Room Noise Floor -59.6DB Argh! [SOLVED]

Hi all,

I just recorded all my files for ACX and ran ACX-check on the raw Room Noise and found that it was -59.6DB :astonished: just outside of Audacity’s requirements. Please, how do I fix this? I’ve seen so much conflicting advice all over the internet about using noise reduction or not.

What tool do I use in Audacity, at what settings, and do I run it over the whole track or just the room noise?

Also, does anyone have a guide to actually use the tools to get the rest of the track configured? On ACX’s video tutorial, it says to use a normaliser, EQ and limiter. While I have found all these tools in Audacity and looked at the Audacity manual, I’m afraid it all looks like Gobbledeygook. I have no idea of actually how to use them to sort out my audio.

Many thanks if anyone can help. I’d be really grateful.

Did you meet the other two, peak and RMS? Do you know what the specs mean?

I posted a process to take you from raw recording to finished ACX-Compliant submission, but if you’re close with what you have, you may only need a gentle push from Noise Reduction. Generally, the less processing, the better.

Drag-Select a portion of Room Tone only. No gasps, no breathing, no mouth ticks, no other sounds. Effect > Noise Reduction > Profile.

Select the whole chapter by clicking just above MUTE. Effect > Noise Reduction: 6, 6, 6 > OK.

Analyze > ACX-Check.

I was clear in the first sentence of the comments that I understand a dividing line between technical conformance and theatrical presentation. Mouth ticks, asthmatic breathing, and harsh SSS sounds will not prevent you from passing the ACX robot, but may very well fail Human Quality Control. Why would anybody pay to listen to vocal presentation that sounds like sandpaper on the ears?

The goal, expressed in different form by ACX, is listening fascinated to somebody telling you a story over cups of tea.

Since your question was about noise floor, I think we’re done, but if you want further analysis and comment, post back.


Thank you so much for helping. I only checked the raw Room tone, because I wanted to know if it passed before I fiddled with anything else. I’m just trying the noise removal process you posted. Do I choose Reduce or Residue in the final step? Sorry, I seem to forgotten everything I used to know about Audacity!

D’oh sorry, figured that one out on my own. Thank you - the noise reduction now has my floor at -65.7db :smiley:

I just checked the whole track. Noise floor is now a Pass, Peak levels are a Pass. Only problem now is RMS level is -30.2db, less than the -23db required.

In that case you still have noise problems. When you amplify the track to get the RMS up to the threshold (7-8 db) your noise floor is going to be back up to ~-57db.

Can you post an unprocessed sample for us? 2-3 seconds of just room tone followed by 2-3 seconds of your normal speech. Post as a wav file. That will allow us to get a better idea as to the source of your noise problems and recommend more targeted solutions.

ACX-Check automatically looks for a “dead” spot in your dialog to measure noise, so you don’t have to find one manually. It does have trouble if you didn’t leave any silent parts for it to measure. Leave about a second of dead air at the end when you read. I believe ACX has ideas about this as well.

So you can always measure the whole performance.

I need to drop for a while.


Behold, the human [flynwill] who developed ACX-Check.

This is the formal version of a test clip.


There are three different versions of “noise.”

There’s microphone or microphone system hiss. All systems have this and it’s your job to choose the system appropriate to your performance…and purse.

There are microphones that allow computer electrical noises to enter the show. Those are pretty evil and the noise is difficult to remove.

And then there are earthquakes. Some microphones make noises that are so low pitched (rumble, thunder) that you can’t hear them, but they mess up noise measurements for no apparent reason.

We can tell which you have when you post. It’s important to post a raw, uncorrected performance. Don’t help and don’t apply effects or filters.


