Getting Noise Floor To -60 dB in Audacity

Hi Everyone,

I need to get my noise floor down to -60 dB in Audacity. So far it is hovering in the -40s.
The only way I have found to achieve -60 dB is by:

1. lowering the mic volume in Audacity to between 0.77 and 0.80,
2. setting the sensitivity on my Roland UA-25EX audio interface to 1:30 (due NE),
3. AND activating the bass cut filter switch on the AKG P200 mic that I use.

I am aware, however, that this is not the correct way to achieve a low noise floor.

In order to test if the source of the background noise is the acoustic treatment in my studio, I wrapped the mic in a towel and shut it in a clothes closet… and the noise floor remained in the -40s (without the bass cut filter activated).

In order to test if the source of the background noise is the mic and audio interface, I plugged in my Samson C03U digital mic… and the noise floor remained in the -40s (without the bass cut filter activated).

In both mic setups (AKG P200 & Samson C03U), the noise floor lowered to -60 dB only when I activated the mic’s bass cut filter.

In order to test if the source of the background noise is Audacity, I checked the same setups (AKG P200 & Samson C03U) with another recording software… and the noise floor remained in the -40s (without the bass cut filter activated).

In order to test if the source of the background noise is the laptop, I plugged in my Samson C03U digital mic into a secondary laptop (the Roland audio interface is not working in the secondary laptop yet)… and the noise floor remained in the -40s (without the bass cut filter activated).

What am I doing wrong? I appreciate any suggestions/observations.

(I’m Using Audacity 2.2.2 in a Windows 8.1 platform)

the noise floor lowered to -60 dB only when I activated the mic’s bass cut filter.

Many home recording systems generate tons of subaudible noise, based, I suppose, on the theory that it’s expensive to fix, nobody can hear it and in most cases, it doesn’t make any difference anyway.

It does if you’re trying to pass ACX. ACX apparently uses unlimited noise measurements (Z Weighing) which responds to acoustical energy in the performance whether you can hear it or not.

We had a significant fuss whether or not to include a High Pass Filter as part of the Audiobook Mastering Suite. We did on the theory that it did more good for technical compliance than voice problems it caused.

So even if you left the high pass filter off in your audio equipment, if you follow the mastering suite, we do the same thing in the first step.

“Low Rolloff” and “High Pass” are the same thing.

Having a rumble filter is perfectly valid. There are such filters on many sound mixers and interfaces. Most of the voices you hear in Hollywood movies have been through such a filter.

There’s a festival of different ways to miss noise conformance. The legacy mic preamp internal noise (Hiss. FFFFF) is joined by computer fan noise, lighting or other home appliance noise, sub audible “rumble noise” and the much more exotic Yeti Curse.

And it’s all made much worse by home style microphones delivering lower volume by design. Generally, only high volume overload and clipping distortion will cause you to send the microphone back.

Most trash is introduced when the sound is still analog. Once you digitise it into ones and zeros, it’s relatively immune to misbehavior.

Nobody wrote you can’t have more than one problem. A majority of home user failures are noise,

Post a sample of the work. We can give you a good idea of what’s gone wrong and how to fix it.

http://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Do Not Process Anything. Leave the filter off. Record it, export it and post it on the forum.

Koz

Hi Koz,

Thanks so much for your feedback. Please excuse my tardy reply, I thought my enquiry had not been approved as I had not received email notification of approval.

I have uploaded the entirely raw recording (with breaths, clicks, etc.) of my audition take for VoiceBunny (more on that below) for your evaluation, as instructed.

BTW, I had also tested with the overhead LED room light off and it made no difference. As well, much to my surprise, I found that the noisy (to my ear) refrigerator in the next room (which I customarily turn off while recording) did not increase the noise floor while running. And interestingly, my laptop’s fan noise does not seem to affect things (when I held the mic up against to the exhaust vent while the fan was definitely running, nothing changed).

I ran into the -60db noise floor barrier when I applied to do VOs for VoiceBunny. Whereas I had been pretty happy thus far with my recordings which were “noisy” and then darn quiet after running Audacity’s standard noise reduction effect (and my client’s were all happy with the products they were receiving), VoiceBunny lists the following requirements (among others) for the audition that is submitted during one’s application:

• Please submit only one take, edited and ready-to-use
• Please do NOT use any compression, noise reduction, or other kinds of post-production effects.
• All audio should be recorded with a “flat” EQ.
• Your noise floor should be at or below -60db.
• Please ensure your audio peaks at or BELOW -3db or it could sound distorted after uploading.

