I finally got a little test clip...

I think I’m fairly close to meeting the ACX requirements, which is goal one for me. If I can do that, I guess I can work on the voice & reading quality next. (I had a little congestion & broke out a neti pot I had used once but put away when my kids warned me about people getting killed by brain eating amoebas after using neti pots. More than likely mine would starve to death, but I boiled the water first anyway. I think it actually helped a little, but still sounds a little nasally or something I think. I probably should do that an hour or so before I try to record something, I guess.

ACX requirements:
measure between -23dB and -18dB RMS and have -3dB peak values and a maximum -60dB noise floor

I used (hopefully) Kozikowski’s method of testing for those:

Click just above the Mute button to select the whole clip.

Analyze > Contrast: Measure Selection. Read -21.1dB.
Effect > Amplify. Read -3dB peaks.

Drag select some portion of the Room Tone silent stretch at the beginning.
Analyze > Contrast: Measure Selection. Read -69.8dB.

So that’s the three numbers. All pass ACX compliance.

I measured them in the Audacity original and came up with:

-21.9 foreground (full 10 second clip)
Amplification (dB) 4.3
-57.8 background

For what it’s worth, there was a slight variation when I did the same tests on the exported WAV clip, not sure if that’s normal or not:

-21.8
4.8 dB
-57.8

This is my second attempt, btw. My first one was too loud and had an Amplification dB of 0 so I redid it. The difference between the Audacity stats and the exported WAV file stats were larger on that one.

Any tips, tricks or recommendations? Would a longer recording like this be salvageable?

Are the contrast results between the foreground and background significant with respect to the ACX requirements? That stat for the attached WAV file was WCAG2 Pass 36.0 dB Average RMS.

Thank you.

I’ll do the full-on analysis when it’s actually daytime in this time zone, but you appear to have a common problem with the specifications.

Remember the oddity that smaller numbers are louder. Zero is maximum loudness. Minus 70 is very quiet.

The first number, loudness is about right, but the other two are off and they’re off in the wrong direction. The peak values (second number) are too low and when you fix that by boosting the volume slightly, the last number, background noise, which is already too loud, is going to get worse.

This is the place where people lunge for Effect > Noise Reduction. Depending on what kind of noise you have, it might be enough to juuuuust push you over compliance. Our measurement techniques and ACX’s are not surgically identical. It’s still possible to get a compliance bounce from them and still pass ours if the show is right on the razor’s edge.

ACX is firmly opposed to tricks like this because they make audiobook production, already a difficult prospect, much more difficult. They stress multiple times in their videos that consistency is a really big deal, and consistency in the face of constant tricks is not likely.

I made a collection of ACX help resources. Pay attention to number two. That’s the video made by an actual recordist and his recommendations for home recording.

\

• ACX AudioBook Resources
  http://blog.acx.com/2014/06/26/how-to-succeed-at-audiobook-production-part-1/
  http://www.acx.com/help/video-lessons-resources/200672590
  http://www.acx.com/help/acx-audio-submission-requirements/201456300
  http://g-ec2.images-amazon.com/images/G/01/Audible/en_US/acx/pdf/ACX_MP3_Encoding_Tutorial_04242014._V338366800_.pdf

Koz

Unfortunately the ACX noise floor spec is for un-weighted noise as well. IMHO your recording is reasonably quiet, but the noise floor is higher than the spec by a few db. I hear a hum that sounds like mains hum or could be a fan of some sort.

Spectral analysis shows a bunch of sub-sonic noise possibly 1/f noise from your microphone preamp. As a test I applied the “high pass filter” effect with a cutoff frequency of 30 Hz and 48 db/octave slope. The result was a nearly 10 dB improvement in your noise floor from -56 dBFS to -65 dBFS. So that’s a repeatable production step that you could just add to the process and get through the gate.

I would try to track down the source of the hum however, as it’s the audible bit in your noise and is unaffected by the filtering mentioned. Attached are spectra from the silent bit at the front of your recording before and after applying the filter mentioned above. The “hum” is those peaks at 60, 120, 240 & 360 Hz (with the ones 240 and 360 contributing much more to what you hear than the big spike at 120.

And that is the magic place for Steve’s voice filter. Everything below about 100Hz is gone. It only affects deep (usually) male voices, but the boost in specification quality is usually worth the effort. Note that sometimes it’s good to run that after Noise Removal because the current version of Noise Removal can add rumble and sub-sonic noise by accident.

Unzip the preset and import it into Effect > Equalization. Run it with the “Length Slider” on the lower right all the way up.

http://manual.audacityteam.org/o/man/equalization.html
Somewhere in there it tells you how to import custom tools.

I expect it to look like the attached picture when you run it.

As above, the decrease in background noise may be enough to get you through the process.

I’ll give you the full list of tools in a second.

Koz

LF_rolloff_for_speech.xml.zip (326 Bytes)

You did one thing wrong right at the top. The Room Tone features you shifting in your seat and adjusting your pants. You need to absolutely freeze and stop breathing for that two seconds. This is an uphill battle because most people see any gap in their daily activities as the place to check their Twitter feed. See: Traffic lights.

