1

I am wondering about two scenarios that I came across for OFDM systems. Say we have a station transmitting to an access point and say we have the following two scenarios:

1- N1 subcarriers and transmit power per subcarrier is Pt1

2- N2 subcarriers and transmit power per subcarrier is Pt2 where Pt2< Pt1 but N1< N2

First question what would the relationship between Pt2, Pt1, N1 and N2 need to be such that the total baseband power consumption that is required by the station in the two cases would be the equivalent?

One other question from implementation perspective which would lead to a better system design? In other words which would make more sense.. does one provide more power consumption savings than the other or how does the 2 compare wrt PAPR, RF design...

Many Thanks as always.

Tyrone
  • 781
  • 1
  • 11
  • 23
  • Is the bandwidth per subcarrier kept the same (such that one case with more subcarriers has wider bandwidth overall with less power density)? Or is the occupied bandwidth the same so the bandwidth per subcarrier has to shrink, but then can you increase the power and modulation order to keep the data-rate the same? What are your actual constraints on all of these parameters so that we can limit the moving parts / combinations? – Dan Boschen Apr 21 '22 at 04:13
  • Thank you for good questions. Yes subcarrier bandwidth is same. So like you said in the first sentence such that one case with more subcarriers has wider bandwidth overall with less power density). – Tyrone Apr 21 '22 at 04:20
  • We can assume that we are trying to transmit the same equal transport block size on both these cases. – Tyrone Apr 21 '22 at 04:21
  • And perhaps in second case, mcs is lower – Tyrone Apr 21 '22 at 04:21

1 Answers1

3

Nice question and I think there will be some better answers to add to this but here are my immediate thoughts:

Let's assume we want to send the same amount of data, over the same range. The OP has clarified that one case has wider overall bandwidth, with lower power density (each sub-carrier for both cases has the same bin spacing), while the other with fewer sub-carriers will occupy less overall bandwidth with higher power density.

I will assume both cases have a large number of subcarriers overall, so therefore there will be no difference in PAPR. Given the central limit theorem, both cases will approach a Gaussian distribution if the number of subcarriers >> 10 and the data in each subcarrier is randomly assigned.

The obvious observation is one choice will be preferred when we care more about spectral efficiency than power efficiency (or vice versa). For the case with a larger number of subcarriers, we can transmit the same amount of data with smaller constellation sizes, which are more power efficient (less power per bit transmitted). When we go to fewer sub-carriers and want to maintain an overall data rate and range, then we must go to larger constellation sizes which are more spectrally efficient but less power efficient. This also adds cost/complexity to the overall transceiver design as a higher EVM must be maintained (so requires better clocks, timing, phase noise, linearity etc all adding cost and complexity).

So I think the consideration comes down to bandwidth available- if the bandwidth was there, the choice would always be for the wider band solution- smaller constellations can be used for the modulation in each subcarrier which are more power efficient and can get away with less accurate implementations in the RF hardware and overall transceiver design. Further, we can reduce overhead given more bins in terms of how many bins are used for pilots and the time slots for the cyclic prefix compared to the symbol duration (with more bins of the same bin spacing, the OFDM symbol duration will therefore increase).

As insight into the power vs bandwidth trade consider the following chart published in EETimes showing symbol error rate vs EbNo for the different modulation choices that could be used for the OFDM subcarriers. The ones with lower SNR per bit also transmit a lower number of bits per symbol (so less bandwidth efficient).

enter image description here

Dan Boschen
  • 50,942
  • 2
  • 57
  • 135
  • Hi Dan. This is incredible. Thank you very much. From baseband power consumption perspective do we think one case would have less power than the other? – Tyrone Apr 21 '22 at 04:36
  • Yes absolutely - Just look at a simple single carrier case and compare the EbNo for 1024QAM vs QPSK. BPSK and QPSK have the same power efficiency and then after that it only goes down. We win in bandwidth efficiency but at the cost of power. – Dan Boschen Apr 21 '22 at 04:39
  • I see. But wouldn't the overall transmit power by the UE in both cases be the same? – Tyrone Apr 21 '22 at 04:44
  • 1
    I don't see how: we will need more power to transmit the same number of bits since the power per bit is going up in order to get that within a smaller overall bandwidth. I added a chart to illustrate what I'm getting at in case that helps. Did I miss something? – Dan Boschen Apr 21 '22 at 04:46
  • @Tyrone Let me elaborate using the chart - let's assume we needed a 1e-5 SER, and we choose a system with enough sub-carriers allowing us to do that with QPSK. Here we need 10 dB per bit, and we send 2 bits in each subcarrier. Next we cut the number of subcarriers in half, to transmit in half the bandwidth. So to keep the same data rate we need to send 4 bits in each subcarrier, which pushes us to 16-QAM. Now we need 14 dB per bit, so the power penalty is 4 dB (More than twice the transmit power, which can really effect the PA and DC power dissipation needed on that end of the link). – Dan Boschen Apr 21 '22 at 04:53