Levels & Pop / click diagnosis?

First post & first attempt at getting my setup right. Two questions:

  1. I have set my levels to pass ACX noise floor and peak level, with RMS level about 3dB too low. My understanding is that if I can only get two requirements right without doing any processing, these are the ones. Am I correct in thinking that raising the RMS a little will require minimal processing?
  2. I’m hearing a clicking / popping sound that I haven’t heard in previous experimenting. Any ideas on how I might hunt down the source of this?
    I’m using Audacity 2.1.2 mono at 44100 Hz with a Blue Yeti usb microphone and a ghetto homemade pop filter and listening to clips using ATH-M50 headphones. My setup is in a walk-in closet with the monitor / keyboard inside and the desktop (HP Pavillion p7-1520t with Windows 10) outside:
    Here is an unprocessed sample clip:


Forget the homemade pop filter, the Hacksaw Jim Duggan 2x4 is the best!

Sounds to me like the CPU was struggling for resources when you were recording. Check what processes are running and see if you can disable any unnecessary things running in the background. You can also try plugging your microphone into a USB port on the back of the PC and not the front. Sometimes the front USB ports have an additional connector to the MB and that can introduce noise while the rear usb ports are mounted directly on the motherboard, so one less link in the chain. This goes for USB extension cables as well (since you have the PC outside the room).

Good luck!

Ditto. Also try increasing the CPU-priority of Audacity so it is less likely to be interrupted by other processes.

https://www.google.com/search?q=Set+“CPU priority”+“Windows 10”

Cool, thanks; I already had the USB plugged into the back of the computer, but I closed firefox and prioritized Audacity to real time, and that seems to have helped. Here’s a new clip after making those changes:

Any thoughts on the general sound quality? Other stuff I should address before I start recording in earnest?

My peak level, RMS level, and noisefloor are now at -2.7dB, -25.3dB, and -64.1dB respectively. Is there a good post somewhere explaining how to post-process to bring the RMS up a little and the peak level down a little (is this compression?). I’ve done some searching but am still a little lost on which settings to mess with.

Thanks again!

“Real-time” CPU priority is not recommended as that can cause the computer to become unresponsive / crash …


Increasing CPU priority by one level could be sufficient to cure the interruptions, (i.e. from “normal” to “Above normal”).

Thanks, makes sense; I’ll change the priority level.

Yes that’s dynamic range compression. A good free multi-band compressor plugin which works in Audacity is Gmulti

4solobo.xml (1.52 KB)
You should get Paul-L’s DeClicker plugin from here.

Thanks for the compression info. I’ve finally had time to give it a try. I want to make sure I understand the settings correctly so I can do this myself in the future. Is this correct?

Before compression:
-2.7dB peak, -25.3dB RMS, -64.1dB noise floor

After compression (using 4solobo settings file imported in Gmulti:
-2.6dB peak, -21.5dB RMS, -57.8dB noise floor

Gmulti Settings

  • 7.0 dB “In Gain” - applies a 7dB gain to the entire signal before doing anything else
  • 30 Hz “Low Cut” - high pass filter with 30Hz cutoff (i.e. get rid of frequencies lower than 30Hz)
  • 300 Hz and 2000 Hz Freq 1 and Freq 2 - splits the signal into three bands
  • 0dB / -33dB / -42 Thresh - any signal above XdB will have compression applied to it (why the difference for the different bands?)
  • 2:1 ratio - any signal above threshold will be compressed by this ratio (I’m a little fuzzy on what this means… my first instinct is that it’s a selective gain only applied to high-volume parts of the signal (i.e. reducing high-volume parts of the signal by a factor of 2, i.e. 3dB), but that doesn’t seem quite right… what’s going on here?)
  • 0.9ms attack (delay before applying the compression)
  • 10ms release (delay before stopping the compression)
  • 0dB / -6.0dB / +1.0dB level - applies an XdB gain to everything after compression (why the difference for the different bands?)

Thanks, and sorry for all the questions: my engineering instincts cry for understanding!

The Gmulti settings I used were what sounded good to me: so they’re not written-in-stone.

The dynamic-range isn’t constant across the whole spectrum, e.g. there is very little variation in the bass, (band1, 30-300Hz) , so that range didn’t get any dynamic-range-compression …
Not much volume-variation in bass, so no compression required_.gif
BTW Band3 functions as a De-esser … De-essing - Wikipedia

GMulti is “downward compression” … Dynamic range compression - Wikipedia

Those settings are another way of applying equalization : IMO the Band2 (300-2KHz) was too loud , hence -6dB gain.

Thanks! I checked out the wikipedia articles: I think I get it now.