h a l f b a k e r y
Baker Street Irregulars
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
or get an account
The problem with digital recordings is, that when the analog sound is too low, it will be cut out during quantization. If the recording system gives unevenly a higher sample rate to a lower analog sound, it can be used to amplify low voices without clipping.
My first "drum sequencer".
[doctorremulac3, Jan 26 2019]
Why John Bonham is better than a drum machine.
There's a couple of reasons. [doctorremulac3, Jan 27 2019]
Yea, you could program a machine to do this...
... but it woudln't be particulary linteresting. Humans performing are more interesting than machines doing what they're told to do. [doctorremulac3, Jan 28 2019]
My key to the music biz.
More like the shaped charge I strapped to that seemingly impenetrable foot thick steel door between a young man of limited financial means and the music business. [doctorremulac3, Jan 28 2019]
Lots of floating point ADC stuff
// Another might be to make a direct floating point analog to digital converter, which I've never heard of being done. // [notexactly, Jan 29 2019]
||But if all of those higher-rate samples are below the
threshold, won't you still clip?
||There is one way to modify digital sampling so that gives it
almost the same dynamic range as analog, even down to
signals well below 1-bit strength. What you need to do is to
add random noise to the signal, equivalent to about 1/2 of
one bit on average. Then your less-than-1-bit signal adds
statistically to the noise, with the result that it's
represented more or less proportionately.
||The downside of this is that you then have noise on your
recording. However, for very weak periodic signals (like
singing, or vowels), the brain is good at picking them out
from the noise, and it all works out OK.
||[+] but I'm not sure I completely grok it.
||I'm a be confused, what's the difference between analog
being too low and "cutting out" and a digital snapshot of
part of the wave not having enough information to put into
||Wouldn't you just put a compressor on the signal, an
automatically acting "robot arm on the volume" to achieve
what you want?
||Seems like a higher sample rate of something too low to
hear would just create a clearer, more perfect
representation of something too low to hear.
||The idea of a continuously variable dynamic sampling rate
is intriguing though. [+]
||//Wouldn't you just put a compressor on the signal// Yes,
||//more perfect representation of something too low to hear//
That's what I thought.
||Adding low-level noise in order to bring weak signals above
threshold is fairly well-established as a method. In fact, the
human visual system probably does it, to improve night-time
||I do not mean bass with a low voice, but more like silent. Let's say that we record a cymbal. First it sounds to resonate nicely. When the sound comes more silent, the high part of the vibration can be heard, but the low vibration is chopped off because it is below the quantization limit. After that there is no sound at all, even if the cymbal is still ringing. If there would be a higher sample rate for more silent sound, it would not lose the information that is in analog sound.
||The higher sample rate won't achieve anything - you're just
taking more samples-per-second of a sound that's too weak to
register. You may be confusing amplitude and frequency.
The cymbal maintains roughly the same frequency (or, rather,
a complex mix of frequencies) as it decays, but their
||Yea, I think you're over complicating digitization. It's just a
lot of little snapshots of information about something that's
already perceivable so it can be re-created.
||And "snapshots" isn't really accurate either. Lot of info in a
snapshot, digitization only has two kinds of information, on or
off, but there's a shit ton of this on or off info that represents
what you've recorded.
||...sorry, "metric" shit ton.
||Lots of Europeans on this site. Need to speak the lingo.
||But again, the idea of a dynamic, continuously variable
sampling rate is kind of interesting. Thought you needed
32 bits for
that sound? Maybe that vowel sounds fine with 8 bits. Then
kick in the higher rate for sounds that need it.
||Is that a need for that? This is the Halfbakery. We don't do
||//Lots of Europeans on this site. Need to speak the lingo.//
No no, it's fine, really. Most of us in England still remember
the old system (or Imperial, as it's known), and it's only a
problem at the interface between the US and the rest of the
planet. Good to know it's still working for you. And on the
plus side, trade between the US and Myanmar is much easier
||Now, where'd I leave that 17/43rds spanner...damn, maybe I
left it out in the 91.44cm.
||I read recently about this, how one bit can be used to switch to a different ceiling. So quiet passages have the extra bit "on" to indicate that they are sampled at a higher amplification. When the signal becomes stronger, the extra bit is switched "off" giving half the resolution but double the headroom. I think this was many decades ago, before audiophiles realised that the signal had a digital stage. Perhaps to do with FM radio? Can't remember.
||//higher sample rate won't achieve anything - you're just taking more samples-per-second of a sound that's too weak to register//
||True, If you can't hear it in analog you won't hear it in digital.
||Analog is all information. Digital is on/off informational parts.
