Using the Amplify Effect When Exporting WAV and MP3 Files

Hello:

I would like to confirm that the AMPLIFY effect setting and process I am using in Audacity to export recordings of records and tapes is not damaging the music.

I am using Windows 10, and Audacity 2.1.3

I am very pleased with the recording quality of Audacity for recording old records and tapes, but have a question about whether the final Amplify settings I am using are damaging the music tracks that I am exporting.

I am first capturing a ‘raw’ recording of an album/tape into Audacity to max out at around -6dB.

Once the original recording is produced in Audacity, I save a copy of that raw file, and then add the songs to the file. Once the songs are added to the file, I then export the tracks as WAV and MP3 files (see below) AFTER applying the Amplify effect to max out at -1dB.

I am exporting three different sets of the same music file tracks with each album I record:

1. A set of tracks at 32Bit/41,100 Hz for high quality listening
2. A set of tracks at 16Bit/41,100 Hz for recording CDs, if needed, and
3. A set of MP3 files at 320Mbps.

To verify that Audacity is exporting these files at the ‘amplified’ volume I chose (-1dB), I test the files by importing them back into Audacity, selecting Amplify, and then checking the Amplification level that Audacity indicates the track is at. The tracks are usually at -1dB, as I requested, but occasionally will indicate fractionally different levels, particularly at the 16dB level, and even more so at the MP3 level indicated above. For example, rather than the -1dB that I set the Amplify effect at for export, I will sometimes see that the songs’ amplitude levels are -1.019, or -.9901, or -1.0001, or -.97091, rather than the -1.0dB that I requested.

My questions are the following:

1. Is setting the amplitude level of -1dB for these three format levels acceptable, in order to prevent song damage/i.e. clipping?
2. Is my methodology for verifying clipping/song damage of importing the files back into Audacity, after having originally exporting them, a sound/logical way to verify that the song(s) have not been ‘clipped’ or damaged? (I have been understanding from various forums that clipping/damage only occurs when the amplitude level of a file goes above 0dB- though, the the exact level above 0dB at which clipping occurs is not known).
3. If setting the amplification level to -1dB for any of the formats listed above is NOT going to guarantee the prevention of song damage under the conditions I have indicated above for capturing music recordings in Audacity, what setting can one set the Amplitude level to, in order to guaranty preventing song damage?

Thank you, in advance, for your reply.

Kind Regards,

BlethiK82

0dBFS is the “digital maximum” for integer formats. It’s as high as you can “count” with a given number of bits.* Floating-point and MP3 files can go over 0dB without clipping . (And since Audacity uses floating-point internally, it can go over 0dB and you are “safe” as long as you adjust-down before exporting.)

Analog-to-digital converters (recording) and digital-to-analog converters (playback) are integer devices and are hard-limited limited to 0dB. If you have an MP3 or floating-point file that goes over 0dB and you play it at “full digital volume” it will clip. It may not clip if your software volume control is turned-down.

MP3 is lossy comression. The wave shape changes making some peaks higher and some lower. It’s not exactly predictable, but usually the highest-peaks are higher after compression, especially if the original was “artificially” dynamically-compressed or limited. However, we already know MP3 is imperfect and I’ve never heard of a case where that slight clipping was audible… If you hear a compression artifact, it’s probably something else and it’s probably not going to go-away if you reduce the volume below clipping. And, these higher short-duration peaks don’t make it sound louder. Many commercial MP3s go a little over 0dB, the ones I make usually do to and personally I don’t worry about it!

If it’s super-important to you to avoid clipping, use an iterative process… Load the MP3 into Audacity and run Amplify to see if it goes over 0dB. If it does, re-load the original (not the MP3**), reduce the volume a little more than before, and re-export to MP3, then check it again, etc.

1. A set of tracks at 32Bit/41,100 Hz for high quality listening

That’s not necessary. You don’t have a 32-bit ADC or DAC, you’re ears don’t have that kind of range, and records & cassettes have nowhere near that kind of resolution (due to the noise floor).

Pros record at 24/96 (which can be argued as overkill, although they do usually leave lots of headroom, which means they aren’t using all 24-bits). Audio processing is usually done in 32-bit or 64-bit floating-point because the math/DSP is easer in floating-point, and because you don’t have to worry about (temporarily) going over 0dB.


The guys who do scientific, blind, [u]ABX tests[/u], have pretty-well demonstrated that nobody can hear the difference between a high-resolution original and copy downsampled to 16/44.1kHz. That shows 16/44.1 is better than human hearing. (“Audiophiles” who claim to hear a difference usually claim there’s something wrong with the test, or they claim that blind listening tests are never valid, etc.)



\

• A 0dB 24-bit file has “bigger numbers” than a 0dB 8-bit file, but you driver scales the numbers to match the resolution of your DAC, so they play back at the same volume. (In floating-point 0dB is 1.0.)

** The “damage” done by MP3 compression is cumulative and you should avoid multiple generations of lossy compression whenever possible.

Thank you Doug.

Based on what you are saying here, it sounds like my logic for verifying amplitude is correct, and that I am setting my amplification levels accurately to guaranty that clipping is not occurring.

Kind Regards,

BlethiK82

it sounds like my logic for verifying amplitude is correct, and that I am setting my amplification levels accurately to guaranty that clipping is not occurring.

Yes. Your procedure is good!!! Except possibly some MP3 files… if you’re worried about that… Again, the MP3 file won’t clip but it can go over 0dB and clip your DAC.

And, you don’t need to leave 1dB of headroom on the WAV files. Nothing bad happens when you get close to the limit, only if you “try” to go over. Some people normalize to -1dB because they worry about “intersample overs”, but there is NO intersample information in the digital audio file itself and there’s no reason the reconstructed analog output from the DAC can’t go over 0dB (it’s just a limitation of the binary-integer sample-values).