How to interpret a QAM display: part 3

Error vector magnitude characterizes actual QAM signals’ deviations from their ideal locations due to nonlinearity and phase noise.

In part 1 of this series, we discussed quadrature amplitude modulation (QAM), which results from summing orthogonal amplitude-modulated cosine and sine waves of the same frequency. In part 2, we looked at assigning bit sequences to QAM symbols, using 16-QAM as an example. We concluded part 2 with a diagram similar to Figure 1, showing that actual QAM signals often deviate from their ideal locations. Note that the symbols representing actual and ideal locations are points at the tips of vectors, as is shown for the symbol representing 0111.

What causes these deviations?

There are several possibilities. One is system nonlinearity caused by gain compression in a power amplifier. This nonlinearity can be expressed in the 1-dB compression point or the third-order intercept. Clock jitter or phase and magnitude errors in the original I/Q signals can also be contributors.

How do we quantify these deviations?

We use a parameter called error vector magnitude (EVM).

Ok, since it’s “magnitude,” we can ignore the phase and subtract the lengths of the two vectors, correct?

Not quite. Look at Figure 2, which shows a closeup of the Figure 1 symbol corresponding to the binary sequence 0111. Your approach gives us the magnitude error — the segment of the vector (red) that extends beyond the light blue arc that intersects the ideal symbol location. Because of the phase error, the error vector (orange) is slightly longer than the magnitude error. EVM is the error vector’s length. The M for magnitude in EVM indicates that the parameter includes no information about the error vector’s direction or phase angle, which is generally random and does not relate to system performance.

We’ve been looking at 16-QAM and higher, but is there a lower version — an 8-QAM or a 4-QAM?

Figure 3 shows four unique symbols representing two bits, each with Gray-scale coding. This scheme is sometimes called 4-QAM, but that’s a bit of a misnomer. There is no amplitude modulation — all the vectors are the same length, and the information resides in each vector’s phase angle. A more accurate term for this scheme is quadrature phase-shift keying (QPSK).

Since there is no amplitude modulation, EVM does not seem to apply to QPSK. Is that correct?

No, EVM remains relevant. Consider Figure 4, which shows three instances of the right half of the I/Q plane. The dashed blue circles represent decision thresholds, regions where a receiver will reliably decode the QPSK symbols. On the left side, EVM (length of the orange error-vector arrow) is nonzero, but the actual symbol (red) lies within the dashed blue circle, so the receiver will reliably decode the symbol to 11. In the center, the phase and magnitude errors place the actual symbol outside the blue circle, and the receiver wouldn’t reliably decode the binary sequence. On the right, the EVM is sufficiently large that it pushes a symbol that is supposed to represent 10 into the adjacent region corresponding to 11, resulting in a bit of error.

What are the units for EVM, and what instrument do I use to measure it?

EVM is typically expressed as an RMS percentage over a sequence of symbols, although it may also be expressed in decibels. The measuring instrument will depend on the device under test. For example, you might use a vector network analyzer for a power amplifier. You might choose a vector signal generator and vector signal analyzer for system-level measurements.

Where can I learn more?

See the resources listed below.

For further reading

How Error Vector Magnitude (EVM) Measurement Improves Your System-Level Performance (Analog Devices)
Making and Interpreting EVM Measurements (Keysight)
EVM: Why it Matters and how it’s Measured (LitePoint)
Do you know your EVM? (Rohde & Schwarz)

Related EE World content

How can I quantify a device’s nonlinearity? part 1
How can I quantify a device’s nonlinearity? part 3
What is phase noise, and how can I measure it? part 1
Measuring and abating jitter
QPSK modulation and generating signals
Analog and digital modulation and modulation measurements