According to the documentation,
"r = snr(x,y) returns the signal-to-noise ratio (SNR) in decibels of a signal, x, by computing the ratio of its summed squared magnitude to that of the noise, y. y must have the same dimensions as x. Use this form when the input signal is not necessarily sinusoidal and you have an estimate of the noise."
In other words,
r can be negative because it's measured in dB, so if the sum of the squares of the noise y exceeds the sum of the squares of the signal x, the argument of the logarithm is between 0 and 1, so the logarithm is negative.
