Attack Sustain Decay Release

Dear all,

I’m new to this forum! :slight_smile:

In this period I’m recording and analizing drum sounds. I would like to know if I can derive Attack, Sustain and decay parameters of a drum shot with Audacity.
If not, can you suggest any other audio software to do it?

best wishes from sunny Italy! :smiley:

You mean in some way other than looking at the waveform?

Hi Steve,

thanks for the replay. I mean that I would like to derive from the waveform the attack and decay time in milliseconds for e.g. 20, 30, or 40 db reduction.
Or in other words in a quantitative way. Any suggestion?

best,

M.

If you select “length”, (dot the circle), the length of the highlighted portion is displayed (in seconds / milliseconds /samples) …
length of (dark blue) highlighted section in milliseconds on Audacity.png
You can select to display the waveform in dB if that helps …
Audacity option to display waveform in dB.png

I don’t know of any analysis tools that will produce numerical outputs, but you can see on this image (a cowbell hit) that the attack is 86 samples duration. The sample rate of this recording is 44100 Hz, so the attack can be calculated as 86/44100 = 0.00195 seconds = 1.95 ms
(Click on image to expand)
attack.png
Drum hits typically have a very sharp attack and a less sharp, roughly exponential decay. This does not work quite the same way as an ASDR envelope which is typically used by synthesizers. If you are looking to synthesize drum sounds, you will need to draw your own approximation as in this example:
ADSR.png

Dear Trebor and Steve,

thanks a lot! I think I’m really close to what I want! :smiley:

What I’m doing is to compare drum sounds e.g. tom-toms with and without dampening rings and see how sustain varies.

The last figure by steve is exactly what I want!

Three questions:

*How did you derive the different lines? Only drawn by hand?

*What do you have on the y axis?

*What would be in this example the duration of the sustain?

thanks a lot!!!

M.

Yes, drawn by hand.

The y axis is a linear scale +/- 1.0 (the default “waveform” view).
If you want the read-out in dB, click on the track name and select “Waveform dB”.

The relationship between the linear scale and dB (approximately)
+/- 1.0 = 0dB
+/- 0.5 = -6dB
+/- 0.25 = -12dB
+/- 0.125 = -18dB
+/- 0.0625 = -24dB

If you’re talking about ADSR envelopes, it is zero, but that’s not very useful for your purposes.
For measuring the sustain of a drum, it would be more useful to decide on what level you are going to consider to be “silence” (if you had an infinitely sensitive microphone and no other noise, the drum would probably ring forever, but at a very very very low level).

You will probably find it most convenient to use the “Silence Finder” tool (in the “Analyze” menu) to mark the point at which the sound becomes “silent”.
Let’s say that you decide that -24dB is close enough to silence, use the Silence Finder tool and set the first slider to -24 and the other 2 sliders to minimum.

It would probably be interesting compare the “sustain” (how long it rings) with/without damping rings using different “threshold” settings (the first slider), that way you will be able to see how much the rings affect the decay from peak to -6/-12/-18/-24/-48 dB which will show how the damping effect varies over time/amplitude. I would expect the damping effect from peak to -24dB to be a much greater difference than the damping effect from peak to -6dB.

I would be interested to know how you get on with this - your results/conclusions.

Dear Steve,

thanks a lot, this was exactly what I was looking fore! You deserve a pint of beer!! :slight_smile:

I think that what you are calculating with the “silence finder” tool is a kind of decay time. I processed my files with the silence finder with db levels of 20, 30 and 40 db as suggested in Toulson et al. 2008 (The perception and importance of drum tuning in live performance and music production). Toulson reports a reduction in decay time from 752 ms to 199 ms after adding a damper ring to a 12" tom-tom. I made a similar experiment with a 12" tom-tom. I compared the sound of the drum with both heads at the same pitch (i.e. 196 Hz, G3) and with the batter head at much higher pitch (i.e. 292 Hz, D4). Here are the results:

Same tension: 20db:430, 30db: 730, 40db: 1450 ms
High pitch: 20db: 320, 30db: 690, 40db: 1130 ms

So the drum with heads tuned to the same pitch have a higher decay time as I would have expected.

I have a couple of questions:

  1. I decided to normalize the audio files to 0 db before analyzing them to avoid differences coming from the possible striking power. Do you think it’s right?

  2. How can I determine if in the drum sound there are beats? I saw that in the “analyze” menu there is a “beat finder” tool. Have you ever tried to use it?

best wishes and thanks again,

M.

Thank you, I’ll have one later :slight_smile:

Yes definitely or else your measurements will be all over the place, but I think that you should also try to hit the drum with as close to equal force as you can so that you are comparing like with like.

Try repeating the experiment, but instead of changing the skin tension, compare the measurements that you get with a soft hit compared with a hard hit. I would not be surprised if there was a measurable difference in the decay rate (even after normalising).

The Beat finder tool is for detecting sudden changes in amplitude and it places a marker at each “beat” that is found.
Try generating a click track (from the “Generate” menu), or play a slow rhythm on a tom and record it, then select the entire track and apply the Beat finder tool. It works quite well on single drum tracks, but becomes less accurate with full music mixes where the “beat” (rhythm) is less well defined.

Hi Steve,

I realized that it is fundamental to normalize the tracks to 0 before applying the “silence finder”, I compared normalized and non normalized tracks and got quite different results. Moreover for my 12 in tom-tom it seems more efficient to use a db reduction of 50 db.

I made a quick experiment.
Decay time without muffling ring: 2260 ms
Decay time with muffling ring: 2120 ms

If I will find the time I will test the striking force!

With this technique I could test e.g. head type, head combination, shell type, tom mounting system effects on decay time!! But it would take a life!! :laughing:

Concerning beats I meant “frequency beats” that cause the sound to a kind of pulse or vibration. Do you know if there is any technique to test if a drum sound is affected by beats? Of course I can hear it but I want an objective measurement of it!! :smiley:

best wishes,

M.

I thought that is probably what you meant, though it will be very difficult to measure using Audacity. The tool for the job is “Plot Spectrum” (Analyze menu), but even with the “Size” setting at maximum (16384) the low frequency resolution is still fairly poor.
You may be able to detect low harmonics (beats) if you “rectify” the waveform and then remove frequencies below your expected beat frequencies. For example, if you are looking for beats in the 15Hz range, then use the following code in the “Nyquist Prompt” effect from the “Effect” menu (cut and paste this code into the Nyquist Prompt text box and apply)

(highpass8 (snd-abs s) 8)

(snd-abs s) makes all the signal positive going (it “rectifies” it - similar to demodulating an AM radio signal)
(highpass8 8) applies a steep high pass filter at 8 Hz.

You may then be able to pick out beat frequencies, though I doubt that you will be able to make many meaningful measurements. It’s much easier to pick out beat frequencies on continuous simple tones than on percussive complex sounds.