I think it is worth pointing out that VoiceBunny actively recognizes/supports the use of Audacity, given that they offer a tutorial, “Learn how to export WAV files with Audacity.” at two points along the application process.

I have not auditioned/applied to ACX but would suppose that the requirement specs are similar to those of VoiceBunny.

Now, all that having been said, if there is no other alternative than to keep the bass cut filter activated on my mic, that would not constitute “compression, noise reduction, or other kinds of post-production effects,” n’est-ce pas? Also, my voice sounds pretty much the same with or without it, as I don’t have a terribly deep voice.

I welcome any input/feedback.

Best regards,
Kevin

ACX Technical limits:
— Peaks no louder than -3dB.
— RMS (performance loudness) between -18dB and -23dB.
— Noise no louder than -60dB.

much to my surprise, I found that the noisy (to my ear) refrigerator in the next room (which I customarily turn off while recording) did not increase the noise floor while running.

How are you listening? Sometimes not hearing something is because your speakers or headphones can’t do it. I have a good quality music speaker system and I regularly hear problems that go right by most people. I have a silly joke that if you can hold your speakers in one hand, they’re probably not going to work particularly well.

Publishing companies regularly forbid the use of post production filtering. I suspect this is not so much because it’s a bad idea, it’s because most performers do it wrong. I should have saved it; one new reader arrived with a long list of effects and processing that they assumed were required to read for audiobooks.

No, not really.

On the other hand, if not for gentle equalization and noise reduction, there would only be four people reading audiobooks.

I subjected your clip to AudioBook Mastering and it passed ACX with no further work or noise reduction. First three sentences and declaration 2/3 down.


Only the first step, equalization did anything significant. Your raw clip is remarkably close. The clip has sub-audible noise which mostly vanishes with Low Rolloff equalization (click the graphics).

Screen Shot 2018-06-03 at 6.34.22.png600×512 33.3 KB

Screen Shot 2018-06-03 at 6.39.14.png597×537 30.5 KB

So I think you’re done. This is the writeup for Audiobook Mastering 4.

The equalization step suppresses rumble and earthquake noises, RMS Normalization step sets loudness and Limiter sets peak/overload. If you read in a quiet environment, you’re done.

ACX Check has one oddity that can cause you to have bad readings. It needs to have at least a half-second of Room Tone (shut up and freeze) somewhere in the performance to read properly. Room Tone is built into the ACX submission format and it may be there for Voice Bunny, too.

Does Voice Bunny have provision to submit a test? I haven’t checked ACX, but they used to make you go through much of the production before they would tell you the quality was trash.

Koz

What about De-Essing ? …

There’s a new de-esser plugin … Please help with de-sser. - #14 by steve

What about De-Essing ? …

That’s not technical conformance. That’s between you and Human Quality Control or the client, the step after the ACX Robot. Also see: removing lip ticks, wet mouth noises, etc.

If you find a De-Esser setting you like, you might see how close you can get mechanically by pulling a sock over the microphone.

One fuzzy goal is to see how few tricks you need to apply since you will be applying them forever.

Koz

A note:

Voice Bunny seems to be a Renta-Voice service. Contact them when you have a voice job, pick a talent, submit a script and write them a check. No mention of hiring on any new announcers, and they list the thousands of announcers they already have available for production.

Where do you fit into this?

Koz

Hi Koz,

Thanks again for your reply.

Actually, I can hear the background noise in my headphones when listening to the audacity project, as well as the WAV file that I submitted to the forum.

It makes sense to me that the background hiss would vanish with Low Rolloff equalization, as it would fall under the same category as the mic’s bass cut filter (although the former is post-production and the latter is pre-production in nature).

Cool that it passed ACX’s AudioBook Mastering, but I don’t understand how an audible hiss that shows in the -40dBs would register as under -60dB when submitted to the AudioBook Mastering.

Regarding that spike at 89Hz, the fridge was off and, as previously mentioned, the laptop fan surprisingly does not seem to contribute to the hiss (maybe I’m wrong there, but I had tested by placing the mic right up to the exhaust port).

Voice Bunny does not have a provision to submit a test. It just threw the aforementioned specs at me at the last click, just as I was about to submit the audition file.

How would I fit in the VoiceBunny realm? I would be one more among those thousands of existing VO talents ready to ‘hop to’ a gig opportunity. They allow you to audition for jobs that fit your profile so, in theory, the early bird might catch the worm. I have read pros and cons about their system. Also, there is no annual membership fee, so nothing to lose… We’ll see.

Yes, they are accepting new VO talents, whereas Upwork is not.

Thanks again!

Best, Kevin

I don’t understand how an audible hiss that shows in the -40dBs would register as under -60dB when submitted to the AudioBook Mastering.

It’s not hiss. It’s rumble, thumping, large truck going by, thunder, earthquakes. Somewhere around 30 Hz and to the left in that illustration is so low pitch you can’t hear at all.

Some microphone system make their own trash just by being alive. Like wrap the microphone in many blankets and it’s still thumping along.

The conspiracy theory has it manufacturers leave it there because nobody can hear it, it mostly doesn’t affect the sound and it’s not audible in podcasts or gaming commentary. Oh, and it’s expensive to fix.

All this is happening on the analog side, so the only way to be certain it’s gone is attack it on the digital side and that’s what Low Rolloff for Speech equalization does.

Both ACX and ACX Check measure “Z” noise which is basically everything whether you can hear it or not. This kills quite a few home recordists whose microphones fail Z noise the instant they come out of the box. Again, according to the conspiracy, this was on purpose to peel off the readers who would require expensive hand-holding. ACX is a business.

Passing ACX Conformance is handy because if you can do that, you can pass anything else. And, as further up the thread, once you pass basic ACX; performance quality, expression, theater, and rhythm is between you and the client. ACX has Human Quality Control which is where you go to die if your submission sounds like a bad cellphone recorded in a bathroom. They have a failure called “Overprocessing.”

Koz

There’s a new de-esser plugin

So de-esser.ny supersedes DeEsser.ny? Or not? Maybe?

Is the developer expected to continue to improve that particular software and run the release number up?

Koz

The new one is by Steve. It’s a lot quicker to apply than Paul-L’s if the problem is whistling at one frequency.
[ New one is single-band, whereas Paul-L’s is multi-band ].

Hi Koz,

Thanks for your reply.

From what your saying about rumble, it is generated, in this case, either by the microphone or vibrations radiating throughout the building (trucks passing by, etc.), in which case I suppose I should just keep the mic’s bass cut filter activated. C’est la vie.

I would like to ask: the Recording Meter shows from -60dB to 0dB. Is it possible to set it to show lower than -60dB? That would be helpful as I would then be able to know more precisely what my noise floor is. I was not able to find this addressed in any of the Audacity forum topics.

Best, Kevin

Is it possible to set it to show lower than -60dB?

Certainly. Preferences > Interface.

I set mine to -96dB and I pull the meters the whole width of the Audacity window (whose graphic won’t fit on the forum message format). If I do that, the critical area between -20dB and 0dB is the same graphic size as the default Audacity meters and it lets me see what’s going on in the sometimes troublesome area below -60db.

Koz

in this case, either by the microphone or vibrations radiating throughout the building (trucks passing by, etc.)

Or not. Some microphone systems don’t filter or process the 5 volts coming up the USB cable. They just jam the 5v onto the mic pre and condenser element in the case of condenser microphones. It’s cheap. So any low pitch trash made by the computer appears in your show (if you know where to look).

Interfaces that provide 48 volt phantom power are largely immune to this problem because they went to the trouble of making clean 48v from 5v anyway, why not just run the microphone electronics from that clean power?

Koz

You can do the analysis and figure out where it’s coming from. Without going back over this message, your microphone may have a shock mount available. That’s what this spider-looking thing does.

This white, rubber band thing is third party. It didn’t come with the microphone.

If your microphone is table mounted, you can use the heavy book and towel trick.

If you think it’s coming through the air, wrap the microphone in heavy towels and see if it changes. If it does, free the microphone from its mount and wave it around. Point it to different parts of the room. Directional microphones should give you some idea the direction of travel (Sometimes, when the whole house is moving, this doesn’t work. Those are seriously magic).

