Semtech's LoRa (Long Range)

#41
Hi,
With all due respect to goeytex, reading datasheets only tell a part of the story. I suppose things were hammered into us so well in early 70s electronics classes
that its hard to visualize a receiver functioning without tuned circuits. Even the ft2000 yaesu I have which is supposed to use modern techniques, still is littered
with tuned circuits in the input stages, and some of the i.f. If you look at the cct of the input for some of these modern transceivers they use multiple varicap diode
tuned circuits. Yaesu even suggest that to improve reception, you can incorporate these micro tuned high q tuned circuits ahead of the radio.
Progress=obsolescence x confusion
john
I
 

srnet

Senior Member
#42
We are talking chalk and cheese really.

A circa £2000 comms receiver, taking up half a bench, has little in common with a single chip transceiver about 5mm square and costing around $1.
 

Goeytex

Senior Member
#43
We are talking chalk and cheese really.

A circa £2000 comms receiver, taking up half a bench, has little in common with a single chip transceiver about 5mm square and costing around $1.
Agreed,

A $15 Semtech SC127X Based RF Transceiver module is not intended to compete with or replace a FTDX-3000D costing > $2000 US. I wouldn't consider putting a 3000D in a weather balloon, or to monitor a humidity sensor in my greenhouse. Neither would I attempt ground to ground transcontinental RF coms with a SX1278.

But, interestingly, looking at the spec sheet, is seems that the Yeasu 3000D uses quite a bit of digital filtering in the IF stage.
 
#44
In a "cost no object" receiver design some things will be done digitally and some the old-fashioned analog way. The designers do whatever does the very best at each stage of the design. But the only thing analog that comes in at the price point of modern digital receiver chips is a crystal radio, and those are still far larger than a digital receiver (or even transceiver) chip, and with much worse performance. There is still room for analog electronics in radio gear, but the place it holds gets smaller every day.

Low noise receive pre-amps followed by a couple of tuned circuits is likely all the analog electronics we'll see in most receivers in another decade, and those only in high grade gear.
 
#45
Hi, I don't seem to be doing a very good job of explaining what I'm trying to find out. I'm keen to know how with such a simple circuit can all the problems that beset most amateur and commercial comms gear.
Even the mod equipment I've used had massive filters. Even satcomms had their own sets of problems.
I own 1 and 2000 transceivers, so cant comment on the 3000. Even the ft1000 I have, to get best results I had to install several crystal filters.
If these modules are so impressive, how is it that if I transmit a signal on 434.650 (Lincoln repeater input frequency) at about 1Watt, most of the car key receivers in our close
get swamped, and refuse to work. I'm trying to comprehend how these modules cope with this when airborne when swamping must be magnitudes higher.
I do know some of the new radios use wide band front ends, but from what I've experienced, their performance in strong signal areas are abysmal.
 

srnet

Senior Member
#46
Some tests on the South beach here at Tenby suggest the range reported by a UKHAS flight is no fluke.

Comparing the performance of the RFM22 as per used in $50SAT, indicate the LoRa modems have a 10 times distance improvement over the RFM22. Granted that the RFM22 is operating at 1kbps and the LoRa modem is at 60bps, but I recall that the RFM22 did not improve much at data rates below 1kbps.

So if your data transfer requirements are modest, and 60bps is enough, then the LoRa modules used with simple quarter wave wires and 10mW ought to give a LOS range of around 125kM, which is very impressive indeed.

Later I will see if adding a LNA improves reception, it did for the RFM22, around 12dB improvement or 4 times distance.
 

MFB

Senior Member
#47
Those results really are impressive. If the addition of an LNA does provide a range of about 500km, it will be possible to track typical up-down latex balloon flights from a single (hopefully, low cost) ground station.
 

srnet

Senior Member
#48
That LOS range was based on hilltop to hilltop testing, I would expect that LOS distance to a balloon on 1/4 waves to be around twice that.

Thats consistent with the reception distance of 394km, reported by a UKHAS listener, they were using a 6dB gain vertical co-linear, so with the 6dB approximately doubling distance, they would likely have got around 200km on 1/4waves.

