Wanted: mastering chain for meeting ACX Guidelines

What else is new? Another acx narrator knowing nothing about sound, trying to “master” for acx.
I was hoping to find the “magic path” of post-prod effects, to pound every recording through, and voila!, but I am learning it doesn’t work like that…
Help?!!

I have learned many things, mostly here, props to these topics:
viewtopic.php?f=64&t=80572#p252057
viewtopic.php?f=28&t=78561#p242164
and others…

The ubquituous Qs:

Using: Audacity 2.0.6 on Mac OsX 10.6.8

Equipment: Tascam DR-05, pop filter

Sample raw (untouched) clip. As imported off the Tascam.

Starting point.png
What I’ve been doing:
Sampling the long quiet space I leave at the beginning for noise removal and noise removing whole recording.
Editing out my faux pas
Saving as a “raw” project before I mess around
Now the dubious stuff-
Compress 3:1, -24 threshold, 10ms attack, 80ms release, 0 gain, noise floor -60
Apple: AUFilter (stripping out highs and lows? hopefully) low at 100hZ, high at 10 000
Steve’s Limiter, to -3dB, from here
Normalize, max amp to -3dB

After Limiter before normalize.png
“After” clip - would it pass? Does it sound right?

Commentary:
I’m happy with the room - I don’t notice any trouble with humm or hiss.
I think it sounds “ok”, but I wouldn’t know. Something seems to go off when I put it through my chain even though the compression at least seems desirable.
My main Q, I think, is how do I achieve the holy -23dB and -18dB RMS (wtf is RMS anyways, or do I need to know?)? How do I know I’ve got it? When I Analyse Contrast on various recordings I get different results between -16.2 and -21. Obvs, I don’t know how to manipulate this.
I’m using in-device noise removal on recording, to 80hZ. The Tascam records on a microSD and then I transfer to comp. My recording level is 69 now and I speak about a foot from it. Recording in mono, 44.1- that’s set on device.
My earlier reading here prompted me to turn up the levels on my mic/recorder, and then not speak so close to the filter, because the meter was running lower than the little “optimum” tick on the display. Maybe that was a mistake - now it seems I hit the “top” (0dB, I think, learned -bad) up to 3x a recording. But then, sometimes not. Depends on the “theatrical-ness” of the excerpt, too, I’ve noticed.
I tried not using the noise removal and then I fancied that some low fuzz came out of nowhere at some step, and also my inhales got amplified into obese caterpillars of wave forms, like gasps before I speak. Not ok, I went back to noise removal.
Should I be doing LowPass instead of the AUFilter (I thought, high and low togehter, boom)?
NoiseGate? Amplify?

Thank you in advance!

If you’ve been following along in your hymnal, you know that first pass is a ten second WAV file, not MP3. Post it raw, no filters, no effects.

Record a 10 second mono test in your normal announcing style with the red sound meters peaking about -6. The first two seconds should be holding your breath and not moving at all (noise test). Export as WAV. Post it here.

https://forum.audacityteam.org/t/how-to-post-an-audio-sample/29851/1

Do not apply any filters or effects. We need a raw clip.


\

Recording in Audacity
2015-02-09

This is one recommended setup for live recording in Audacity.

Undock the meters and make them enormously bigger. Click the ribbed control strip on the meter left edge and the control corner in the lower right. Make sure the meters are set for about 60 on the left. I think it will actually read -57 or so. That’s the right sound range.

Change the range if needed in Audacity > Edit > Preferences > Interface: Meter dB range: -60dB…

Then, while you’re performing, make sure your bouncing red sound meter regularly peaks around -6 and never goes all the way up to 0. Note my blue waves are generally the same size and don’t wander up and down by very much.

Yes, you do have to watch the meters and read copy at the same time. The board operator on the other side of the glass in a real studio is adjusting levels and watching the meters as you perform. Now you have to do that. You get used to not being wildly theatrical while you read.

If you can get a stable live recording, then chances are you will need very little help to get to ACX compliance. If not, we’ll try to tell you what to do to improve. If you cover three chapters with “Mastering” filters and effects, you’re probably doing something wrong. That and it’s exhausting going through all that for every performance.

Let us know.

Koz
LiveRecordingLevels-650.jpg

And yes, I did shoot a simple sound test like that, and with two very simple, gentle filters fell into ACX compliance… and didn’t sound too bad, either. Oddly, I don’t remember where I left that clip. I need to search for it. It should have been in my collection.

RMS is Root Mean Square. It means roughly overall loudness. You have to meet three specs. The blue wave peaks should be close to but not over -3, the holding-your-breath noise should be -60 or below, and the overall loudness should be between -18 and -23.

