EVM — Error vector magnitude — the transmitter's own noise floor
RF and system
Where it sits
What it is
What it is. Error vector magnitude is the distance between where a transmitted symbol should have landed and where it actually landed, as a fraction of the constellation's scale. It is the transmitter's contribution to the blur described above, and it is a hard requirement, not a design goal: TS 38.101-1 clause 6.4.2, Table 6.4.2.1-1.
| Modulation | Average EVM | In dB | Best SNR the transmitter can offer |
|---|---|---|---|
| π/2-BPSK | 30 % | −10.5 dB | 10.5 dB |
| QPSK | 17.5 % | −15.1 dB | 15.1 dB |
| 16QAM | 12.5 % | −18.1 dB | 18.1 dB |
| 64QAM | 8 % | −21.9 dB | 21.9 dB |
| 256QAM | 3.5 % | −29.1 dB | 29.1 dB |
Read the last column as a ceiling. A transmitter meeting exactly the 256QAM limit has an intrinsic signal-to-noise ratio of 29 dB. No amount of transmit power improves that — the error scales with the signal. So a device that only just passes the requirement can never quite deliver the performance the constellation implies, and the margin a vendor keeps here is a real product decision.
Note the shape of the numbers. From QPSK to 256QAM the EVM limit tightens by 14 dB while the constellation demands 19 dB — 3GPP does not require the transmitter to be perfect, only good enough that it is not the dominant impairment.
This is the hand-off to the front-end half of the folder. What it costs in back-off, linearity
and heat to hit 3.5 % is fe-components.md and fe-thermal-power.md, and this table is the
requirement those notes have to meet.
Read on
This concept was first written up in ref-modulation, which reads the whole group as one argument.
Before this concept, the hierarchy says to learn the following — the full chain, in order: