Probably a silly dynamics question

[Farview]

Got it, thanks.
Each sample must have a value that translates into volume level. Yes?

Yes. Sample level is a little too small. Remember, the attack time on most compressors starts at a couple ms. That's 441 samples per ms.

It also takes time for a compressor to back the gain off a certain amount. Then it takes time for the compressor to recover the gain after the signal has dropped below the threshold.

I think you know how your volume works, I don't think you know what kind of a complicated volume curve a compressor would create if you could see it.

 

[dobro]

I believe what you say about my understanding.

This is off-topic, but now I'm wondering if it was possible to actually hear a single sample of, say, a flute - would you be able to recognize the sound as that of a flute? Probably not, right? Probably you'd need thousands of samples to be able to recognize the flute timbre. I'm wondering how much of the distinctive sound of an instrument a sample contains.

 

[flatfinger]

Got it, thanks.

So, it would be possible to draw volume automation in a way that would sound like a compressor that's really working?

Well the semantics of "really working " can be confusing because I would normally interperate that as doing some serious gain reduction .( I think that you meant "working like "?)

Usually the way it's done in a DAW (volume automation) results in a gentler , slowish "volume leveler" type of compressor at work. ( Like a rms or opto = slower envelope behaviour. Gentle.

Just get down to the sample level, in other words?

Well if you were to zoom in like mad in a dedicated editor like wave lab I suppose you could get that surgical .... But what I meant was that most folks who draw volume automation in there DAW are going to probably draw w/less resolution and work on a more Macro than micro level. I mean your ear probably isn't going to be able to differentiate if you make a 10db change within 10 milliseconds or something.

One thing this thread has done for me is to make me realize I don't have the first idea about how my software works for even a simple thing like volume/level. Each sample must have a value that translates into volume level. Yes?

The thing to realize here is that each sample is not a signal level( it's called Pulse Code Modulation... it's "coded". An analog signal is continuous ( that is in nature, sound waves) , we humans are the ones who slice it up into descreet samples ( so we can store them ). When we go back from digital sample to analog voltages , the digital to analog converters "reconstruct" things ...... Smoke and mirrors and what not .

Most DAWS are not going to get down to the sample by sample level . You could drop one sample out (completely remove it from the wave , and it can sometimes be inaudable . Clicks are usually caused by more than one sample missing or out of step with each other..............

The important thing to remember is that a sample that you see in a wave editor is'nt "really" an audio signal . The illustrations that are always presented showing a stair-stepped "digitized " sine wave superimposed over it's a nice curvy sine wave counterpart are just for a generalized , conceptual way that's become popular .
This makes it seem like each bit is a "level " it's not really .. every sample uses all bits . Wav editors are great , but as all the guys are always saying , go with what the flaps on the side of your head tell you .




P.S. compressors and distortion


The main example that is given to illustrate the effects of in harmonic distortion that is generated by a compressor is that you must imagine a sine wave ( and 99.9% of musical signals are not simple sine waves; it's just a way to try and imagine whats happening! ).
Wave shaping distortion happens when the gain element is being told by the attack and release settings to change the levels so fast that there is'nt time for the nice curvature of the sine wave to get through the process intact; so it comes out as a square wave. Square waves have more than one harmonic ( sine waves only have one harmonic ... thats why they sound so simple and pure).

So it's called inharmonic distortion because what come out has harmonics that were'nt there ( they were generated by the process of controlling the signal)
use slower attack and release settings ( or automatic release ) and you will probably do less damage to your signals .( especially when first getting into it)




P.S. as odd as it sounds , a little bit of dabbling in synths and sound design can result in some insights when engineering some audio!!

 

[flatfinger]

I believe what you say about my understanding.

This is off-topic, but now I'm wondering if it was possible to actually hear a single sample of, say, a flute - would you be able to recognize the sound as that of a flute? Probably not, right? Probably you'd need thousands of samples to be able to recognize the flute timbre. I'm wondering how much of the distinctive sound of an instrument a sample contains.

A single sample can't sound like anything by itself .

 

[dobro]

Okay, so what's the minimum number of samples you need to be able to capture and recognize the timbre of an instrument?

 

[flatfinger]

Okay, so what's the minimum number of samples you need to be able to capture and recognize the timbre of an instrument?

I don't know !
At least enough to get you 10 -20 milliseconds...

I know the hass effect says that two sounds will merger together ( in our perception that is ) if they are less that 30 or 40 milliseconds apart ..... But I don't know what the minimum amount of time would be to even be audible ......

Anyone ?? How fast is a transient ?? ( not one who's running from the cops !)

 

[dobro]

There's an idiosyncratic definition of an atom of water as being the smallest amount of water that will display all the characteristics of water. It's way more than a chemical molecule of water. I'm wondering about the equivalent in sound.

