"WMA has a richer, warmer sound"
2003-09-04 00:26:33
I especially find that WMA has a richer, warmer sound than MP3 (yes, even LAME APS MP3s), although WMA chokes on some of the high notes, (the sssss's slur and guitar solos can be watery). WMA has rich, full bass. That post in http://www.hydrogenaudio.org/forums/index....pic=12436&st=0& made me curious and as I don't want to add another off-topic post I start a new thread here. *** Mod's/Admins - feel free to move the other WMA-related posts here (and edit the title if necessary) *** 1st thing I've been curious about: Are rumours like "WMA encodes louder than original" (see Chrisgeleven's 1st post in the other thread) or similar which could be a reason for sound perceived as "richer" or "warmer" true? Test: I encoded a track ripped from CD (Omara Portuondo - Quizas, Quizas, Quizas) with CEP2.1 to WMA 9 128kbps CBR and WMA 9 pro VBR 75 (closest to 128kbps) and had a look at CEP's statistics and frequency analysis: Statistics:Uncompressed Left Right Min Sample Value: -32393 -32393 Max Sample Value: 32392 32392 Peak Amplitude: -.1 dB -.1 dB Possibly Clipped: 0 0 DC Offset: -.008 -.007 Minimum RMS Power: -82.9 dB -83.94 dB Maximum RMS Power: -7.93 dB -6.43 dB Average RMS Power: -16.89 dB -16.95 dB Total RMS Power: -15.98 dB -15.97 dB Actual Bit Depth: 16 Bits 16 Bits Using RMS Window of 50 ms WMA 9 CBR 128 CEP2.1 Left Right Min Sample Value: -32768 -32768 Max Sample Value: 32767 32767 Peak Amplitude: 0 dB 0 dB Possibly Clipped: 20 114 DC Offset: -.006 -.006 Minimum RMS Power: -85.96 dB -85.88 dB Maximum RMS Power: -7.95 dB -6.44 dB Average RMS Power: -16.88 dB -16.93 dB Total RMS Power: -15.97 dB -15.95 dB Actual Bit Depth: 16 Bits 16 Bits Using RMS Window of 50 ms WMA 9 pro VBR 75 CEP2.1 Left Right Min Sample Value: -32768 -32768 Max Sample Value: 32767 32767 Peak Amplitude: 0 dB 0 dB Possibly Clipped: 29 146 DC Offset: -.01 -.008 Minimum RMS Power: -82.55 dB -82.75 dB Maximum RMS Power: -7.95 dB -6.44 dB Average RMS Power: -16.87 dB -16.93 dB Total RMS Power: -15.96 dB -15.95 dB Actual Bit Depth: 16 Bits 16 Bits Using RMS Window of 50 ms Result : WMA isn't louder - it's slightly quieter (but the difference is way too small to be noticable) 2nd question: Does WMA "equalize" (i.e. change the relative volume of certain frequency bands)? Test: I took the files from above and Musepac 1.15r --standard --xlevel as reference and let CEP2.1 perform a frequency analysis. Then I copied the results to Exel and summarized the values to 1/2 octave bands: (CBR: WMA9 128kbps CBR; VBR: WMA9Pro VBR Q75) Band Original CBR VBR MPC (all in dB) - 00022 Hz -64.35 -61.60 -61.30 -63.82 - 00030 Hz -51.34 -51.20 -51.13 -51.35 - 00043 Hz -49.23 -49.14 -49.11 -49.23 - 00061 Hz -37.63 -37.62 -37.62 -37.63 - 00086 Hz -30.10 -30.10 -30.11 -30.10 - 00122 Hz -31.73 -31.72 -31.73 -31.73 - 00172 Hz -37.96 -37.95 -37.95 -37.96 - 00244 Hz -34.22 -34.22 -34.21 -34.22 - 00345 Hz -30.85 -30.84 -30.84 -30.85 - 00487 Hz -34.32 -34.31 -34.31 -34.32 - 00689 Hz -34.34 -34.33 -34.34 -34.34 - 00974 Hz -36.43 -36.42 -36.42 -36.43 - 01378 Hz -39.59 -39.57 -39.56 -39.58 - 01949 Hz -43.41 -43.33 -43.33 -43.40 - 02756 Hz -49.36 -49.25 -49.22 -49.34 - 03898 Hz -53.11 -52.97 -52.93 -53.10 - 05513 Hz -58.21 -58.00 -57.97 -58.20 - 07796 Hz -59.70 -59.48 -59.44 -59.71 - 11025 Hz -61.85 -61.60 -61.58 -61.83 - 15591 Hz -66.90 -66.57 -66.54 -66.87 - 22050 Hz -76.98 -102.45 -92.81 -95.46 [/font] Results/Conclusions: Between 50 and 1000 Hz the difference is >= 0.01dB for both WMA samples and MPC. Below 50 Hz WMA starts to become louder with arround 3dB difference for frequencies > 22Hz. MPC is only 0.53dB louder < 22Hz. Between 1378Hz and 15591 Hz WMA starts to become louder again with a maximum at the 11025-15591Hz band: CBR + 0.33dB, VBR + 0.36dB (MPC +0.03dB). The average for 1949-15591Hz band is: CBR + 0.21dB; VBR + 0.24dB; MPC + 0.01dB. Having a look at the highest band (15591-22050Hz) doesn't make much sense here because the codecs' lowpasses kick in here. From these measurements it seems like WMA amplifies very low and high frequencies. Frequencies < 22Hz can't be played back properly with most equipment and don't occur that frequently in music that they could lead to a general change in sound. About 2000Hz+ frequencies: AFAIK an overall volume change of 0.2-0.4dB can't be noticed/ABXed, but I have no idea about changing only a part of the spectrum. Anyone? I think most likely the richer/warmer sound is immagination, but as we know how WMA equalizes the music encoded, this could be simulated using FFT filters/equalizers and ABXed. Volunteers? edited: clarification, typos