People doing very casual recording in hostile rooms never meet noise and loudness at the same time. Those are the performers that drill themselves into the mud with lists of effects and tools.

Koz

Don’t go too crazy at the top. Nobody is expecting Warner Brothers Music first time out. The “optimal mark” on the Tascam is probably fine.

I won’t keep it a secret. I’ll talk it up as I do it so you can do it without us. That raw clip is pure gold. In ten seconds it’s going to reveal your basic recording problems— if any.

Koz

I did a first quick analysis on the first MP3. You have fans on in the room with you? Can you not? That one thing is going to be a lot of the room noise. Also, as in the comments above, we only need two seconds or so of silence at the top of the test clip and you’re not allowed to check FaceBook or clean your desk while you’re doing it.

My joke is of one poster who was moving around so much I could tell what kind of pants he was wearing. That’s the level of noise we’re talking about.

Nice voice.

Koz

If you’re wondering why I’m banging on about the noise, it can kill some of the other tools. If you need to bring the fullness or loudness of your voice up to make RMS, those fans are going to come right up with you and miss the noise measurement.

Koz

I will get back to you with sample. Wasn’t thinking wav it comes off the tascam as mp3. Do I get a wav by importing mp3 then exporting from Audacity as wav?
FAns!? Could it be picking up my laptop? There’s nothing else electrical around. I must try with and without laptop around.
I read that first post before, that you copied to me in your first reply, and to be honest I don’t understand it. Maybe bc I’m not plugged in to the comp while recording? I’m not recording into Audacity. Should I? I didn’t want my laptop working while I record, bc then it would warm up and hum.
Will be 24 hrs before I’m back here. Thanks for your attention.

I lied. Not gone quite yet, and it doesn’t take v. long to record 10 secs.
Of course it does come off Tascam in a wav, sorry.

Of course it does come off Tascam in a wav, sorry.

Lots of grownups use Tascam for some very high-end applications (NPR), so it’s a good bet it supports WAV, yes.

I need to go have a life for a little while. I want to show you how I found what I’m assuming is a fan sound, so you can do it, too.

Are you in the US or Canada?

There are magic (in the Arthur C. Clarke sense) ways of having hum problems. I have a dead-quiet “Studio” (third bedroom) that has hum. I think it’s coming from power line radiation. I know exactly what I’m listening for and I can’t hear squat. And yet, there’s the clip with hum in it.

Just to keep you from getting suicidal, you’re very close. No obvious distortion, no P Popping, OK noise floor, etc. We’re just mopping up around the edges.

Koz

Well I’ll be waiting with breath that is bated til you get back from having your life.
I am in Canada atm and 4 time zones from you.
Now that I suspect my laptop the noise it emits seems so loud, and I never heard it before. Power lines are not an issue.
I have one other suspect noise source now and want to try another room tone recording with that off, tomorrow.
Thank you!

We’re baaaaack.

I didn’t have any trouble making conformance with either clip, although the one with the laptop running is very slightly more difficult.

I’ll do the noise thing later.

This is going to read a little pedantic like “How To Ride a Bicycle,” but stick with it.

Open up “Without Laptop.” I cheated a little here because I know you’re way further along than many newbies, so I didn’t need the fire extinguisher or the bomb squad. You’re already close.

Select the whole clip by clicking just above MUTE.

Effect > Normalize: [X]Normalize to -3 [X]Remove DC

Drag-select a chunk of silence (more toward the left). Analyze > Contrast: Foreground > Measure Selection. That should be below -60 (higher number).

Drag-select the voice portion (see attached). Analyze > Contrast: Foreground > Measure Selection. I get -21.6 which is between 18 and 23.

OK, we’re done [dusting off hands]. That was ACX Conformance.


Notes. Technically, you shouldn’t normalize to -3. If you now make your MP3 submission from a performance like this one, there is the slight chance that the encoder is going to push your peaks up a little higher than -3 which will ring bells.

Open the MP3 show after you get done and Effect > Amplify to check it. Don’t apply Amplify. just look at it. You should get a number -3 or lower volume (higher number). If you get something like -2.8 instead of -3, then you should trash that MP3 (DO NOT patch it) and correct the WAV. Normalize at, oh, say, -3.2. Make a new MP3. Maybe that should be the goal along along. I’m still making pieces of this up.


You can do that same dance with the “Laptop” clip, except you’re going to find the noise test is closer to failing the critical -60 point because of the laptop fan noise.

