Quote:
In the time domain, the amplitude of that note at any moment will be the sum of the amplitudes of the that fundamental and all these harmonics. The familiar time domain shape of a square wave occurs because the fundamental and the harmonics are occurring at the same time. . . .

A full-band compressor operates on the entire spectrum simultaneously.

While the total amplitude of an instrument's note is indeed the fundamental and its harmonics, a compressor does not act on the entire frequency spectrum simultaneously.

Even a recording of an entire symphony orchestra is a single wave form. The blast of a trumpet playing a high note and the whisper of a gently bowed double bass are both contained in this same waveform. If they are playing the same pitch three octaves apart, the trumpet is, in essence, playing the harmonics of the bass.

If one applies a compressor to decrease the intensity of the trumpet, this changes the relative amplitude of these higher frequencies to the bass's fundamental. Thus, the "harmonics' of the bass have changed.

Of course, we can also have a double bass playing loudly up front and the trumpet playing softly off-stage with a dense cup-mute. Apply the compressor and the double bass decreases in intensity, and we hear more of the trumpet. That is, we hear more of the bass's higher frequency "harmonics". The character of the sound once again changes.

A compressor does not know that certain frequencies belong to one instrument or that the belong to another's. It makes the louder frequencies softer, while leaving the softer frequencies alone. Thus, it changes the relative differences between fundamentals and harmonics, as well as the balance between instruments.

In summary, a dynamic range compressor is not a volume control. It changes relative volumes so that they are more the same - it compresses the dynamic range of the input material.

The attack and release settings of a compressor are something else entirely. Bad settings can indeed make the sound quite odd. For example a very quick attack on a snare drum destroys the "pop" of the snare. It the attack is slower more of the initial quick transient sound of the snare is let through - keeping more of the characteristic sound of a rim shot. Of course, sometimes the sound of a heavily compressed snare is perfect for a particular mix.

On the other topic, the Aphex Aural Exciter adds harmonics to a waveform. Used judiciously it adds a certain kind of presence or excitement that can be appealing on a pop recording or dull sounding live sound reinforcement. It is an effect however, not a sound improvement device per se.

Some like the sound on a pop recording. Others can't stand them and won't touch one in a studio.

I can't imagine using one in a hi-fi system, especially not on anything acoustic.

The DBX expanders can be fun to play with (the increase dynamic range) but they never worked well enough to offset their obvious artifacts, in my opinion.

Very nice examples Elk.

Dave

Hope you're having fun with this thread - I am. Quoting from this tutorial on audio compression:

Quote:
Be careful when compressing an entire mix. Very often in pop music there is a bass line which is generally constant in level. If there are sudden very loud peaks such as brass stabs, the compressor will lower the whole track at that point, with the result that although you may get an overall levelling of dynamics and increased loudness, the bass line will dip at this point and lose its flow. This is sometimes called pumping. This problem can be overcome by using a multiband compressor, which splits the signal into different frequency ranges, and compresses them individually. In the above case the brass stab would be compressed, but the lower frequencies can pass through a lower band of the compressor with no or very little compression. It is usually best to mix a track with no, or very gentle, compression, then apply compression at a later stage (final mastering).

This tracks well with your example of a trumpet and double-bass. As described in this quote, a multi-band compressor will allow leveling of the overall dynamics of a track by altering the attack and decay envelope in a particular band while not causing "pumping" of instruments in other bands. This type of compression can cause altertion of the ratio of harmonics (timbre) for a particular instrument as the fundamental may be in a different band than its harmonics. This is not the case for the full-band compressor that operates on the entire spectrum simultaeously.

The previous bullet in the tutorial is also pertinent:

Quote:
Listen carefully. Compression can affect the timbre of an instrument, either because of the inherent sound of the compressor, or because the peaks of an instrument may have a different tone to the troughs, so reduction in level of the peaks relative to the troughs changes the overall tone. This can be especially true when applying fast compression to instruments with broad vibrato.

In this case, the attack time applied by the compressor is shorter than the period of the vibrato for that instrument. This, of course, can radically alter the timbre of that instrument.