I have said before that balloons don't move around the sky very quickly, so it ought to be simple enough to point a 10dB DIY yagi in the right direction, that would get you 3 times further again, and the real issue now is the height of the balloon and the radio horizon.

I dont have my LNA with me, but for that testing to be practical, I need to cut the TX power further still, I had it at 4dBm out then with a 20dB attenuator, so about 25uW output, and still I had to walk 1km there and back.
 

manuka

Senior Member
#50
Great stuff guys! I'm still in rural Vietnam but similar amazing LoRa performance is reported from a Kiwi mate I left Dorji modules with. Expect more from me in about 10 days.

Antenna-at low UHF bands classic Yagi are usually best for point to point work. Gains of 6-10dB are typical at each end.
Stan.
 

srnet

Senior Member
#51
What sort of range would you expect from a Moxon antenna without LNA? Very simple to make and portable.
Indeed it is, packs flat too.

I would guess the Moxon just a little bit behind the Vertical co-linear the UKHAS guy was using (for 394km) around 5dB gain versus 6-7dB gain, but accurate figures are kind of hard to find for the Moxon, might test it over the field one day.

By fitting a terminator to the antenna socket I found the radiation from the LoRa TX is cut dramatically and range is down to 30M or so, which will make LNA gain (and Moxon) very easy to work out. For instance if the distance with a 1/4wave is 50M, and the distance with a Moxon is 100M, the Moxon has a gain of 6dB, all other things being the same.
 

srnet

Senior Member
#52
I have the PICAXE TX side of the LoRa link working, the PICAXE TX is sending "Hello" and it being picked up by the Arduino driven RX.

I altered the Arduino code to print all the SPI register writes and reads and sent them to the serial terminal, so I could be sure what configuration was required.

Anyway, enough computers for one day, and its very noisy outside here in Tenby, lots of very tired looking people passing the window who seem to have been for a short swim, cycle and run, the Welsh Ironman maybe ?
 

srnet

Senior Member
#53
I now have the TX and RX side of the LoRa modem link working under PICAXE.

The program addresses the SX1278 SPI device directly. The code is very similar to that for an RFM22, the biggest difference is in the way in which the FIFO for RX and TX sides is addressed, although as for the RFM22 the FIFO is actually accessed via burst reads and writes.

The TX program sends the ASCII representation of a word variable which increases by one at every TX, the RX receives the count and displays it to terminal.

Its probably more appropriate to put the code under a separate thread in the Projects area, I will do that as soon I have tidied it up.

The provided Arduino code was written in such a way as to way over complicate the configuration for the various coding (error correction) factors, bandwidth and spreading factors. The setup initially looks very complex, but when you actually read the manual (silly me) it’s quite straightforward and easily handled in PICAXE basic with some appropriate symbol definitions and the actual calculations done by the compiler at program time.
 

srnet

Senior Member
#54
I had a chance to try the effect of a very low noise LNA (<0.4db noise, 41db gain) on the RX side.

Having read some of the material on the LoRa modem, I had a suspicion that the LNA may not make a lot of difference.

The distance\range improvement with the LNA was only about 50%, or 3dB. The RFM22 however, with its relatively insensitive receiver, showed a 12dB improvement in sensitivity with the LNA.

For a fuller picture of the benefits (or not) of an LNA with the LoRa device, I will need to set the data rate closer to 1kbps, so a direct comparison with the RFM22 can be made.
 

MFB

Senior Member
#55
Unfortunately, its a tradition amongst Arduino programmers to write code that that is both overly complicated and verbose. Just look at how many lines of code are required to set up the C++ complier and interfaces, before you even get to the main loop function. Not to mention having to install a library just to use I2C. Anyway, it all helps them feel a cut above mere Basic programmers.
 

Circuit

Senior Member
#56
Its probably more appropriate to put the code under a separate thread in the Projects area, I will do that as soon I have tidied it up.
This is something I shall really look forward to; such code can be most educative. A note on this thread to indicate when you have done so would be a helpful reminder to check out the Projects area.
 