OK so I have checked all my files after noise reduction. None of them pass DMS. Only 2 of them don’t pass peak:

Dedication -35.0DB
Contents -33.0DB
About -30.2DB.
Part 1 -30.8DB
Part 2 -27.6DB
Part 3 -27.2DB. Peak is also off at -2,2DB
Part 4 -27.0DB. Peak is also off at -2.4DB
Glossary -28.0DB
Opening Credits -28.6DB
Foreword -29.9DB
Introduction -29.4DB
Closing Credits -28.4DB
Summary -25.3

I’ve put a WAV file sample of Dedication, since that’s the furthest off RMS and of Part 4 as that is off on both Peak and RMS:

Sorry I couldn’t post the non-noise-reduced files. I saved copies of all my raw files (as aup) but when I try to open them up in audacity now, a message comes up saying there are missing audio data (.au) blockfiles, which can’t be recovered!

The only unadulterated file is my Room Tone, with no noise reduction or anything, which is below:

Thank you for all your help!

Sorry I couldn’t post the non-noise-reduced files.

That’s not useful because the best we can do is pile our custom corrections on top of the corrections you already have. The possibility of sound damage goes way up. One of ACX’s failures is “Overprocessing.”

The goal is a few corrections as possible.

Burn us a fresh test clip according to that recipe.

Did all of your Audacity Projects (AUP/_DATA) fail? That’s super serious because we’re going to give you corrections for the raw performance.

It is recommended that you Export a WAV (Microsoft), 16-bit of each chapter as an archive. That will, for the most part, prevent you from ever having to re-read any work. I know that’s what you thought you were doing. Projects can have problems.


OK, I have a test clip with my original setup I used to record the original files here:

It fails Peak and Noisefloor but passes RMS.

And one with the gain changed to give me a lower noise floor (not my original settings):

This Passes peak and Noisefloor but fails RMS.

All other settings are the same. I’m using a Blue Yeti condenser in a small storeroom filled with foam and boxes and a thin blanket over the mic to stop plosives etc. I’ve moved my laptop as far from the mic as possible as it does have a fan. I’m speaking fairly loudly into the mic and it is close to me.

Yes, none of the aup copies will open :frowning:

none of the aup copies will open

Let’s do that one first. That’s pretty serious.

This is an Audacity Project.

They come as a set, you can’t separate them and it’s a bad idea to move them once Audacity places them somewhere. Did you “clean up” or decide to “organize” your work?


OK, this is your clip.
Screen Shot 2017-03-25 at 14.43.42.png
Your microphone produces a lot of rumble and I sheared most of it off with a special, custom tool. You know when a thunderstorm goes through and the windows start moving/rattling before you can actually hear the boom? Your microphone is making that kind of sound. No, you can’t hear most of it.

Then I ran a custom tool that forces the RMS (Loudness) to be correct. That usually throws peak off, so I ran Effect > Limiter to bring the peaks back into compliance. SetRMS is such a welcome relief. We never used to have tools that directly affected RMS (loudness), so we had to use work-arounds and odd tools.

That suite of corrections will pass all by itself, but the noise is right on the edge, so I ran Effect > Noise Reduction to produce a plesant-sounding clip that easily passes ACX.

But before you get all excited, I can’t easily hear all the details in the sound. I’m working from the field and it’s noisy here. But from what I’ve been able to see, it should be good to go.

There are three custom downloads associated with this. You already have ACX-Check. You will need SetRMS and LF-Rolloff for Speech.

I am basically following this posting.

Everything is custom, so the instructions go forever. You in particular will be applying LF-Rolloff and Noise Reduction out of order from those instructions. You get rid of the thunder first so Noise Reduction has less work to do.

I need to drop out.


Firstly, thank you so much for your help. I will download the plugins you mentioned.

No, I didn’t reorganise any files. What I did was save all the tracks as aup projects, then copy and paste them all next to the originals, so they should have been exact copies.

I still have the original folders with the data files in them. They are unedited. Is there some way of reconstructing my tracks from those folders? They each have several files in them, named things like I seem to recall it might be possible.

I am eternally grateful for your help so far. Thank you, Koz.

What I did was save all the tracks as aup projects, then copy and paste them all in the same folder as the originals, so they should have been exact copies.

They shouldn’t be moved like that. Can you put everything back the way it was?

