0

I am a bit new to this field and I was reviewing some speech enhancement papers. These two terms keep popping up every once in a while so I was very confused as to what the difference between these two exactly is. I know that SNR is defined as logarithm of signal/noise, but I haven't come across SDR and haven't been able to understand the differences between these two.

user297016
  • 11
  • 2
  • 2
    In every context I've ever seen, SNR stands for *signal* to noise ratio. – Hearth Oct 01 '21 at 14:24
  • SIGNAL-to-noise ratio and SIGNAL-to-distortion ratio – DKNguyen Oct 01 '21 at 14:36
  • @DKNguyen yes sorry i meant that only, that was a typo from my end – user297016 Oct 01 '21 at 14:55
  • 1
    I would think it is the same as THD, total harmonic distortion. – DKNguyen Oct 01 '21 at 14:57
  • @DKNguyen what is the difference between signal to noise and signal to distortion specifically? I am sorry but I am very new to this field so I might be asking very silly questions, I am still learning my way around it – user297016 Oct 01 '21 at 14:59
  • Signal to noise is just signal to noise. Distortion is when you inject a sinusoidal input into something and how many frequencies (and how strong relative to the original input sine) that aren't that original sine wave come out. Or, a real signal made of many sine waves via Fourier. It's how the frequency spectrum of the signal changes passing through: change in the relative strengths of pre-existing frequencies or brand new ones. An overall increase/decrease without change in relative strength would just be change in volume. – DKNguyen Oct 01 '21 at 15:00
  • @DKNguyen that makes sense to me thanks a lot! – user297016 Oct 01 '21 at 15:02

1 Answers1

1

In that case, you'll like another term, SINAD, Signal to Noise and Distortion.

Noise is unwanted signal correlated with no other signal. It's present whether the signal is there or not.

Distortion is unwanted signal correlated with the input signal. It's absent until the signal appears, and is usually not significant until the signal is occupying most of the dynamic range of the system.

I was aghast to learn recently that, when de-noising some recordings using Audacity, I discovered that 30dB was sufficiently good SNR during music passages that I couldn't tell the difference between that and perfect SNR. How my ears have aged!

Neil_UK
  • 158,152
  • 3
  • 173
  • 387
  • Damn this mamallian hearing that trades regeneration for higher , but ephemeral frequency response. I'd rather have a bird's hearing that only goes to 12kHz but stays that way. – DKNguyen Oct 01 '21 at 18:05