# MFSK vs FSK: Why BER decreases increasing the symbols

My question is related to the performance of 2-FSK modulation scheme against M-FSK (M > 2). From the SNR vs BER curve it seems that as the number of tones (frequencies) increases the BER decrease for the same SNR compared to 2-FSK(pics attached). (I assume maintaining the same symbol duration)

As the number of tones increases the bandwidth increases but i can't figure out why the performance of MFSK also increases. At first glance i would say that the resilience to the noise should be the same as the 2-FSK.

Which is the benefit of using 2-FSK instead of M-FSK (M>2) given a fixed available bandwidth?

Peter

Hence, with higher modulation order, you get less symbols per bit, and hence, for constant $E_b$ more $E_{sym}$.
Now, errors happen on a symbol level – and a Symbol Error Rate of $r$ doesn't imply a BER of $r$ – in fact, a wrong symbol has at most its worth in bits in wrong bits, but usually less (see: Gray code), so $\text{BER}\le\text{SER}$ generally holds, and one can in an uncorrelated noise scenario be almost certain that $\text{BER}<\text{SER}$.