Goeytex

Senior Member
#57
Unfortunately, its a tradition amongst Arduino programmers to write code that that is both overly complicated and verbose.
I see no such "tradition". I primarily use Picaxe BASIC for my projects, but sometimes I use Arduino "C" , and other versions of C when necessary. C is certainly different from Picaxe BASIC and may seem overly complicated to those who have not learned it or those that refuse to try. But to state that Arduino programmers have a "tradition" of writing code that is overcomplicated and verbose IMO is simply false.

Just look at how many lines of code are required to set up the C++ compiler and interfaces, before you even get to the main loop function.
This is the nature of C programming with Microcontrollers. The programmer has the responsibility (and flexibility) to set the processor type, configuration bits, and other things, where in Picaxe BASIC this is done for you in firmware. However the programmer loses the flexibility to set things up differently. The lines of code are there in Picaxe Basic you just can't see them and you can't change them.

Not to mention having to install a library just to use I2C
.
Having the option to include or not include certain libraries means the programmer has more control over what is complied and loaded into memory. For example, if the program does not need I2C coms then that library does not need to be loaded thus saving memory. You do not have this flexibility with Picaxe Basic or other interpreted Basics. The memory is used whether or not the function is needed. There are trade offs either way.

Anyway, it all helps them feel a cut above mere Basic programmers.
This is unnecessary ad hominem banter. How can you possibly know how any anyone feels? How can you with any intellectual honesty make such a statement? How does this kind of banter help anything at all?
 

MFB

Senior Member
#58
Of course there are trade-off's between compilers and assemblers. The definition of banter is "to assail with good-humoured raillery; to joke or jest" and that was exactly the spirit of my comments. Sorry if it hit a raw nerve.
 

manuka

Senior Member
#60
Bravo! I'm just back in NZ & intend dusting off the workroom's cobwebs to explore my SX127xx LoRa&#8482; Dorji modules. Stay tuned.

Extra: A casual smartphone Google while away spotted Semtech's AN1200.14 mention (below) that LoRa&#8482; is Doppler shift immune. I've yet to look at this in depth, but ponder it's an attribute of linear frequency modulated (LFM) pulse SS "chirping". Although valid for the likes of 2000 km/h aircraft,you may even have some mileage from this with your next LEO $50 sat? Yah! Nanotron's CSS pages make good reading too

Semtech&#8217;s proprietary modulation is a spread spectrum technique that uses wideband linear frequency modulated pulses to encode information, whose frequency increases or decreases over a certain amount of time.

As with other spread spectrum modulation techniques LoRa uses the entire channel bandwidth to broadcast a signal, making it robust to channel noise. In addition, because LoRa modulation uses a broad band of the spectrum it is also resistant to long term relative frequency error and multi-path fading and Doppler effects.
Stan.
 

Attachments

Last edited:

srnet

Senior Member
#61
There are two issues with Doppler, the first is the shift itself, the signal can be up to 10khz high as the satellite is coming towards you and 10khz low as its going away, remember $50SAT is moving along at a brisk 17,000mph.

The AFC in the RFM22B copes well enough with stable frequency shifts of up to 10khz and was fairly easy to test for on the ground, I just set TX and RX 10khz apart.

What I could not test for was the rapid change in frequency as the satellite approaches apogee, directly overhead. At the point the base frequency is actually changing at around 100hz/sec. It was envisaged that the PLL that locks onto the bit stream could have trouble coping with such a rapidly changing frequency. In practice however it’s not been a problem.

What will be most interesting is what happens at higher data rates, say 1kbps or more. The LoRa calculator does suggest an improvement over the RFM22B here, although for satellite operation it’s the combination of the RFM22B + LNA it has to beat. If the combination of LoRa device + LNA were to improve over the RFM22B combination by 6-10db, that would make a significant difference.
 

Circuit

Senior Member
#62
Preliminary program, TX only, to drive the LoRa device is here;
Excellent, thank you. There is so much to be learnt by studying how an expert in the particular field constructs such a program. I will look at it in detail; not necessarily to use it for a LoRa device, but to understand the structure of the program and simply learn.
 

srnet

Senior Member
#65
The attention that you put into the annotation and structuring of this program is much appreciated. High educational value, thank you.
For my benefit as well really, without the commnets its easy to loose track of what and why you have done, when you re-vist the code maybe months or years later.
 

srnet

Senior Member
#66
Well, I have to say the perforamance of these LoRa modules is really stunning.

