[linux-audio-user] Re: producing a drum sample library for hydrogen

R Parker rtp405 at yahoo.com
Fri Jun 18 10:37:26 EDT 2004


Hi Juhana,

PROGRESS UPDATE
I did snare drum test prints after staring this thread
but a mono playback bug is preventing me from forging
ahead.

Beginning with loudest and working towards quietest,
we recorded nine volumes for one tone. I'll probably
produce three or four tones

--- Juhana Sadeharju <kouhia at nic.funet.fi> wrote:
> >From: R Parker <rtp405 at yahoo.com>
> >
> >I have been discussing the design of a drum sample
> >library for Hydrogen with Comix (Hydrogen author),
> 
> Such a sample set would be useful outside of
> Hydrogen as well.
> How about "public domain" license? We just have to
> make sure
> people can get the originals if they wish. Is there
> GPL equivalent
> for samples?

Someone suggested a specific license, I haven't had
time to look at it yet but will.

> Here are my non-professional tips:
> 
> Tip 1: Listening to existing libraries may not help.
> It could
> be that most drum samples are not done as good as
> possible.
> 
> Tip 2: Provide as many snare (say) samples as
> possible. If you
> want create a limited preset for Hydrogen, you can
> do that later
> but leave an option for us others who want a
> randomizing
> element to the drums. Press record and hit the snare
> for an hour
> in various ways.

To begin with we'll have either 27 or 36 volumes for
three or four tones; center hit, rim shot, mute. From
these mono samples, I'll produce a series of snares
with 1 in, 2 out reverb sounds. The first series of
tests I ran used a dozen or more reverbs per sound and
level. It's a alot of work and I might draw a line
somewhere.

> Tip 3: Group the samples so that all samples from
> one drum kit
> are together.

Organization is mandatory. Nothing worse than trying
to find what you want:
1. snare_M_rimshot_10
2. snare_M_rimshot_09
3.
...
9 snare_M_rimshot_01
_M_ = Mono
10 = Loudest

1. snare_S_H-lrg_rimshot_10
...
9. snare_S_H-lrg_rimshot_01
S = stereo
H-lrg = Hall large
10 = Loudest

We'll need a legend of sorts that explains naming
conventions.

> Tip 4: No effects whatsoever at recording time.
> Reverbs, gates etc
> can be added later, but cannot be removed if they
> already are there.

Right, with an expception being the stereo snare drums
where the whole point is to design sounds.

> Tip 5: If you record stereo, make the other channel
> so that it can
> be played mono.

I probably won't use this strategy. I'll mic for
stereo and then do another session for mono. One
reason is because my mono signal path is much higher
quality than my stereo path. :)

 Channels should not be mixed for
> obtaining the mono
> version. First mic is for mono and the second mic is
> to get a different
> perpective from the same sound. The first mic should
> be placed to best
> position. Snare is often recorded both from top and
> bottom (where the
> spring is heard best). Stereo effects can be done
> with these later;
> the mics does not need to be in a true stereo
> configuration.
> 
> Tip 6: When two close mics are used to record a
> drum, check
> any phasiness effects by mixing the mics. Fix it by
> moving the
> second microphone. It may not be possible to fix it
> later.

I don't have phase problems. Fixing any audio problems
after the fact is always sucks. It's never necesarry
to do that with my work.

> Tip 7: The overall stereo mic could be in a
> human-ear-stereo
> configuration above the drum kit. Those mics should
> be kept in fixed
> position during all recordings. So, for each drum we
> would get both
> the mono, the second mic, and overall stereo
> recording --- four
> channels in total. That should give freedom at mix
> time.
> 
> Tip 8: When samples are trimmed, don't cut decays
> too early.
> I use spectrum view in editing long decays because
> the spectrum
> display can see throught the waveform noises.
> Listening to the
> sample helps but not many editors have +64 dB output
> gains (Audacity
> has ok +36 dB max output gain).

If you can't hear anything then all useful audio is
gone. I've been allowing the tail to fade completely
out of the audible level range. I think long tails are
the way to go because size is of little consequence
and subsequent triggers in a sequence will simply play
over any tails.

 Leave good space to
> the start as well;
> trigger point can always be adjusted later in
> software.

In my tests, I trimmed heads to be tight. I can't
think of any situation where there's a benefit to
leaving space in front of the sample. If there isn't a
benefit to space in the head and I'm already editing
the files I should just get it right so nobody has to
deal with it again. Of course this is a point for
which I could be entirely ignorant.

Please provide specific examples that clarify why room
in the head is the most useful strategy. 

> >mic pattern. When sequencing ride cymbals they need
> to
> >bleed from the listeners Hard left side across the
> >stereo field and into the Hard right side. If we
> don't
> >treat cymbals as stereo images they could fail to
> >sound like a live player when sequenced.
> 
> Why? Do you sit on drummer's lap in concerts?
> Does your drummer's cymbal walk from left to right
> when hit? :-)
> 
> As far as I know the cymbal sound (one cymbal) comes
> from one
> fixed position. If I stand at 50 feet from the
> drummer, the
> sound is practically mono. Any left to right
> movement coming from
> speakers is just an effect. Effects can be done
> later. Though, use
> of two mics makes panning effect less dull.

When you're fifty feet from source, you're listening
to a room whether it's the Ordway Theatre, Lee's
Liquor Lounge or the Great Outdoors it is for our
purposes a room. Imaging is the artificial
construction of a room and in our case done with Left
and Right sides. This is just tried and true standard
stuff. There's nothing to debate.

A second version of the library could be for surround
where anything is open for debate. Incidentally, I
have begun to research surround, not because of market
demand because I have none. I'd simply love to mix
surround. To the best of our abilities, we designed
our mastering room so it could be certified for
surround work. That's alot simpler than it sounds. In
addition to tried and true accoustical design methods,
you simply have a minimum total cubic footage
requirement. I think we are short about 100 cubic
feet. Not that it matters to me because I'd never pay
for a certificate.

> >from 1 in 2 out reverbs. In addition to dry mono
> snare
> >drums, we intend to produce a number of reverb
> >processed (wet) stereo samples.
> 
> I can help here and apply some of my own reverbs.

What reverb instruments do you have?

I've got Lexicon LXP-15, TC Works and Waves.

> can also
> help in trimming and in labeling the samples from
> the raw
> recordings (if you don't have time). I can also
> provide
> space from here ftp.funet.fi.

At one time, I was gonna ask if someone could do the
trimming and labeling. I since concluded it would be
easiest to do it myself. Now you've got me thinking
and I appreciate the offer. When production starts
I'll know if it's reasonable to ask that of you.
 
> >The drums will sound great because Bill,
> >Bobby and I know what we're doing. If we don't know
> >what we're doing someone should shoot us--we've got
> >about 90 years of combined experienced with playing
> >and recording drums.
> 
> I have recorded zero seconds of drums, but giving
> "tips"
> won't harm anyone. If I'm mistagen, I'm sure you see
> it instantly.

Your feedback is definitely appreciated. I seldomly
sequence "keeper" drums and have never produced a
sound library that's intended to be of some
consequence. If 20 years of experience has taught me
anything it's to listen and carefully interpret what a
client wants.

When the mono playback bug is fixed, I'll start work
again.

ron

> Juhana
> 



		
__________________________________
Do you Yahoo!?
Yahoo! Mail - Helps protect you from nasty viruses.
http://promotions.yahoo.com/new_mail



More information about the Linux-audio-user mailing list