2
\$\begingroup\$

When researching long range medium bitrate RF comms I came across the TI range. The TI CC1201 transceiver claims -97dBm sensitivity at 500-1000kbps. If the system is on the 868MHz band which is apparently free to use at transmit levels up to 25mW, a full system with two 2dBi antennas should have a max range of ~1.8km - based on the Silicon Labs calculator.

Will other losses in the system lead to much worse performance? (i.e. if the design between the transceiver and antenna is 'normal' with usual losses, not badly designed but realistic and compact)

EDIT: here's another potential frequency band, 863-865MHz with no duty cycle limits. The maximum power is only 10mW but this could still be viable? enter image description here

\$\endgroup\$
  • \$\begingroup\$ All I can say is that I used C1111 pre-assembled modules and they were very disappointing (Not to mention that TI does not seem to know how to implement a FIFO) \$\endgroup\$ – Oldfart Feb 20 at 17:23
  • 1
    \$\begingroup\$ Be very wary of silicon vendor claims. What they probably don't mention is that at 500kbps, your transmitter will be extremely broad, to the point where it won't even fit inside any license-free band in the known world. We're likely talking several MHz channel width here. And then mediocre -97dBm sensitivity, 10mW power, 500kbps and 1.8km range simply don't add up - I'd say that's complete hogwash and someone is trying to scam you. \$\endgroup\$ – Lundin Feb 21 at 10:43
  • 3
    \$\begingroup\$ Specific chip issues aside, note that the 868 band is a "crap band" where you'll find lots of interferences. If you squeeze in a broadband device there, you'll end up in a competition between yourself and everyone else over who will block out who. On top of that you get to fight the first harmonic of the 434 crap band too. I wouldn't even consider using anything that isn't narrowband. And whoops, there goes your baudrate. \$\endgroup\$ – Lundin Feb 21 at 10:47
  • 1
    \$\begingroup\$ @Lundin Could I reduce the wideness of the band with a band pass filter between the transceiver and antenna? Though perhaps I guess that could lead to impadance matching issues. Or is the increased transmit bandwidth just a symptom of high data rates here which can't be filtered out without the data itself being filtered out? \$\endgroup\$ – BenAdamson Feb 22 at 9:13
  • 1
    \$\begingroup\$ @BenAdamson Band pass filters are to block out noise, not the signal itself. The bandwidth does indeed come from the high baud rate, but of course also from the modulation. OOK and similar will be broad, while various GFSK will be narrower. I've got quite a lot of bad experience from sub-GHz TI and Silabs modules where you buy an evaluation board with some pre-programmed packet transmission, hook it up on your spec... and then wonder on which planet the silicon vendor lives on. As the broad signal output can't be used on any known band in the world. \$\endgroup\$ – Lundin Feb 22 at 9:57
6
\$\begingroup\$

Here is the outline of a link analysis:

we'll assume you want a low bit-error-rate (BER), maybe 1 in a Billion; for normal OOK on-off-keying transmission, and similar for BPSK, etc, you'll need a signal-noise-ratio SNR of about 20dB with the noise measured in the data-bandwidth (or for some maths, in data-BW/2).

Let's compute the room-temperature noise floor:

-174 dBm/Hertz [**]

+60 dB for 1MHz data (would be +50dB for 100,000Hertz data) [****]

+4 dB for antenna interfacing losses [antenna, balun and LNA losses/VSWR]

+5 dB for Low Noise Amplifier noise figure (perhaps overly cautious)

+20 dB for SignalNoiseRatio [for 1 error in 1 Billion bits; [***]]; this is conservative (12 or 13 dB may suffice, if you have perfect data recovery AKA no sampling-timing-error of a matched-filter)

000 dB for the benefit of error-correction (would be -15dB, for example)


-174 + 60 + 4 + 5 + 20 == -174 + 89 == -85dBm required OUT OF THE ANTENNA

Thus we may conclude your chipset either uses a much lower datarate, or has lots of bit-error-correction.

Now lets examine the effects of (non-multipath, non-tree-ground losses) pure energy-spreading-as-range_squared transmission:

Path Loss is 22dB + 10*log10[ (Distance/Wavelength)^2 ] dB

where the 22dB comes from hemispherical energy spreading factors; a 4*pi is buried in there.