I have the LoRa module working well under PICAXE and transmitting decending power packets, 17dBm to 5dBm. I setup a RFM22B to do the same. Both devices were transmitting the same payload of 12 bytes, at about 1kbps.

The idea behind the decending power packets is that if the transmitters (DRF1278 and RFM22B) are in the same place, and you swap the recieving antenna between the 2 types of reciever, you can measure the performance differance between the two devices. If with one setup you recieve packets OK at say 11dBm, but the other needs 17dBm, the performance differance between the two is 6dBm, and that can then be used to translate into a potential distance improvement.

For this test the DRF1278 and RFM22M transmitters were together at about chest hieght at a location 750M from my shed. I adjusted tha hieght of the antenna on my mast so that I was only getting the most powerful 17dBm packets from the RFM22B.

I then swapped across to the DRF1278 reciever, and I was also only getting the most powerful 17dBm packets.

This does not seem a big deal, apart from the fact that the antenna on the DRF1278 transmitter was fitted with a inline 20dB attenuator !

So the DRF1278 has a link budget gain of around 20dB over the RFM22, which equates approximately to a range differance of 10 times.

What further suprised me was that the testing was in a dense urban loction, and with the LoRa TX and RX antennas (simple 1/4 wave wires) at chest height, I was also able to recieve the 1kbps packets at 750M, when the effective TX power was a mere 0.5mW.
 
Last edited:

Goeytex

Senior Member
#67
Thanks for testing and sharing your information.

I will be receiving several of these LoRa front end modules in about a week. The plan is to write firmware that allows a Picaxe to send serial commands to the control the module. But unlike most other serial enabled modules, the firmware will allow the frequency and other settings to be changed without having to write to EEPROM.
 

srnet

Senior Member
#68
Thanks for testing and sharing your information.

I will be receiving several of these LoRa front end modules in about a week. The plan is to write firmware that allows a Picaxe to send serial commands to the control the module. But unlike most other serial enabled modules, the firmware will allow the frequency and other settings to be changed without having to write to EEPROM.
I can understand the atraction of the serial enabled modules, but driving the raw module is really not so hard, easier than the RFM22 for instance.

What do you mean by 'firmware' ?
 

srnet

Senior Member
#70
So the plan is to take the Dorji front end modules (which already have thier own firmware that allows them to be controlled via serial) and add some additional hardware and software that also allows the module to be controlled via serial ?

I am confused.
 

Goeytex

Senior Member
#71
How confused could you really be?

I plan to use a DRF1278F which DOES NOT support is RS232/TTL serial and add either a PIC or Picaxe MCU ( firmware) and produce something like the DRF1278DM.

I do not particularly like the firmware on the DRF1278MD because it forces a write to the MCU EEPROM every time a setting is changed. This makes frequency hopping and other on the fly changes impractical as the EEPROM would quickly wear out. So I will write the firmware so that certain on the fly changes can be made without having to write to EEPROM.

Clear now?
 

manuka

Senior Member
#72
Stuart: Many thanks for these MOST impressive "shootout" trial reports. NZ/Australian 433 MHz regs. allow us 25mW TX power, & with our more open spaces (& timber buildings) performance should be even better- especially in outback Australia. NZ's more hilly terrain & thick "bush" forests may well be an issue however for local environmental monitoring - yet to test decently. Check your email for possible collaboration thoughts!

Initial urban NZ trials have used Dorji's MCU version (pictured)- Chinese makers Appcon & HopeRF already offer similar. For most PICAXE type investigators, configuring such a MCU LoRa module should be easy enough (once EN is grounded!) via the likes of Dorji's software & DAC02 (or similar) USB-serial adapter. Stan.
 

Attachments

srnet

Senior Member
#73
I thought that a test in an Urban area was of more interest to the average user.

However since the comparison between the two devices was the same location\route and RX antenna, the comparison would be valid in relation to the known performance for the RFM22B.