That’s how I found my music bass cabinet had two problems: It hummed/rumbled when it wasn’t being used, and it didn’t go off when I turned it off. I pulled the plug and all the problems vanished.

Koz

Forgot one. The wall brick that runs my sound mixer radiates electrical hum. Get the microphone closer than about a third meter or so and the sound goes mmmmmmmmmmm.

Koz

Wide and deep sound meters (posted through the magic of Photoshop).

Koz

Hi Koz,

Thank you for so much input!

I will follow your instructions.

Best, Kevin

Hi Koz,

Please forgive my absence…

Thanks for the pics of the shockmount. My mic is suspended inside one.

Regarding: “Some microphone systems don’t filter or process the 5 volts coming up the USB cable…. Interfaces that provide 48 volt phantom power are largely immune to this problem.”
The AKG P220 that I use is run through a ROLAND UA-25EX audio interface, which provides the mic with 48 volt phantom power. Interestingly, when I hook up my SAMSON C03U USB mic, I get the same background noise level. In both cases, the background noise drops to about -60dB when I activate the mic’s bass cut filter.

Regarding: “If you think it’s coming through the air, wrap the microphone in heavy towels and see if it changes.”
As mentioned previously, I wrapped the mic in towels and placed it in a clothes closet, and the background noise was not reduced at all.

However, I went ahead and waved the mic around the room as per your recommendation and:
I found that a ventilation duct in the back of the room (runs up and out through the roof - we live in an historic building and it has all kinds of interesting details) was a source of street noise.
So, I wrapped over this duct with cloths and a thick jacket and turned my studio recording system on as usual, but interestingly, the noise level was not any lower.
C’est la vie…

I would like to ask about debreathing:
Trebor posted about de-essing. My sibilance is not that much of an issue, but I spend a good amount of time removing breaths manually - whereas clicks are pretty easy and quick.
I have read articles that suggest that it is unnecessary to remove breaths, except for commercials, but I personally prefer to have them absent or really minimal.
Different tutorials suggest using limiters or reducing the volume during breaths (in order to minimize them), or copying ambient silence over the breath (which is what I currently do).
I have also seen some videos speaking about breathing from the diaphragm (which I already do as a singer) and “stealth breathing” or breathing silently.
Would like to know if you have a ‘take’ on this, as I could reduce my editing time considerably if I could get a better handle on this issue.
I downloaded a trial of Wave’s debreath, as I had seen it mentioned in a forum post, but could not make it work in Audacity. And, although I was able to install and try it in Reaper, I found it to be very insensitive.
I used to paste ambient “silence” into spaces where I wish to eliminate unwanted clicks, breaths, etc. between phrases, but I currently am using the “Amplify” effect, but specifying a negative value, such as -7.2, in order to reduce the breath in a selected area. Then, if it has not been reduced sufficiently, I simply use “Ctrl+R” (RepeatLastEffect) in order to reduce it further. This is much quicker and sensitive than copying ambient silence into the space AND I find that it is indeed nice to hear a slight breath sound at times (makes it sound more human).

Best,
Kevin

Would like to know if you have a ‘take’ on this, as I could reduce my editing time considerably if I could get a better handle on this issue.

I don’t read professionally, but on an informal poll of performers; leave breathing alone. My actual question was: Has anybody ever had a rejection because of normal human noises such as breathing, and does ACX forbid breath noises?

Zero people and no hits. We have had audiobook people rejected because of audible processing, patching, correcting, pasting and messing with the sound. What you do in one-off voice work is up to you, but audiobook production doesn’t mind breathing. One performer said it’s normal to build breathing into the theater of the show.

ACX is clear they want you telling a story over cups of tea, breathing and all.

Post Production is a bookkeeping problem. That’s one reason Audiobook Mastering is basically only three tools. Four with noise reduction. That was not easy. People run immediately to compressors and other processing, but those are much harder to use repeatedly than you think. Properly applied, Mastering 4 can’t be heard working and has no settings past the first application.

Also try and imagine getting popular. Too many readers build hard limits into their career. What happens when you get to read four books? You will not have time to remove each air intake from each sentence.

Koz