Why are Detach At Silences spots removed from audio when exporting to .WAV ?

MP3 (and similar) encode the audio data into a “compressed” format (which is why MP3 files are relatively small). WAV does not do that (which is why WAV files are relatively large). However, Audacity works internally in 32-bit PCM, whereas a 16-bit WAV file is 16-bit PCM, so there is still a “conversion” that occurs.

Because “32-bit PCM” has much higher precision than 16-bit, exporting as 16-bit WAV has to “round” the sample values from 32-bit numbers to 16-bit numbers. One problem inherent in this conversion is that simple rounding causes a problem called “quantization distortion”, which has quite an unpleasant sound on very low level signals. It is therefore customary to apply a small amount of special constructed “randomization” to the rounding process, which replaces the unpleasant harmonic distortion caused by “quantization”, with a less intrusive, extremely low level “dither” noise. More information about “dither” here: https://manual.audacityteam.org/man/dither.html

I order to export an Audacity track with no “conversion losses” at all, you can export as “32-bit float WAV”. The disadvantage of the format is that a lot of other applications don’t support the format (though most “professional” audio applications do).

The conversion from 16-bit to 32-bit is perfect (completely lossless). Any mixing or processing that you do benefits from the extremely high precision of 32-bit float. The “problem” comes when mixing back down to 16-bit (exporting). Most 32-bit float sample values lie between 16-bit values, so cannot be converted exactly, so requires some kind of “rounding”.

It’s probably easiest, and sufficient, to think of this by way of analogy.

Say you are measuring something that is about one metre long, and your tape measure is marked in centimetres. From the tape measure, you write down the answer. Say the answer is 97 cm. Now convert that to millimetres and the answer is 970 mm. The conversion from cm to mm is “lossless” because “97 cm” can be expressed exactly as 970 mm.

Similarly, every 16-bit sample value may be expressed exactly as a 32-bit float number. “32-bit float” has much finer precision than 16-bit.

With the “Amplify” effect. People tend to think of “amplifying” as “increasing” the level, but that is only “positive amplification”. You can also amplify by a negative amount, which reduces the level - this may be called “negative amplification” or “negative gain” or “attenuation”, but whatever the word, it is amplifying by a negative amount.

The Amplify effect in Audacity defaults to a setting that will amplify (up or down) to bring the peak level to 0 dB.

A word of caution about the red “clipping” lines. They do not indicate that the audio IS clipped, only that it “may” be clipped (is “probably” clipped). If a sample value is exactly at 0 dB, then a red clip line will appear. Because of this, even after amplifying to 0 dB, there may be an occasional red “clip” line, even though the audio isn’t actually clipped.

Note also that filters may cause the peak level to increase.
MP3 encoding may cause the peak level to increase.
These increases are generally small, but illustrate why it’s a good idea to always allow a little “head room” in your exported files. Modern CD recordings frequently go right up to 0 dB, (which ideally they wouldn’t, but many do).

There is a universal level at which digital audio is clipped, and that is “0 dB” (also known as “full scale” because that it the full range for “valid” digital audio). Yes that’s where “dBFS” comes in (dB relative to Full Scale).

“32-bit float format” (in fact, any “float” format, but 32-bit float is the only float format commonly used in audio) is peculiar in that it can go over 0 dB (higher than “full scale”). “Integer formats” (such as normal 16-bit and 24-bit) cannot go over 0 dB. For integer formats, 0 dB is an absolute limit.

All current sound cards / audio devices work with “integer format” digital audio (nearly always 16-bit or 24-bit), so they are also limited to an absolute maximum level of 0 dB.

Clipping is potentially damaging to speakers. Clipping (especially “digital clipping”) creates high levels of very high frequencies. If played at high volume (the amp turned up high), there is a risk that it could blow the tweeters.

The “universal level” refers to the “digital signal”. The “analog level” (the signal / sound level at the speakers) is determined by a combination of the signal level going into the amp, and the amount of amplification (how high the amp is turned up).

This is one of the benefits of mic’ing up a guitar amp rather than “DI” (direct inject) the stomp box straight into an audio interface. Guitar amps (usually) don’t have a tweeter, and naturally filter out a lot of the extreme high frequencies.

Some stomp boxes include a speaker / cab simulator, which has a similar “filtering” effect as a real guitar cab, but many stomp boxes don’t.