For instance the pivotal test for the $50SAT project was a 40kM hilltop to hilltop test (100mW). If a DRF1278 had been used then you would expect an increase in range to 400km, which is really quite silly for such a simple and lowe cost module (I realise thats not entirely accurate as gaussian white noise becomes more significant over very long distances)

The decending power tests also made it easy to see how much differance an LNA makes (I used the outrageously expensive mast head Superamp SP7000), in the case of the DRF1278, the LNA added around a further 10dB of signal gain, and that equates to a distance gain of about 3 times.

I should aslo say that the tests were done at an RF friendly bandwidth of 20.8khz, with the coding rate and spreading factor adjusted to give a data rate that matched that of the RFM22.

The tests also sent out a tone and carrier marker at each power level, so I was able to confirm the DRF1278 output level was 20dB bellow that of the RFM22, its easy to check on the spectrum display of an SDR such as my Funcube dongle.

The same test software also sends out at decending power FSK RTTY, AFSK RTTY and FM Morse, I will run a recording later so you can hear what the Morse sounds like at the Limit of LoRa reception.
 
Last edited:
#75
Manuka: I was after the possible impact on neighbour's DTTV.
Personal experience with 10mW 434 had some 'nuisance' effects - but where I live the DTTV signal is slightly weak so was easily swamped.
 

manuka

Senior Member
#76
Dippy: OK - have purged post accordingly-wrong side of your morning coffee perhaps ?
Weak LoRa&#8482; interference with DTTV (Digital Terrestrial TeleVision) is likely to be small beer ?
However srnet's shootout (post #66) was in a dense urban environment!
 
#77
Thanks Stan. No coffee + long thread = missed things.
But, unless I missed it again, I see no comments on interference... 'small beer' during the final game at Wimbledon is very annoying.
(Note, in my previous post it turned out that I was the 'nuisance' ;))

It may all be happiness all round but some tests would be appreciated.
 

srnet

Senior Member
#79
For the $50SAT project I carried out a 40km hilltop to hilltop line of sight test with the RFM22B modules. The test was necessary to provide a baseline for the receiver&#8217;s performance.

For that test, using just 1\4 wave wire antennas, I got reliable communication at 1kbps with 20dBm (100mW), at the time that seemed impressive&#8230;&#8230;&#8230;...

I recently repeated the same test using the LoRa modules, also at 1kbps, 434Mhz

LoRa gave me reliable communication over 40km at 3dBm (2mW) also with simple 1\4 wave vertical antennas.

That would equate to 90km at 10mW which is the UK licence exempt limit.

If you dropped the data rate to 100bps, still a very useable rate for remote monitoring, you should be (according to the LoRa calculator) getting approximately 280km at 10mW. Proving that is going to need and airborne test, finding a ground based location with a LOS of 280km is not so easy.
 
Last edited:

manuka

Senior Member
#80
- finding a ground based location with a LOS of 280km is not so easy
Ha! Especially if you live in the near billiard table flat UK !

The earth's curvature indeed constrains LOS data comms., & my own on going 433 MHz LoRa™ trials (using our legal 25mW Tx power) show LOS is the limiting LoRa™ factor. Although we've lots of "hills" here in NZ (Mt. Cook - the highest- is ~4000m ASL), accessing them with suitable gear can be a real challenge.

However the onset of our southern summer holds promise for a ~250km volcanic link trial I've in mind. On clear days (& especially during eruptions!) 2 of these peaks are just visible from hills near where I live in Wellington. By chance my Picaxe mate Andrew lives near Mt. Taranaki & may be tempted to take a stroll up there for LoRa™ trials. In EU terms such a LOS span is equiv. to "seeing" Corsica's highest mountain (Monte Cinto ~2,706m / 8,878ft ASL) from mainland southern France.

FWIW Some years back I met a WiFi guru who'd managed a 382km 2.4GHz link using enormous dishes from the Sth American Andes far out into the Amazonian hinterland. This is/was particularly impressive, as he was probably NOT aided by VHF/UHF "tropo" (tropospheric ducting) that abounds in summer over oceans. Here in NZ we can receive powerful East Coast Australia ~100MHz FM stations under such conditions, & ~162 MHz AIS sea level marine tracking can extend 100s-1000 km with just a few TX watts.

Stan.
 

Attachments

Top