, CC1310, CC1120

Semtech provides a new family LoRa with extremly high link budget and low power consumption.
For example SX1276 has a 168 dBm link budget. This is much more that CC1125 provides (145 dBm).
Difference is 23 dBm. It's approximately 16x times more range. It's possible?
Semtech presented a new type of modulation in this product family - LoRa. What can TI provide for competing with this technology?
Application requirements - 868 MHz, 4.8 kBod, output power 10-100 mW, lowest power consumption in receiving mode. SX1276 has a 9.9 mA in RX mode, TI has a maximum current 17 mA.

  • The 168 dBm link budget comes from the SX datasheet number of -148 dBm sensitivity. This sensitivity is at 18 bps (yes, 18 bits per second), which is not a usable rate for any application, just a showcase number for the datasheet frontpage. For a rate of 4.8 kbps, a solution based on TI Narrowband technology will give more range and a more robust link than a link using LoRa.

    The modulation format in LoRa is Chirp Spread Spectrum (CSS), a technology invented in the 1930's. CSS is widely used in radar and ranging applications, but not in communications systems. The main reason CSS is not used in communication systems is the very weak co-existence performance, i.e. the ability to reliably operate in presence of interference. LoRa spread the signal information over a wide bandwidth (500 kHz), but will be jammed by any other system operating in this bandwidth (check out the co-channel rejection on page 21 in the SX1276 datasheet.....). Semtech claim LoRa has great co-existence, but their own datasheet measurements show performance 50 dB worse than TI narrowband solutions.

    Using a wide bandwidth to send a low rate signal, using a modulation format that does not tolerate any activity in the used bandwidth, this is something that will not work in a real environment with real interference. The de-facto standard for long range communication is narrow band, this offers the best range, co-existence and robustness. For more details also check out the white paper on long range RF communication:

    http://www.ti.com/lit/wp/swry006/swry006.pdf

  • Note also that you can take advantage of the RF Sniff Mode feature to lower the CC112x/CC120x RX current consumption. Please refer to  www.ti.com/lit/SWRA428

  • I have no affiliation with Semtech, but the CC1125 is obsolete versus the SX1276.  If you compare co-channel rejection, and sensitivity at equivalent bit-rates, the SX1276 wins, in both narrow and wide band specs, and power consumption.  That being said, Semtech may be fudging their specs, who knows unless a real world comparison is done.

    For their Lora receiver, they spec positive co-channel rejection numbers. This means it is pulling a signal out of the noise!!!! If you look at page 87 of the SX1276 spec, "Due to the nature of the LoRa modulation, it is possible to receive packets below the noise floor."  See page 26, for an explanation, and table showing negative SNRs.

    I like TI's high speed solutions, but they are behind in the sub-GHz ISM transceiver low-speed, long-distance solutions assuming Semtech's specs are true.  I would also say that TI could likely implement a similar solution to Semtech's, but with better specifications.  It's just a matter of taking the time to do it.  The CC1310 looks more competitive since it has an integrated CPU, but it's still not available.  It appears IOT support won out over making the transceiver specs better, but that's unfortunate since the power consumed by the transceiver outweighs the power consumed by the CPU. The SX1276 wins in the power consumption category vs the CC1125, but not against the pre-release specs for the CC1310.  Unfortunately, the CC1310 still can't compete with the SX1276 in terms of transceiver performance, but that's more palatable given the clear power consumption savings.

  • I would recommend to test the CC112x and the Lora system together in parallel and to perform a range test. We have performed this test, as well as several customers and the results have been the same every time. CC112x is far superior with respect to blocking and selectivity.

    With the IoT trend, blocking and selectivity are becoming more important since the number of wireless devices are increasing on a daily basis.

    Perform the range test and let me know your results.

    Regards,

       Richard

  • Hi Richard, thanks for the response.  I don't really have time to do this test, which is why I searched online hoping your applications engineers had already done so.  Until then, I will use the specifications.


    Thanks,

    David

  • Hi David,

    Recommend the following white paper as well for further reading: www.ti.com/.../swry006

    Regards,
    Richard
  • David,

    If you take a look at the white paper linked below (www.ti.com/.../swry006), you will find a detailed technical explanation of difference between coded (wideband) and non-coded (narrow band) signals. You do not gain sensitivity by adding coding gain, as you use the coding gain to combat the higher noise you get from the wide channel. The practical sensitivity is a function of the datarate, lower datarate, better sensitivity, longer range, pretty straight forward.

    The problem of using modulations like LoRa, is that you waste a lot of bandwidth. This becomes a big issue for capacity, escpecially in a long range network. Semtech knows this fact very well, but this is of course not part of their marketing pitch. Semtech have done system simulations of LoRa vs. narrowband and found that narrowband has 13000x better capacity compared to LoRa for a given bandwidth, you can find the simulations in the following public document, a Semtech proposal document from March 2015:

    http://www.3gpp.org/ftp/Meetings_3GPP_SYNC/GERAN/Docs/GP-150075.zip

    Using spectrally in-efficient modulation for the limited frequencies available for ISM band communication is just not a good idea, it does not scale

    Moreover, using wideband modulation for low rate, you are also very vulnerable to interference. We have performed apples to apples comparison of CC1120 and SX1276, using same frequency (868 MHz) and same datarate (1.2 kbps), running range test side by side. As you can see from the video below, the LoRa link totally breaks down in presence of a single interferer, even if the interferer is several 100 meters away. I expected it to be bad from the datasheet values, but  I was shocked by how bad it handled a real world environment. See for yourself and judge, would you like to deploy a system with this kind of issues?