For 3,000 meters TX-RX separation, and 1/3 meter wavelength, the Distance/Wavelength is 10,000X. Squared is 100Million; Log10 of that is 8; scaled by 10 ==> 80. Add on the 22, and the PathLoss is 102 dB. Note this is for 3,000 meters and exactly 900MHz.

The starting power is +25dBm. We need -85 dBm (and we included no margin for multipathing nor for foliage losses nor for operation inside buildings).

So what do we have? We have +25 - (-85) = 110dB to work with. The two antenna gains (2dB each, if properly installed, with zero-loss coax) boost this to 114.

Thus with 114 predict, and 102 due to energy-spreading-losses, this system has only 12 dB margin. At 3,000 meter distance. Which is about 6dB hit compared to 2,000 meters.

What if you want to use lower power or achieve longer range? What can you do? If you drop the frequency 10:1, the wavelength increase 10:1 and the Antennas increase in size by 10:1. Thus much bigger antennas reduce the Path Loss by 20dB. You can use +5dBm Transmitter power at 1/10th the frequency. Or retain +25dBm output power, but communicate 30,000 meters.

What if you drop datarate (by 30,000:1 to only 30 bits per second) and you drop the frequency (by 10:1, thus 10X larger antenna)? That +5dBm will drop, by 45dB, to -40dBm. Given 0dB is 0.632 volts Peakpeak across 50 ohms, the -40dBm is 0.632/100 or only 6.32 milliVolts PP, which greatly increases battery life or allows operation on solar cells and super-capacitors.

Thus at very low data rates and with large antennas, your sensors can cover large geographic areas. However, you need very accurate frequency control.

[**] read about "fluctuation dissipation" for the fundamentals of electron noise; -174dBm is KT at T=290degree Kelvin, where KT is 4.00e-21 watts/Hertz. To convert to dBm, we take 10*log10(XX) and simply add +30.

[***] search for "bit error rate for binary phase-shift keying"

[****] close, but depends on how the symbols are designed;

Another approach is to use the GPS method: emit 1MHz modulation, with a pseudo-random-number noise-whiting bandwidth-spreading underpinning as used by Qualcomm (and GPS). Simply flip the phase of the PRN transmission at your (much lower) data rate. You'll need a high-current LNA to tolerate possible blockers, and supportive correlation algorithms to recover the (low rate) data.

\$\endgroup\$
  • \$\begingroup\$ Thank you for your comprehensive answer - it's been a while since I studied RF/antenna theory so this was a really helpful refresher. I've put your formulae into an excel sheet so I can automatically calculate the effect of changing each value. I've found that if I'm transmitting on 865MHz at 500kbps at 10dBm output over 1km, I get a margin of 11dB (before multipath, obstacle, terrain etc losses). I noted you said "only 12dB margin" on your solution - I guess you'd need a much better margin to overcome these other losses? Or would the RX need much higher margin even after other losses? \$\endgroup\$ – BenAdamson Feb 21 at 11:24
  • 1
    \$\begingroup\$ @ BenAdamson The example I provided should be labeled "free space" losses, which I call "airpath". Placing TX and RX atop hills should come close to the numbers I gave you, unless the terrain between has irksome EM reflections. Note the world-record for communication for low-power is well under a milliWatt for reaching halfway around the world: read on "QRP" experiments by ARRL (ham-radio folk); that may have been over the oceans, and using ionosphere reflection (multi-path that helps). I've heard of multipath needing 30dB or more margin. \$\endgroup\$ – analogsystemsrf Feb 21 at 18:21
  • 1
    \$\begingroup\$ I've builtin lots of margin-params: lower bitrate to 10,000 bits per second, and get +20dB. Add error-correction, and get about +15dB. Lower the frequency by 10X (use larger antennas, which also changes multipath responses) and get +20dB. In recent years, MIMO multi-in-multi-out methods allow crude phased-array-antenna behavior, where energy nulls can be overcome by the Receiver, and TX energy given a bit of focusing for modest ERP boosts. Oh, forgot; NASA gets the photos from Pluto, quickly, by cooling the very first RF amplifier to about 10 degrees Kelvin; still only get 100 bits per second. \$\endgroup\$ – analogsystemsrf Feb 21 at 18:26
  • \$\begingroup\$ Where does 25dBm power come from? The OP said 25mW but then for the 868 band that implies duty cycle that kills 90% of the baud rate. More realistically, the band is limited to 10mW = 10dBm for continuous transmission. The power being E.R.P from the antenna. \$\endgroup\$ – Lundin Feb 22 at 10:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.