MDCT excels for coding but lacks robustness to time-frequency domain edits

Evidence

The modified discrete cosine transform (MDCT) is a perfect reconstruction, real-valued, and critically sampled transform. The MDCT is widely applied in audio coding due to its nonredundancy and its property of representing narrowband signals with a relatively small number of prominent spectral coefficients (Geiger et al., 2001; Neuendorf et al., 2013). However, the MDCT is not designed to be robust for considerable signal adjustments in the time–frequency domain. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 2115

Categorization

• DEFINE: Key distinction between a coding transform (MDCT, in MP3/AAC) and a processing transform (STFT).
• XREF: Directly connects to why spectral editing/denoising tools prefer STFT — the standard practical choice across audio restoration literature.
• ADDENDUM: Supports a follow-up angle on which transforms suit which audio tasks, usable with the grafting/trees infrastructure piece’s method of matching tool to purpose.