Do PSK/QAM modulators (e.g. 4G/5G) fundamentally use more energy per packet than CSS modulators (e.g. LoRa)?

Question

Not sure if this is the right place or even the right question to ask, but googling did not get me much further, so here goes:

I'm trying to assess whether the power consumed by an IoT device, based on (future) 5G technology, will ever come close to the power used by a device based on (current) LoRa technology. For the purpose of this comparison, the use case would be a battery powered device that transmits one small packet per minute.

Experience tells me that there currently is a large difference between technologies. Peak current drawn and total energy used for sending a single packet with LoRa is almost an order of magnitude smaller than with NB-IoT, one of the 3GPP protocols that is promoted as "Ultra Low Power" and part of 5G (and currently part of 4G).

As any google search for power consumption on 4G/5G or LoRa will return mostly vague promises made by marketing people, I have tried to dive a little deeper. The fundamental difference between the technologies lies in the modulation scheme used: PSK/QAM vs CSS.

The way I understand it, CSS (Chirp spread spectrum) spreads the signal in time and frequency, in such a way that in demodulation the effective noise level is reduced significantly, allowing for communication below the noise floor. PSK/QAM requires a signal above the noise floor, plus some margin. This then should explain the difference in the observed power consumption.

As the benefits of CSS come at the cost of reduced spectral efficiency, it may not be a likely choice for the industry, given the high cost of licensed spectrum and the market demand for high datarates. Apart from that, integrating CSS-like modulation within the framework of 3GPP technologies may pose various other obstacles way beyond my limited knowledge. My assumption is that we will not see CSS(-like) modulation within 3GPP standards anytime soon.

So my question is: Do PSK/QAM modulators (e.g. 4G/5G) fundamentally use more power per packet than CSS modulators (e.g. LoRa)? And does that rule out that any 5G module in the foreseeable future will come close to the power consumption of LoRa modules? Am I missing relevant factors in my assessment of 5G/LoRa power consumption?

Answer 1

Yes PSK/QAM modulators fundamentally use more energy than LoRa chirp spread spectrum (specifically it is the transmit power required for transceivers using PSK/QAM modulators in order to close the link that is greater for a given data rate). This is a classic trade of bandwidth efficiency versus spectral efficiency, and is made clear from the "apples to apples" comparison of BER vs Eb/No under AWGN conditions for the various modulations. QAM is only favored when we wish spectrum efficiency at the expense of power efficiency. The choice of modulation specifically is a major factor in the resulting power / energy efficiency of the transceiver, and is typically constrained by a minimum required spectral efficiency.

See this related post where user "sugur_ros" references BER vs Eb/No for LoRa CSS, with his results as copied below:

Compare that to this plot specific to uncoded QAM:

The above plot is from "Analysis of the Probability of Bit Error Performance on Different Digital Modulation Techniques over AWGN Channel Using Matlab" by Diponkor Bala et al. JEEECCS, Vol 7, Issue 25, pages 9-18, 2021.

What these plots show is that (under additive white Gaussian noise conditions) to achieve a 10^{-4} bit rate with LoRa, a 5 dB SNR per bit transmitted is required. Comparing that to the (uncoded) results for 4 to 256 QAM and we see that 8 to 21 dB SNR per bit is required! Coding will improve this but no where near the achieved results with LoRa. What we get for the higher energy needed with QAM that isn't indicated in the above plots is spectral efficiency- we can transmit more bits in less bandwidth, and this is detailed further with the plot below.

This plot below from "Optical Fiber Telecommunications (Sixth Edition), 2014 by Masataka Nakazawa et al, and other plots like it, are useful in showing both metrics together with the Shannon limit where spectral efficiency would be high on the vertical axis and power efficiency would be low on the horizontal axis.

Further and significantly is the fact that the chirp spread spectrum waveform is also a constant envelope modulation (as AlexTP mentioned in the comments). This allows for the power amplifier in the transmitter to run much closer to saturated output power which leads to significant power efficiency.