OT: Software of your choice for sampling an instrument

infuzion · 17 Jan 2008

gilwe wrote: What would be your choice when sampling an instrument ?

What do you mean? Sampling electronic (MIDI) instrument, or acoustic? & there is a difference between sampling & editing & mastering...

TomC · 17 Jan 2008

gilwe wrote: What would be your choice when sampling an instrument ?

Soundforge has always sounded a bit gritty to me, but I liked Audition a lot on the other hand 😉

You can use whatever audio edit software you own. You already mentioned Soundforge, so I would use this. If it sounds 'gritty' you did something wrong . Sampling a real instrument isn't as easy as it seems to be!

Tom

electrogear · 18 Jan 2008

Sampling a real instrument isn't as easy as it seems to be!

Certainly not - I'm doing a degree in this, and it's not as easy as slapping a microphone down anywhere and recording it into a HD24 or PC. There are lots of things to take into consideration - levels, ambience, reverberation, EQ, microphone type, etc. Drum kit's, piano's and brass instruments can all be difficult to capture in a realistic way.

I find guitars and vocals pretty easy, but if you get things wrong, the sound may well be different from the sound you expected to hear. I find that a sound that peaks at -3dB with no effects/EQ processing (and if you can help it as little compression as possible) is the best to handle, and effects/eq/more compression/gating can be added later. Obviously there are different mic placement techniques but these are often down to you.

Anyway I'll shut up for now - I'm a little bit pished.....

(being a student I guess you would expect that!) :P

Mikey

infuzion · 18 Jan 2008

electrogear wrote: I find that a sound that peaks at -3dB with no effects/EQ processing (and if you can help it as little compression as possible) is the best to handle, and effects/eq/more compression/gating can be added later.

Good tip! Though I thought you want to go as close as db0 as possible, so you don't lose bit depth... but you'll need a real peak detection program like SSL's X-ISM.

electrogear · 18 Jan 2008

Well, there were certain reasons at the University studios not to go above -3 dB (these being that the desk output and HD24 input were not calibrated with eachother, so a -3dB would equal about -1dB on the HD24 input (I imagine a professional studio would have everything calibrated correctly). However, I always find it best to compromise a tiny little bit of sound quality to ensure that the signal never rises above 0dB - after all, 3dB isn't a real lot in terms of reduction in amplitude from 0dB, and the gain used to bring it back up (if you thought it was needed) would be hardly noticeable, if at all.

stefan · 18 Jan 2008

And, fwiw, if the actual wave file peaks at 0dB, you'll most likely get clipping in the DA conversion due to intersample peaks being above 0dB. You can easily see/understand why if zooming in real close in Audition, so that you can see what the reconstructed waveform looks like between the recorded samples.

infuzion · 18 Jan 2008

stefan wrote: And, fwiw, if the actual wave file peaks at 0dB, you'll most likely get clipping in the DA conversion due to intersample peaks being above 0dB.

Yes, & that is why I recommended SSL's X-ISM, which calculates in those intersample peaks. I'm thinking Live7 does that also, but I have no proof aside from casual observation of their peaks running higher than other meters.

& a -3db ceiling wouldn't be bad if you used 24bit recording.

TomC · 18 Jan 2008

electrogear wrote: Sampling a real instrument isn't as easy as it seems to be!

Certainly not - [...] I find that a sound that peaks at -3dB with no effects/EQ processing (and if you can help it as little compression as possible) is the best to handle, and effects/eq/more compression/gating can be added later.

-3db way to hot. My advice is to sample in 24 bit (so you don't have to care about noise) and sample at about -9 db (with sharp transients (like percussion) reaching -6db.

24 bit has THAT much dynamic range that it won't hurt to lose a bit or two because of less input level. But, if you record too hot you risk a lot of inter-sample peaks in AD and in the further processing. In fact that's one of THE main reasons why digital is(was) known to sound bad compared to analog recording.

Tom

electrogear · 18 Jan 2008

Inter-Sample peaks? Have you ever experienced this? I mean, even if it did happen I'm not sure you'd notice it would you?

TomC · 18 Jan 2008

electrogear wrote: Inter-Sample peaks? Have you ever experienced this? I mean, even if it did happen I'm not sure you'd notice it would you?