ACX wants a much longer 30 second Room Tone and they want you to get up and go for coffee during that time (you don’t need to stop breathing for theirs).

Koz

See, not that hard. Attached. Run the three tests on that.

How to:
Select the whole (raw) clip by clicking just above the MUTE button.
Effect > Normalize: [X] Remove DC, [X] Normalize to -3.2dB
Effect > Equalize: LF rolloff for Speech plugin.

Done.

Even better, it still sounds like you. You should really try to figure out where that hum is coming from. Mine is coming from the Great Hummy Monster that Lives in the Attic.

Koz

A note: Everybody is expecting you to make normal human noises during the presentation. Adjust your pants then. One common problem with people using Noise Removal is they have no pants.

That two seconds is magic because it’s used for the Noise Removal step if necessary. You body has to be exactly there for the sound profile, but making no noise.

Koz

Thanks much, I will check those out and bookmark them. Actually, if I can figure this out reasonably well, I’d like to either put up a web page with a step by step tutorial on best practices with links to resources such as these & maybe to discussions on this forum on some topics… or maybe to put that into your Wiki, if that’s something we’re allowed to do.

Duh, I think I had left my stereo on which was causing that hum. I had been using it to listen to streams from my computer via the UCA202 interface. It definitely does have an audible hum to it, but I think that I’ve become so acclimated to it that I often don’t notice it anymore. My wife does, though, when she comes into my office.

Spectral analysis shows a bunch of sub-sonic noise possibly 1/f noise from your microphone preamp. As a test I applied the “high pass filter” effect with a cutoff frequency of 30 Hz and 48 db/octave slope. The result was a nearly 10 dB improvement in your noise floor from -56 dBFS to -65 dBFS. So that’s a repeatable production step that you could just add to the process and get through the gate.

That is a really neat tool to have in one’s toolbox. Is that something that can be done through Audacity, or are you hooking up an oscilloscope or some other specialty piece of hardware for those graphs?

I would try to track down the source of the hum however, as it’s the audible bit in your noise and is unaffected by the filtering mentioned. Attached are spectra from the silent bit at the front of your recording before and after applying the filter mentioned above. The “hum” is those peaks at 60, 120, 240 & 360 Hz (with the ones 240 and 360 contributing much more to what you hear than the big spike at 120.

It’s possible that my laptop might be contributing some noise as well. It’s usually fairly quiet as it’s a hybrid with a small SSD for the common functions and a bigger regular hard drive for storage or when otherwise needed. The fan only kicks on occasionally & doesn’t seem real loud. I may need to rearrange my desk a bit & see if I can move it more behind the mic, outside of its cardioid pickup pattern.

Thank you very much for those graphs. That is an awesome way to see what’s going on. I’m actually a little surprised that it picked up stuff below 20 Hz, though, as the specs for the mic stop there. Maybe they only graph down as far as human hearing, or don’t want you to know that it records things down that low.

Yes, that’s what I thought. They do a really good job of step-by-step. Watch the videos. The pieces they don’t cover are generally custom applications that don’t lend themselves to general rules or videos. They do many descriptions of putting padding on the walls and beach towels on the desk under the microphone. That’s the low-hanging fruit. What do you do when you have Frying Mosquitoes in the background when you stop talking? How do you get rid of a dog barking in post production?

There are people who fall off the bottom of the production process. “No matter what you do, that microphone is never going to make ACX compliance.” That’s the sort of thing that doesn’t lend itself to easy answers.

But yes. Third party submissions are welcome. I wrote the original overdubbing tutorial.

Koz

That stuff below 20Hz is a little magic. I’m not entirely sure where it’s coming from and it’s deadly because it can throw off noise measurements with “noise” nobody can hear. I suspect it’s coming from the digital/analog interface system. Filtering that stuff out of an analog sound device is expensive and nobody can hear it anyway, so why bother?

Koz

Thanks much. I’m adding this to my growing “How To” list. Hopefully just turning off my stereo will solve a lot of that low hum problem, but I imagine this might help with the occasional rumble of trucks in the distance, or maybe an airplane flying overhead. I’ve heard a few small planes in the background today & decided to look up airports in the area on Google Earth. It looks like the closest one is a small airport across the river, but less than 3.5 miles away, and a couple more within 10 air miles, including St. Louis International, which has a runway about 6 miles from our home. I guess we must be in a higher air traffic area than I had suspected.

It will not cure “sub harmonics.” That’s only good for problems above 60 (in the US). The stuff below that is very likely coming from your equipment, as you point out, the microphone isn’t rated that low for actual sound.

But now you know how to test for problems and in general, how to fix them.

ACX has an interesting process for noise, having, no doubt struggled with this for years. Use their 30 second Room Tone recording technique and then play it back at full volume. What do you hear? They’re going a different direction than we are, so you have to pay attention to their instructions.

That will give you the dog next door, the refrigerator and the stereo hum, but not that stuff below 20. For that you need the analysis tools.