  • For an RF system, it is really the range in a real life environment that matters. See for yourself how Richard Wallace and Rodney Farrow achieved 114 km (!) range using CC1120 devices. As you can understand from this test, it is not the sensitivity that limits the range, it is the ability to operate in a real RF environment - The CC112x family has the best co-existence properties of any ISM band devices on the market

  • Good technical discussion, however it all depends on the application. If I have a sensor that sends small amount data every 15 mins or hour, LoRa is better solution. It has better link budget and range since I can live with < 200bps. However, CC112x supports SigFox stack, which makes it competitive to LoRa.
  • Hi Terje,

    What was the configuration of the interfering singal; output power, bandwidth, modulation....
  • The interferer is a best case type of signal, CW (continuous wave / single tone / unmodulated), +10 dBm. Using a real signal, the difference between a robust narrow band system and the wideband LoRa system will be even bigger.

  • Was it a co-channel or adjacent channel interferer?
  • Co-channel is when the interferer has the same center frequency as the desired. Adjacent is when the interferer is offset from the desired.
  • The whole point here is to show the difference between wideband, coding gain solution vs. narrowband. The LoRa claim is that it handles well interference using coding gain - shown here is that it totally fails with a single tone as in-band interference. For the narrowband system, it is a neighbor channel, showing that coding cannot beat filtering
  • The above is correct for a narrowband signal, but not that accurate for a wideband, high coding gain system

    CC1120: 868.0 MHz, 25 kHz channel, 1.2 kbps
    LoRa: 868.0 MHz, 500 kHz channel, 1.2 kbps
    Interferer: 868.05 MHz, CW
    So for this demo, the interferer is in-band for the wideband, but neighbor channel for narrowband. The LoRa claim is that the coding will somehow handle this, which is shown here to be not correct.

    If you have a LoRa kit and 3 CC1120 kits, it is very easy to reproduce the result, you will be quite amazed by the massive loss of range for the LoRa kit
  • If you had recorded a little more detail, including transmit power, RSSI, and modulation type, this would have been an excellent comparison and may have disproved their specs. I have also had bad experience with coding gain which is why I am skeptical of their claims.
  • Guys, being a brave TI customer with tens of thousand CC112x in the field I need to put a finger into this wound.
    One year ahead of the start of this thread and having CC13xx introduced:
    are you TI guys still convinced that Spread Spectrum can't improve data trasnmission in certain situations?
    In CC13xx datsheet you reference e.g. to receive parameters with "2.4 kbps, using 19.2 ksps DSSS with spreading factor 4 and ½-rate FEC,".

    I understand the physical discussion A vs. B and I doubt that Semtech's device can deal with interfering. I also haven't found any convincing article explaining how Semtech can pull information from below the noise floor.
    But for a 'stupid few byte transfer' we did the test on our own (1120-DK against Semtech-DK at 433MHz with 1kBaud), finding that the 1276 with similar datarate performed better. Sure without interferer etc.,
    But this turns out it's not black or white.
    I doubt too that Microchip would ride that horse without having trust in certain LoRa abilities; their test at last Electronica performed with a lot of simultaneous devices.

    I hade some hope that TI's setting up something similar (better) in CC13xx and let customers chose what Layer1 communication the CC13xx should use in certain environments: A or B ....
    /GG

  • PS
    I'm for sure not trying to talk against narrow band, but the market and at least our customers will take this decision for us if we are not prepared.

  • I would also like to know this. This makes a big difference on if this is a real world test. LoRa can not tolerate a constant carrier wave tone inside its BW, but theoretically should be able to handle the short time duration packets of normal systems. I could totally jam a narrow band system too by broadcasting on its intended frequency constantly.
  • All, 

    Semtech LORA is not based on DSSS, it is based on a proprietary modulation that only Semtech implements.

    We do believe the DSSS technology has merits and our new CC13xx implements various long range modes based on forward error correction in conjunction with low order DSSS modes. Starting with the CC13xx we implement 2:1, 4:1 and 8:1 DSSS modes. We do not go higher than that due to poor return on investment, as you loose a lot of data rate for small gains in sensitivity.

    Regards,
    /TA

  • I think this test is not a valid comparison. With LoRa it is using spreading with 125khz BW so the interferer is probably positioned in the middle of the LoRa band while out of band for the desired TI signal conveniently for the purpose of this test. LoRa can demodulate 19.5dB below the interfering signal while if the interfering signal was placed in band for narrow-band solution it would need to be 8dB above the interfering signal to demodulate. This argument would only be valid if all the interference in the ISM band were narrow band but if you scan the ISM band most of the interference is wide wand from metering and other control applications using from 9.6kbps to 100kbps. With wide band interference LoRa performs far superior to narrow band.
  • Hi Terje,

    I think it's important that you describe the characteristics of the jamming signal. Thank You
  • The type of jamming signal is described in an earlier replay in this thread.
  • Has anyone gotten an answer as to what the characteristics of the jamming signal was? The test was performed with what power level and modulation? What was the power level and modulation of the jamming signal.
  • Hello All,

    What was the detailed settings of the Jammer. Frequency, power level, modulation type?

    How was the TI EVB set up? (that was able to be immune to the jamming.)

    How was the Semtech EVB set up that was jammed?

    Best Regards,

    Wilson

  • Jammer: See post April 30 from Terje: Interferer: 868.05 MHz, CW. I believe we used +14 dBm jammer
    The TI EM used the 1.2 kbps settings that are available in SmartRF Studio (1.2kbps, 4kHz deviation 2FSK). The Semtech used the default setting, selectable from theit menu system.