I attached the two pictures of a spectrum analysis you’re going to get when you analyze the silent portion of the two clips. Reading along the bottom, you have a vertical purple spike at 240 in one which is missing in the other analysis. That’s the tone, we’re guessing, which is made by your laptop. It’s very quiet and without the test instruments I would never have found it, but if the conformance started to run into trouble, that’s the first thing that would have to go.

To give you an idea of what that is, 440 is the oboe note at the front of the orchestra and 256 is Middle C. So even though it’s really quiet, it’s also right there in the middle of the piano.

Enjoy.

Koz
Screen Shot 2015-02-18 at 17.01.51.png

Sorry. Forgot to post the frequency analysis. Click on the pictures.

Koz
Screen Shot 2015-02-18 at 16.44.36.png
Screen Shot 2015-02-18 at 16.45.45.png

The steep slope on the left side of the spectrum is DC Offset (not a lot, but definitely there and easy to fix).

That DC Offset error should die in the Normalize step.

I think you mentioned this a million miles up the thread. Yes, I think you should be recording with the 80Hz vocal filter applied to the Tascam. That will help suppress any rumble and thunder sounds in your environment. It will help with P Popping and wind, but you don’t seem to have any of that.

Koz

Oh, trust me, I need

, and I’m sure I’m not the only one.

Ok, yay, the noise floor conforms, even with the laptop. I totally get it - there will be no removing the laptop in post, because the tone is right in the middle. It won’t filter out. For here on I’ll record without it.

Yes, DC Offset is selected in Normalize, so I think it’s working.

So, I revisited a pile of recordings I’ve already done, to see if they’re salvagable (all with laptop on, too). I normalized to -3.2, and for every one, the noise floor was excellent (-63.5 to -69). Yay! But…

The voice was uncannily consistent over a half dozen recordings, even different books, different styles, between 26.9 and 28.9. Not so good. So WHY did my 10 seconds conform, if I thought I was doing nothing different?? Is this my levels?

Can this be salvaged?

Then I started fooling around with one, actually the original clip I sent a piece of. Compressor as before, without makeup gain ticked. Worse! Voice up to -30s.
Undo. Compressor with make up gain ticked. Voice -20.6, but now the noise is -56.9.
Is there something to do to fix, or is that my recording levels?

Also, the waveform after normalizing looks nothing like the mastering “video resource” example they give. This bothers me, makes me think there’s more to be done to make it better, even excellent, not just adequate. Isn’t compression vital, for the consistent listening volume? Salvaging comes first though, if you’re willing:)

Still fooling around - can make it work with Normalize, Compression (3:1, NF -50, Thres -21, NO mu gain), and then Amplify. The silence comes up and I think it sounds loud, but it passes, -61.
Seems like too much messing about (why turn it down and then amplify?). Still looks like a ragged, inconsistent wform.
Screen shot 2015-02-19 at 11.55.13 AM.png

pulls out hair

more reading and experimenting:
http://blog.acx.com/2014/07/11/how-to-succeed-at-audiobook-production-part-3/ (normalize to -6!?)
I am not succeeding.

The noise is not a problem, but getting both peaks and RMS compliant is- it does not happen at the same time. My theory is it needs to be compressed more (but I don’t know how to do that, ratio?)- high points squished down, and then all volume up together. Amplify is too smart for me though. I set the high peak to -3.2 and then the RMS is never enough.

The noise is not a problem, but getting both peaks and RMS compliant is

Not exactly. A marginal noise floor is going to interfere with balancing the other two settings. As you noted a bit up the thread, compressing the voice to make it denser and “louder” oozes up the noise floor. Those two operate in direct opposition to each other.

A good dead noise floor opens up possibilities for the other tools.

And yes, once you start missing the settings, it’s a serious juggling act to hit everything at the same time without destroying your voice.

I need to look more.

As we go.

Koz

Absolutely. More flexibility will come when I reduce recorded noise more. But with the fidgeting I’ve been doing with various exisitng tracks, the silence passes. It’s Amplify that’s most dangerous bumping up noise, I find, less than compression.

In compressor effects, setting the “Noise Floor” setting just above the actual noise floor of your recording should prevent, or at least minimise bumping up the noise level.
If you are using a compression effect that does not have a “Noise Floor” setting, then gentle application of a Noise Gate effect before compression can give the same benefit.

One difficulty that will be encountered is that some settings may be “peak level” and others “RMS level”. For real-world audio, “peak level” is always higher than “RMS level” (RMS being a kind of "average level). When using the Audacity Compressor, I prefer to use the “Compress based on Peaks” setting as that uses “peak levels” throughout, which I find much easier to use and to visualise.