Talk:Compressor

Discussions archived from the page

RMS

Gale: haven't followed the discussions on this, but needs explaining why RMS threshold line is not at -18 dB. How was the graph produced? The Audacity compressor graph can't produce an input/output at -60 dB with a -18 dB threshold.

Steve the Fiddle As is indicated on the graph, the x axis shows peak dB. A waveform that has an RMS level of -18dB will have a peak level somewhat higher than -18dB (as is indicated on the graph). The exact peak dB level is dependent on the waveform but for a sine wave is approximately 3dB higher than the RMS level.

When using RMS compression, the compression begins at a level several dB higher than the peak level that the user can observe. This impacts on the user and so has been included in the illustration.

For example, if the compressor threshold is set to -18dB and the effect is set to compress based on RMS, the effect on a sine wave will be that compression does not start to occur until the RMS level reaches -18dB which is equivalent to a peak level of -15dB. For square waves the peak and RMS levels are equal. For white noise the RMS level is approximately 4.5dB below the peak level. For white noise, while the compression begins when the RMS level reaches -18dB, the peak level will be around 13.5dB

The graph, as it says in the text directly above the graph, is an illustration. It does not represent the graph display in the compressor GUI, which as explained in the previous section is not an accurate plot of the actual input/output levels. The graph that is displayed in the GUI is based only on the Threshold and Ratio slider settings. It does not take account of any other settings such as make-up gain or if RMS based compression is selected (in which case downward compression is used).

Looking at this again I can see that the graphs are not quite correct. I'll have another go at this. Complete accuracy can not be attained with a simple 2 axis graph because the actual output is also dependant on the Noise floor setting and the time domain so we are just trying to indicate a general expectation of what will occur without getting bogged down in too much technical detail. (and the precise working of this compressor are very complex with peculiarities unknown in any other compressor).

Gale: Several people have complained about the illustrative graph and I don't have knowledge enough or time enough to research the subject. But fundamentally the message seems to be the illustrative graph is confusing (novices) and it seems to beg questions that (as you suggest) may be too complex to explore. As it is I think it needs to be explained at least a bit more, or retired.

• On the input axis of the graph, the RMS level appears to be above the peak level (louder) because of the arrow labelling, yet people can see the input waveform and RMS there is of course below peak. The blue dotted vertical line suggests that when the effect is set to compress based on RMS, compression starts at a peak level of -15 dB or whatever, true? OK, so I suggest we say the graph values are peak values and the arrows should say "-18 dB threshold using peak compression" and "-15 dB threshold using RMS compression".
• Does the red line in the non-GUI graph actually illustrate peak (upwards) compression?

Bill 02Jul11: I vote to retire the graph, despite the work that Steve and I put into it. I don't think we'll ever reach a satisfactory version that is both accurate (and thus appropriate for audio pros) and understandable to novices. As for those experienced with hardware compressors and who want that type of experience, they're going to have to use a DAW that has real-time effects, since IMO you really need real time feedback as you make adjustments to get the right settings for the material you are compressing.

• Peter 3Jul11: ok, gets my vote for graph removal too - not worth the hassle
• Steve 4Jul11: OK, graph removed - does the compressor require any further explanation or is it sufficiently clear as it is?
  • Peter 7Jul11: seems clear enough to me.

Regarding the example

Gale 04Nov11:

• Should we not mark the threshold line?
• Is it a good idea to show what "some" compressors do rather than say what ours does? Our release does seem to affect audio below the threshold.

Steve 04Nov11:

• Images updated to show compression with Audacity 1.3.x Compressor.
• Gale 23Nov12: These images are still pretty confusing IMO. People complain and I don't really understand enough. Do they ever come to the Forum?
  • "A simple sine wave that drops off by 12 dB". Does the first image show that? Do you mean it has already dropped off from 1 to 0.25? It doesn't look like 0.25. Where is the selection region you mention in the text?
    • Gale 11Dec12: Steve has updated one of the images, but we say "when Compress based on Peaks is selected, audio with a peak level within the threshold range will be boosted". Yes, but we also say peaks compression "makes the entire audio louder, but amplifies the louder sounds above the threshold less than those below it". This still holds true but only so far that the louder part doesn't appear to be amplified at all...

      Also as complained about before, the second image shows the gain reducing inside the threshold when it is supposed to be increasing (and before the attack time, the audio is already louder than it was before, where you imply the compressor is not working).