Almost every digital summing device (like e.g. the out bus of your DAW) has this problem. You put in a sample less or equal 0db and get out a signal less or equal 0db, but internal processing steps would move it temporarily above 0 db if that'd be possible. It isn't, so the signal gets clipped.

If you'd (like me) mix the songs of bands (and with 'mixing' I mean mixing recorded tracks, not this re-mixing piece of crap that happens a lot these days) you'd learn the hard way HOW much difference in sound quality that means.

You have only so much bit of dynamics to process your audio data, and it's always preferable to run out of bits on the non audible end of your 24 bit. If you mix 2 signals with -3db together the result reaches 0db[1]. If one is just a little bit hotter you already have clipping.

When I mix records I have usually between 20 and 70 tracks to mix, so you can imagine what that means.

Cheers,

Tom

[1]: that's the real life scenario. Of course you can phase reverse two tracks and when summing they cancel instead of add. Also the frequency spectrum of each track must be considered.

infuzion · 19 Jan 2008

TomC wrote: electrogear wrote: Inter-Sample peaks? Have you ever experienced this? I mean, even if it did happen I'm not sure you'd notice it would you?

Almost every digital summing device (like e.g. the out bus of your DAW) has this problem. You put in a sample less or equal 0db and get out a signal less or equal 0db, but internal processing steps would move it temporarily above 0 db if that'd be possible. It isn't, so the signal gets clipped

....If you mix 2 signals with -3db together the result reaches 0db[1]. If one is just a little bit hotter you already have clipping.

If you are lucky enough to have high-end gear (or that free VST I mentioned 2 times), then your meter will compensate for those funky peaks. But it is safer to go lower. & many new dance MP3s are compressed too hot, so there are issues when I DJ mix.

I'm not sure, but I think those "1bit" recorders don't have the inter-sample peak problems?

But record at -9db?!? With all due respect, I prefer to record accurately, aim to -3 peaks, use a limiter to tame those mistake 0db peaks, then use the fader almost all "mixers" come with so when I mix 2 signals together they won't reach 0db. (But I'm into dance music which is compressed & needs to be hot anyway. ) That way, you can still capture the little nuances, room ambiance, etc clearer. Also, I think gilwe is under ideal circumstances; he is taking the time for each sample. If you're recording a vocalist who doesn't know how to use a mic, or a session player who has time for only 1-2 takes, then yea -9 or even lower average peaks.

TomC · 19 Jan 2008

infuzion wrote: But record at -9db?!? With all due respect, I prefer to record accurately, aim to -3 peaks, use a limiter to tame those mistake 0db peaks, then use the fader almost all "mixers" come with so when I mix 2 signals together they won't reach 0db. (But I'm into dance music which is compressed & needs to be hot anyway. ) That way, you can still capture the little nuances, room ambiance, etc clearer. Also, I think gilwe is under ideal circumstances; he is taking the time for each sample. If you're recording a vocalist who doesn't know how to use a mic, or a session player who has time for only 1-2 takes, then yea -9 or even lower average peaks.

Lets do the maths: 24 bit word length covers a dynamic range of 144 db. If we record at -9 db maximum level, we're losing this 9 db, so the recording as only 135 db dynamic range. The final recording will end one a CD which has 16 db, that's 96 db dynamic range. Still plenty of room. It's a myth back from the analog recording days that you should record as hot as possible. True for analog, wrong (within limits of course) for digital.

What you WANT to do (especially if you use a crappy recording chain) is to have a as hot as possible analog signal to use the maximum dynamic range your analog gear can deliver. But before you hit the AD converter you can savely reduce the signal (the noise will be reduced as well, so your overall dynamic range is about the same).

Cheers,

Tom

PS: ...and I won't even mention that these days a fine 24 bit recording will end compressed to death with a final dynamic range of less than 10 db...

PPS: ...on an MP3 player in 128 kbps...

infuzion · 19 Jan 2008

Good info! Why don't you add a tutorial in the WIKI Tom? TomC wrote: PS: ...and I won't even mention that these days a fine 24 bit recording will end compressed to death with a final dynamic range of less than 10 db...

& Dance is like -6db; ARGGG!