 

[flatfinger]

There's an idiosyncratic definition of an atom of water as being the smallest amount of water that will display all the characteristics of water. It's way more than a chemical molecule of water. I'm wondering about the equivalent in sound.

I know that sometimes sound is conceptually thought of as particles instead of waves ( just like light), but thats getting way to esoteric for me.

Going to have to take it over to the physics forum now !! What the quark


Where the hell is my Steven Hawkings vocoder when I need it !!!!!!!!!!!!

 

[dobro]

Yeah, if you could post your answer in a Hawkings voice, that'd be brilliant!

 

[Farview]

There's an idiosyncratic definition of an atom of water as being the smallest amount of water that will display all the characteristics of water. It's way more than a chemical molecule of water. I'm wondering about the equivalent in sound.

Well it would depend on the frequency of the note being played. You would need at least one full wave, which will take different amounts of samples depending on the frequency. One 60hz wave is 683 samples long, one 440 hz wave is 93 samples long.

The other variable in here is how short can the tone be for you to recognize the sound.

 

[noisewreck]

I believe what you say about my understanding.

This is off-topic, but now I'm wondering if it was possible to actually hear a single sample of, say, a flute - would you be able to recognize the sound as that of a flute? Probably not, right? Probably you'd need thousands of samples to be able to recognize the flute timbre. I'm wondering how much of the distinctive sound of an instrument a sample contains.

A sample is simply a value representing a level between +1 and -1 at a given point in time (on a CD it lasts 1/44100 of a second). A single sample cannot convey anything such as frequency, amplitude, let alone timbre.

Furthermore, you need much more information than timbre alone to recognize an instrument. For example, I have turned piano samples into long, evolving pads that don't sound like a piano, not because I have fundamentally altered it's frequency content, but because I have taken the attack away, make it swell up slowly for example, and then let it loop seemlessly. Now, because you KNOW what a piano sounds like (relatively sharp attack, followed by continuously decaying sound) the result of my simple manipulation is that it stops sounding like piano altogether.

I'll try to post an example in the next few days. Won't happen tonight though, I'll likely be at work still until around 1am.

An analogy I can give for sample is to look at your computer monitor with a magnifying glass just so you can only see one pixel. Can you tell what the rest of the image is by just looking at the pixel itself?

 

[noisewreck]

I don't know !
At least enough to get you 10 -20 milliseconds...

I know the hass effect says that two sounds will merger together ( in our perception that is ) if they are less that 30 or 40 milliseconds apart ..... But I don't know what the minimum amount of time would be to even be audible ......

Anyone ?? How fast is a transient ?? ( not one who's running from the cops !)

The real answer is, it depends on the frequency. Low frequency sounds require more time for it to complete one cycle (i.e start from 0, go to +1, go back down to 0, go down to -1, and go back up to 0 again)... or in degrees, to start from 0, go to 90, then 180 (which is the same as 0) then 270 (or -90 if you will) and back to 360 (same as 0).

Now imagine you draw this shape on a piece of rubber. Low frequency sound would be like stretching that rubber which will stretch the drawing itself, which means it will be longer. Higher frequency sounds are the opposite, you'd be shrinking the rubber, thus making the drawing shorter. This is why high frequencies for example are also referred to "short wave"... usually in the radio frequency realm.

1 Herz (Hz) is defined as the frequency of sound that makes full cycle in 1 second. So, 50Hz will take 1/50th of a second, 200Hz will take 1/200th of a second, 3kHz will take 1/3000th of a second to complete one full circle and so on.

So, once you start looking at this, and you take simple wav file recorded at 44.1kHz, you realise that 1 sample cannot accurately represent any frequency in the audible spectrum

 

[SouthSIDE Glen]

Just to add a couple of sheckels to the pot on this really cool discussion, noisewreck is right in that the timbre of an instrument depends greatly on how you affect it's envelope. Think of a wah pedal on a guitar. Now we all know what wah pedals sound like now, but try to imagine hearing one for the very first time with no prior knowledge. How long would one have to sustain with the wah open before you recognized the sound as coming from a vibrating guitar string? It's difficult to put a number on that, but I think we'd probably all agree that the longer it was held open, the more likely we'd wind up making the mental connection, even though the overall envelope is still wahed.

At least part of the reason fir this is because one of the properties within the timbre of an instrument is the complexity of the sound it creates, which may require x number of cycles of many of the component frequencies generated to "firm up" the overall timbre picture in our heads. I'm not positive of this, but my edumacated guess is that one could recognize a middle C sine wave test tone in fewer cycles than it would take for a mid-C vibrating piano string (which is actually multiple vibrating strings plus vibrating sound board and shell, etc.), let alone something really dirty like a mid-C played on a saxophone. The sound modulates and changes over time with all these different vibrations and resonances interacting together, so theoretically it should (I think) take more time for us to build the picture.