  • Is Release Time "How soon the compressor starts to release the volume level back to normal after the level drops below the threshold"? Or is it "how long it takes to return to unity gain", as most docs describe it and your second image appears to show?
  • Our Attack Time description "How soon the compressor starts to compress the dynamics after the threshold is exceeded" is a common description, but users may see this as a "waiting time" before the compressor does anything. But you state the compressor is working within the attack time. One user is very confused that the audio is being reduced within the attack time rather than increased when compression based on peaks is supposed to be on.
  • Several people have suggested we need four images:
    1. Before compression
    2. After RMS compression, make up gain "off"
    3. After RMS compression, make up gain "on"
    4. After Peak compression

Not sure about #3 but I strongly agree otherwise. And I think images like the ones I use below (block tones, but with the time range added) would be much better. Perhaps we should have a more hand-holding description with more images on http://wiki.audacityteam.org/wiki/Compressor rather than even attempt it here?

• A very common source of confusion is how much the compressor controls are interdependent. Let's look at this which is a 6 seconds tone:

  Then apply compressor to the entire audio at Threshold -12, Ratio 6, Attack 0.2, Release 1, compress based on rms, make up gain off:

  

  • The length between the end of the higher tone and the resumption of steady amplitude in the lower tone is 0.5 seconds, not 1 second.
  • Why is the difference between the peak of the lower tone and the peak of the higher tone not 1 dB, since the increase between them was 6 dB and the ratio is 6:1?
  • If we Undo then rerun the compressor at a threshold of -3 dB, why is the higher tone untouched, but at threshold -4 dB, the higher tone is reduced?
  • Why is the lower tone affected at all at the start if the threshold is - 3 dB or -4 dB?
  • Also we need to explain why the whole result is at a much lower amplitude if the threshold is say -24 dB than say -12 dB (with make up gain off). Users often cannot understand this.

• Steve 24Nov12:As this is such a complex issue I have commented on the forum: http://forum.audacityteam.org/viewtopic.php?p=198262#p198262

Older archived material

Bill 28Nov09: I'd like to keeps these graphs and discussion here for a while. False starts, old page text, and text that made it to the main page have been removed.

Technical details and input-output graphs

When "Compress based on peaks" is not checked

The following image shows the input-output characteristic of the Compressor effect when "Compress based on peaks" is not checked.

• The light blue line shows the characteristic when "Make-up gain for 0 dB after compressing" is not checked.

In this mode the Compressor effect behaves like a traditional hardware compressor.

• The dark blue line shows the characteristic when "Make-up gain for 0 dB after compressing" is checked, and corresponds to the line shown on the graph in the Compressor dialog.

In this mode the Compressor effect behaves like a traditional hardware compressor with a "smart" output gain control.

Note that the line from the threshold point to the left of the graph always has a slope of 1.

This is implied by the graph on the effect, but the graph is wrong. After make up gain to 0dB has been applied the slope below the threshold is no longer 1. This is easy to see when you consider the case of silence (-inf dB) - both input and output are silence. Steve the Fiddle
Bill 29Nov09: Steve, try this. Generate chirp 440 Hz, 5 secs, 0.01 to 0.1 (-40 to -20 dB). Duplicate track. Apply Compressor to first track Threshold -26 dB, Ratio 4:1, RMS, make-up gain ON. Apply Compressor to second trach same except make-up gain OFF, then Normalize. They are virtually the same. And the bit at the beginning that started at -40 dB is now at about -20 dB.

When "Compress based on peaks" is checked

The following image shows the input-output characteristic of the Compressor effect when "Compress based on peaks" is checked.

• The red line shows the characteristic when "Make-up gain for 0 dB after compressing" is not checked.
• The orange line shows the characteristic when "Make-up gain for 0 dB after compressing" is checked and the input signal had a maximum peak value of less than 0 dB.

Note that the line from the threshold point to the left of the graph no longer has a slope of 1.

When "Compress based on peaks" is checked the Compressor effect applies "upward compression" to the input. This is very different from the behaviour of a traditional hardware compressor. Note that when the volume of the input is below the threshold the volume is "stretched" to fit the new post-compression range. Sounds above the threshold still receive downward compression. The result is that very quiet sounds are not significantly amplified.

Here's my suggestion: Steve the Fiddle

Then Steve's suggestion made it to the main page unaltered.

Old page text removed.

Steve the Fiddle Compression graph on page updated. I believe that it is finally correct. Perhaps a bit large - I have the original SVG file so can easily provide a resized version if required.