I looked for a good description (hopefully with pictures) of how compression can affect timbre. Unfortunately most articles just state that compression affects timbre, but don't go on to explain why in any meaningful sense.

Audible pumping doesn't occur unless one is particularly ham-fisted with the settings - then it can get pretty bad.

JA pointed out some time ago that compression can also affect relative timings between frequencies (we were discussing PRaT at the time). I took some uncompressed recordings I had made and listened carefully and found that he was right. It's very subtle but different frequencies can be affected differently in the time domain as well.

Quote:

Quote:
Listen carefully. Compression can affect the timbre of an instrument, either because of the inherent sound of the compressor, or because the peaks of an instrument may have a different tone to the troughs, so reduction in level of the peaks relative to the troughs changes the overall tone. This can be especially true when applying fast compression to instruments with broad vibrato.

In re-reading this it occurs to me that the writer is actually referring to tremelo, which is a variation in amplitude - not vibrato, which is a variation in pitch.

(The writer is probably a guitarist. Electric guitar players reverse the meaning of the two terms. I think Leo Fender got them confused early on and the misuse stuck.)

Quote:
This tracks well with your example of a trumpet and double-bass. As described in this quote, a multi-band compressor will allow leveling of the overall dynamics of a track by altering the attack and decay envelope in a particular band while not causing "pumping" of instruments in other bands. This type of compression can cause altertion of the ratio of harmonics (timbre) for a particular instrument as the fundamental may be in a different band than its harmonics. This is not the case for the full-band compressor that operates on the entire spectrum simultaeously.

I played with a single band compressor last night and conducted a little experiment.

I created a test WAV file with two generated tones, a 500Hz sine wave and a 1kHz wave - that is, one octave apart. I made the 500Hz wave to have twice the amplitude (6dB) of the 1kHz tone.

Applying 2:1 compression with a -10dB threshold greatly decreased the level of the louder tone. The audible difference (timbre) in the combination of tones was readily discernible.

I created more complex combinations of sine waves approximating the harmonic structure of an instrument. Whenever compression was applied the relative strengths of the harmonics changed. This was both audible and easy to measure using a spectrogram.

A compressor decreases the amplitude of the frequencies that are louder than the set threshold by the ratio of compression chosen by the user. This changes the relative amplitude of the involved frequencies.

Those louder than the threshold (fundamental and loud partials) are brought down in level. Those softer than the threshold (quieter harmonics) are left untouched. This changes the ratio between the fundamental tone and the harmonics, changing the resulting timbre.

Interesting how much easier this is to hear on steady state generated test tones than on real instruments. My hypothesis is that the ear has nothing else to listen to than the single change when considering test tones. Real instruments have much more complicated spectra which changes over time.

Now if a speaker dynamically compresses the sound signal in a non-linear fashion this will change the timbre of the instruments as well as change how the space around the instruments sounds as well.

I assume your test waveform is comprised of the 500Hz and 1KHz signals mixed at constant amplitude with a total peak amplitude of something significantly higher than -10dB (0dB?). Could you please:

• Set the compressor threshold to 0dB/2:1 ratio, record the signal processed by the compressor and perform an FFT on the recovered signal.
• The FFT should show the two peaks, one at 500Hz with 0dB amplitude and the other at 1kHz at -6dB. This would demonstrate that the compressor does not have inherent distortion.
• Decrease the threshold to -10dB on the compressor, record the original signal as processed at this setting and perform an FFT on the result (using the same window and sample size).
• Post the result.

My expectation is to see a peak at 500Hz that is at -6dB (half amplitude) and a peak at 1kHz at -16dB.

A full-band compressor uses a processed version of the RMS amplitude of the incoming waveform to control that gain of an amplifier. This processing is what controls the attack and decay of the compression.

To exhibit the behavior you describe, the compressor would have to somehow operate on the various frequencies in a musical waveform independently. This could be done with a multi-band approach in which several compressors, each controlling a separate frequency band, are operating in parallel with the outputs summed. This process could also be done digitally, but the implementation would be a lot more complex.

I'll do some more experimenting when I get a chance. Good suggestions.