As far as transients, in practice we often can hear a single-sample "click" or "thump" of high enough amplitude over our loudspeakers, but the ballistics of the loudspeaker itself will slew that transient and "smear" it in time, which lengthens the effect into a full cycle of movement (with another cycle or two of decaying amplitude probably also). It is, in a weird sort of way, not the sample we are hearing so much as the loudspeaker itself. If we had a theoretically perfect loudspeaker - AFIAK, impossible; at least at this resolution - we likely wouldn't hear the single-sample transient as a sound, because it would be a pulse of zero frequency (zero cycles).

Which also brings up the point that even repeated clicks or pulses that give us an actually frequency that could be measured, we still have the lower limits of human hearing where it won't be heard as a tone until we get well up above 20 of those cycles per second.

G.

 

[Farview]

Check this out https://www.youtube.com/watch?v=7IRe6HUo7Pc

Somewhere around 7 minutes, you will see drums slowed down. Notice that the tranient actually seems to take a long time. The stick pushes into the drum before it bounces back.

This footage is taken at 5000 frames per second. You are sampling at going on 9 times faster (44,100 times per second), so the 'transient' will be spread out over several samples and it will take another bunch of samples to represent one cycle of the drum hit.

'Transient' and 'instant' are anything but at these sample rates.

 

[mixsit]

So, in light of all that.. and kicking this over a bit, I keep coming back to maybe I asked the wrong question, and what seems to me the common difference in this idea and a 'compressor. Throw out 'timing?
If you scale each sample, no attack, no release, no 'musical best guess ramping the gain up and down on samples that maybe didn't need their individual relationships changed, wouldn't it (could it?) sound totally natural? or why not..?

(Maybe I should go back and reread my very good Valley People manual..

 

[LemonTree]

I thought when you scaled back the sound without attack or release it was done with a TRIM knob

 

[mixsit]

I thought when you scaled back the sound without attack or release it was done with a TRIM knob

You just dropped in at the end right? It's about ratio. Change the dynamic range..

 

[oretez]

"Okay, so what's the minimum number of samples you need to be able to capture and recognize the timbre of an instrument?" Dobro

limiting to coherent timbers (i.e. not noise) and very roughly speaking

1.5 X frequency of fundamental

but what any individual can interpret depends on both internal and ambient variables as well

(you question is actually compound: 1. mathematical limit necessary to model a specific 'driver'. 2. limits of human 'resonator' to interpret that driver. Then of course the two are typically, unless 'driver' is jacked straight into neural net, linked via an intermediate 'ambient' stage. Frequently what humans detect is not the thing itself but a change (we hear because of changes induced by a pressure wave). Roughly speaking most humans have no trouble detecting 3 dB changes in S/N, but generally speaking the higher the fundamental the more energy the driver has to supply for the changes to be recognized. In any case understanding response to the question requires some awareness of the physics of pressure waves, how those waves interact in 3D environment (4D if you count time) and human psycho acoustics . . . a slightly practical response suggests, for most instruments, a number between 1.5Xfundamental and 14000 samples (@44.1 kHz sampling rate) (off the top of my head)

 

[SouthSIDE Glen]

So, in light of all that.. and kicking this over a bit, I keep coming back to maybe I asked the wrong question, and what seems to me the common difference in this idea and a 'compressor. Throw out 'timing?
If you scale each sample, no attack, no release, no 'musical best guess ramping the gain up and down on samples that maybe didn't need their individual relationships changed, wouldn't it (could it?) sound totally natural? or why not..?)

That's actually an interesting question that, after thinking about it, results in a perhaps surprisingly obvious answer.

It really is a simple volume change, but is volume change by multiplication/division, not addition/subtraction. If you apply an instant 2:1 downward compression across the board, you are, in effect saying cut the overall volume by one half. This does not mean subtracting a specific number of dB like with a trim pad, but rather a halving of overall volume, whatever that volume may be at any given time. It is a true reduction in the amount of volume gain - hence the name "gain reduction" on the ratio control!!

Such a gain reduction isn't changing the true character of the waveform other than the overall volume, and theoretically should sound the same. There'd be a quieter overall volume, of course, but one should be able to cancel that out by applying makeup gain in the output stage.

So, assuming a "perfect" compression algorithm or circuit that added no noise, coloration or artifacting, an even, equal overall change in compression ratio across the board, when adjusted for playback volume should sound no different than the original, because the compression, when applied that way, really is nothing more than an evenly applied difference in signal gain.

G.

 

[Farview]

But he said throw out attack and release, not threshold.

If you leave everything below the threshold alone and only affect the wave above the threshold, you will be changing the rise and slope of the wave. If you did it perfectly, it would probably sound like crap because of the distortion of the wave.