Typical values for RMS Time Constant, Decay Constant in Speech applications
Hello,
In several SigmaStudio Algorithms, time constant parameters like RMS TC, Decay are used to set attack / release / hold times for audio signal processing. 1) When dealing with human speech input, what are the typical values for these parameters?
Using modulated sinewaves as inputs, I am trying to test the functionality of dynamic processor algorithms. I am unable to see the desired effect in the output signal and I am guessing that the time constants in my DSP Project are not set right.
2) What does a RMS Time Constant of 10 dB/s mean?
3) What does a Hold Time of 10 ms mean?
4) In Compression / Limiter / other dynamic processor algorithms, is 0 dB equivalent to 1VRMS (1.4Vpp)?
5) What is the difference between No Slew (Standard) and Clickless SW Slew Algorithms?
Thanks,
Shyam.
rms
Audio Signal Processors
tc
sigmastudio
SigmaDSP Audio Processors
adau1761
Show More
2) This means that if you set a limiting curve at -10 dB and input a signal at 0 dB, it will take 1 second (approximately) for the output signal to be compressed to -10 dB.
More information about the concept of RMS time constants is here: https://ez.analog.com/message/20622#20622 and here https://ez.analog.com/message/14930#14930
3) This means that when compression is occuring, after the input signal becomes lower than the compression threshold, the output signal's gain will not start to rise until the hold time of 10 ms has expired.
4) No. Signals inside the DSP is in the digital domain. In other words, all signals you see in SigmaStudio are referenced to 0 dBFS. In order to figure out how this relates to the analog world, you'll have to check the datasheet to see what the full-scale analog ranges are on the ADC and DAC - usually they will be different.
5) Non-slew algorithms change the gain instantaneously. This saves instructions, but can lead to zipper noise, pops, and clicks when the gain is suddenly changed. Slew algorithms slowly change the gain so that there are no audio artifacts on the output, but require more instructions.
1) Typically you'll want the attack to be quite quick for voice, only a few tens of milliseconds. The decay (release) should be longer, like 1 second or more. You can use the equations I gave in the other post to calculate suitable values. However, like anything in audio, it's usually much better to use your ears and tune the parameters rather than relying on theory and calculations.
Please log in to reply.