nan value getting used in float reference code

Test case:

// OMASA 1Dir2TC 4ISM at br sw techs 13.2 to 512 kbps start 80 kbps, 48kHz in, 48kHz out, HOA3 out
../IVAS_cod -ism_masa 4 2 NULL testv/stvISM2.csv testv/stvISM3.csv testv/stvISM4.csv testv/stv1MASA2TC48c.met ../scripts/switchPaths/sw_13k2_512k_2fr_start_80k_omasatechs_4ism.bin 48 testv/stvOMASA_4ISM_1MASA2TC48c.wav bit
../IVAS_dec HOA3 48 bit testv/stvOMASA_4ISM_1MASA2TC48c.wav_HOA3_sw_48-48.tst

After first few frames, it is observed in file ivas_stereo_dft_enc_itd.c, for the statement "cohSNR = 20 * log10f( mEr );", mEr gets a negative value assigned due to which cohSNR becomes nan. Further, cohSNR is used for comparison.