If you get far enough into their videos, they do recommend the Shure A15HP hardware device that does a similar job to Steve’s custom equalizer filter.

http://www.shure.com/americas/products/accessories/microphones/microphone-problem-solvers/a15hp-inline-high-pass-filter

The Radio/TV/Movie people have the same problem you do. Nobody is going to pay me for sound nobody can hear, but which may damage my show.

Koz

One other note. I have gone on “hum hunting” trips by running my sound recording system in monitor mode into my headphones (right-click in the red recording meters > Start Monitoring). It does everything but “roll tape.”

Turn it up and walk around and see if I can figure out where the hum is coming from. That’s how I found the location and direction of the microphone that I used to shoot my sound tests. "OK, up a little, up a little. Woof. Turn towards the street…little more. Woof. Record the show quick before that $!@#) comes back.

Or: “Hey, I didn’t know my external disk drive made hum noises…”

“You know that device power supply under the desk with the little green light on it? Hum city.”

Koz

Yes it’s in off-the-shelf Audacity. Analyze → Plot spectrum.

Set “Size” to 65536 and “Axis” to “Log Frequency” to get plots Like I posted.

Back in the ol’ days spectrum analyzers were horribly expensive test instruments (10x the price of a good oscilloscope), and most of us just did without, but now with computers it’s trivial.

The computer will happily record frequencies as low as you are willing to wait for. It is unlikely the noise is from any acoustic source as I’m sure you would be aware of it if it were. I suspect that it is in your microphone pre-amp. Here’s a relatively technical treatise on the subject http://www.analog.com/media/en/training-seminars/tutorials/MT-048.pdf/

I tend to not think that. I think it’s Frying Mosquitoes where the audible portion of the USB/digital noise has been filtered out and the manufacturers called it good and shipped it. Who else but us would be concerned with noise at 7 Hz?

Low frequencies are precisely where filtering hardware starts getting expensive.

Koz

Thanks. That is a very, very handy tool. It really did seem to fix things for me by doing the equalizer thing and start dropping the volume of frequencies below 100 Hz. That alone was dropping the noise floor for -50s to -70s which makes all the difference in the world when the ACX cutoff is -60 dB. It’s the difference between passing and failing, at least that important test. If I couldn’t get past that hurdle, I couldn’t even get out of the gate.

Back in the ol’ days spectrum analyzers were horribly expensive test instruments (10x the price of a good oscilloscope), and most of us just did without, but now with computers it’s trivial.

We’re very blessed indeed. It’s hard to believe how far we’ve come in just my lifetime. My parents witnessed some major advances in their lives & my grandparents would probably have been astounded to see the technology that would be available to their grandchildren.

[quote=“JeffB”]
Thank you very much for those graphs. That is an awesome way to see what’s going on. I’m actually a little surprised that it picked up stuff below 20 Hz, though, as the specs for the mic stop there. Maybe they only graph down as far as human hearing, or don’t want you to know that it records things down that low.
[/quote]

The computer will happily record frequencies as low as you are willing to wait for. It is unlikely the noise is from any acoustic source as I’m sure you would be aware of it if it were. I suspect that it is in your microphone pre-amp. Here’s a relatively technical treatise on the subject > http://www.analog.com/media/en/training-seminars/tutorials/MT-048.pdf> /

Well, I thought the computer would only be able to record what the microphone picked up, and the mic’s chart stopped at 20 Hz. Live and learn, I guess… and that’s a critically important thing to learn in my particular case.

Turning off the stereo helped some. I waited until this evening to try and record some again and I heard what sounded like an airplane at some distance, so I waited for a bit, but it never seemed to go away. I went outside to verify that it wasn’t coming from in the house, and sure enough it was louder out there. I’m thinking now that there’s a constant hum from planes… there are 3 airports within 10 air miles of here per Google Earth, and then there’s an interstate highway that’s .82 miles from our house as a crow flies and curves around some staying pretty close to us for a few miles. I’ve probably gotten used to it and just don’t notice it anymore.

Thanks for the link to the article, btw. Some of that was a bit too technical for me, but I don’t think I needed to know those parts anyway, and I can still pick up some things here and there. One thing they mentioned was white noise, and if I recall a couple of you guys have talked about white &/or pink noise generators on here. Is that something that Audacity can do? Would that be used at all in the context of audiobook production?

Yeah, it seems like ACX should probably modify their requirements a bit to accommodate that type of thing. It’s a good thing to have quality control, but there’s no real benefit to having noise outside the range of human hearing to reject recordings. I wonder how many would be producers had a product the consumers would have enjoyed but were rejected and eventually gave up trying.

Yeah, it seems like ACX should probably modify their requirements a bit to accommodate that type of thing. It’s a good thing to have quality control, but there’s no real benefit to rejecting recordings because of noise outside the range of human hearing. I wonder how many would be producers had a product the consumers would have enjoyed but were rejected and eventually gave up trying.

Yeah, it seems like ACX should probably modify their requirements a bit to accommodate that type of thing.

It’s not up to them. They’re in the business of supplying MP3 sound files and MP3s can get seriously screwed up with sound that’s either too low or too high.

I got your clip to easily pass with two simple tools. Are we fighting some other battle?

Koz