I seem to recall it might be possible.

Let’s go with no. Audacity Projects are robustly resistant to Disaster Recovery. You should make every effort to put all your copies and movements back the way they were.


I’m back home (quietly) and you do have an odd tiny whistle in the background of your clip. I was able to get rid of it with yet another custom tool, but it’s not a “normal” failing and I strongly suspect that’s the cooling fan in your computer.


The voice will never pass Human Quality Control like that. You are about to experience a shortcoming of USB microphones. You can’t ever get more than one USB away from the computer.

I’m trying to remember if I ever had to juggle a problem like this. I’ve never used USB microphones to do anything serious.


It may be possible to rescue your recordings with some combination of LF rolloff filter, noise reduction and a bit of compression. I’ll leave it to Koz to develop the recipe, he’s really good at that.

But before you do any additional recording you need to track down and fix that 2 kHz whine. Even when below the specified -60 dB noise floor, it it very apparent and may well cause your recordings to be rejected. This is not the usual USB digital noise (which has a fundamental frequency of 1 kHz and has a lot more harmonic content). Tell us a bit about your setup, what sort of microphone, preamp, mixer, digital interface, computer, etc do you have?

One thing you might try, is if you are currently recording with a 44.1 kHz sample rate, try switching to 48kHz. (Or if you are currently using 48 try 44.1)

I know, right? You bought the Yeti because it was easy and convenient to produce AudioBooks at home, make a mint and retire to a cottage on Majorca.

I produced a clip with the background noise boosted briefly. Nobody’s worried about the rushing water sound, but that eeeee whine is a problem. Is that your computer?

There are physical ways to help with that if that’s what it is.



I didn’t move the files. I had everything in a Library, and just did a copy and paste of the original Trackexample.aup, pasting it in next to the original in the library, so I had Trackexample.aup and Trackexample - copy.aup. The copies didn’t create any extra data folders or files. The original ones are still there.

I have no idea if running noise compression will have altered those data files or not. I’m assuming not, as I still have folders from where I spliced together two tracks to make one - so I’m assuming they are unedited, raw data.

As for the Yeti, I used to make videos and it was good for those as my video camera microphone is weak, and it worked well to provide a clear voiceover for my videos - good enough for you tube anyway. all my mics are USB as my computer does not have any other type of microphone port. I’m just using the Blue Yeti and an HP laptop to record, nothing else.

I do live in London, but my room was fairly well padded. When I was in the room all I could hear was the laptop fan. I did place it as far away as possible, and turn the mic away from it, but I assume that may have been the whine as the fan is rather noisy. Is there any way to take the whine out, or will I have to re-record anyway? I’m not sure what I could do, other than put the computer out of the storeroom, and pull the door as closed as possible.

Thanks for all your patience and help.

OK I have moved the computer out of the room.

Here is the sample:

I now meet:
Noise floor -62.8
and peak -5.4
But RMS is too low -30.4.

Would something recorded with this set-up be salvageable??


OK I ran the following on my sample with the new set-up:

EQ: LF Rolloff filter around 5000 (5015)

Run Nyquist Prompt – OK (SetRMS: Target RMS -20.00, Linked Stereo > OK)

Limiter: Soft Limit, 0.00, 0.00, -3.50, 10.00, No > OK

Apply Noise Reduction
— Drag-select Room Tone, silence or the flat area between spoken phrases.
— Effect > Noise Reduction: Profile
— Select the whole clip or show by clicking just above MUTE.
— Effect > Noise Reduction: Settings 6, 6, 6 > OK

I now have this, which passes all ACX checks and no whistle in the background (THANK YOU KOZ et. al :smiley: )

My voice still sounds a little sharp. Any final tips to make it perfect? I am so grateful for all your help.

Would something recorded with this set-up be salvageable??

Yes, but you also have The Yeti Curse. We have a pre-baked filter for that.
Screen Shot 2017-03-26 at 11.17.52 AM.png
Let me make a fresh cuppa and catch up.