Theoretically there are more sounds in analog than digital.
A higher rate is always going to be more information which the brain can possibly interpolate more sounds.
||Maybe there is a prime imperial function of sampling that helps specific sounds with loss of others.
||Finally we're getting into a subject I actually know a little
about. I used to be in the music biz in my youth at the
time of transition between analog and digital. I've
mentioned that I never got a college or even high school
diploma but I do have a wall full of gold and platinum
albums so I could probably get a job teaching something
music related in college, like "How to get a wall full of
gold and platinum albums 1.01". (Lesson 1 would be "Give
it up, it ain't gonna happen.")
||Anyway, digital recording on the production side got a bad
rap when it first came out as sounding "harsh" because
people didn't know
how to use it. It recorded the object making the noise
very clearly, whereas tape would soften, crush and smash
everything in a very pleasing way. It's been called tape
compression among other things. Also, the analog circuits
involved in these very expensive mixing boards and
equipment (hundreds of thousands of dollars for the
equipment in my studio) would caress, massage and
lovingly process the sounds in a very pleasant way as
well. The old "analog sound".
||Now music in real life isn't pleasant to listen to. Stand
next to a drum set and tell me that cymbal in your ear is
music. It's pain. Analog can soothe that beast, digital
gives you everything that's actually happening. So you can
still make things sound pleasant with digital, you're just
going to have to develop the skill to deal with all those
transients and spikes that were previously smoothed over
by the old analog recording gear. Analog tends to
naturally sound good, digital needs to be controlled by
somebody with a very good ear.
||Anyway, big problem with digital recording and
sequencing is ear fatigue. Your ears get tired of being
beat up on the sonic side, and on the musical side, the
perfect music repetition like sequenced drum tones the
same way it processes wall paper with repeating pictures
of flowers on it. It doesn't go "Flower, flower, flower,
flower..." the mind goes: "Shitload of flowers." and moves
on. This is the problem sometimes with very sequenced
perfectly timed and especially repeating stuff, the mind
processes it and gets bored of it so you need
to take drugs on the dance floor to tolerate it. A drum set
played by a human, a skilled human anyway, moves with
the music and gives you little changes and variations that
keep the brain interested rather than being a glorified
||That's why the old stuff has survived the decades. It's
sloppy, it's imperfect, it's human.
||By the way, go see the Queen movie. I thought it was
beautiful. I also personally know a guy featured in it.
||It's not name dropping if you don't drop the name.
||<feeling reverence> Yeah, ", nice. </feeling reverence>
||//I wonder if anyone is writing 'sloppy' synthesizer
software. If they aren't, they are now.//
||Well as far as sound processing, there's a person I
know who started a digital plugin company called
The Bomb Factory that's now owned by Avid that
makes stuff that completely "analogizes" (new
word) stuff. These things are incredible. He
carefully analyzed tons classic analog rack mount
processing gear and made amazing digital clones of
them. So you can purchase a plugin for a couple of
hundred bucks that completely replicates the
sound of a five thousand dollar classic tube
compressor for instance. There's lots of these
sound processing plugins of course but his actually
||But I'm not sure how you'd "humanize" sequenced
drum or sequencer tracks other than backing off
on the quantization. (the thing that neatens up
sloppy playing) Of course you can just turn
quantization off but for some reason nobody does
that. I never did. Guess it's like a gal wearing
makeup. Yea, it's not natural, but if you look
better with it, why not? I quantized all my
percussion stuff except drum travels that sound
like fire crackers going off if they're sequenced.
||Plus playing instruments in real time and making
them sound good takes talent and practice. John
Bonham is the greatest drummer who's ever lived,
he's the one you'd want to study to find out how to
properly "humanize" or "sloppyize" sequenced stuff.
That would actually be a fascinating study. You'd
write some kind of program to see when and how
he led or lagged hits compared to a perfect
timeline. Of course it's not just timing you'd study,
||Been so long since I've been in the biz, wonder if
anybody is doing anything like that.
||Uh oh, just realized John Bonham, my number one
musical hero, was English. Oh well, gotta give
where praise is due.
||This is a tricky area, because for a beat to groove,
it really does have to be "in the pocket" as we used
||Here's an interesting experiment:
||I actually have a pretty cool collection of classic
rock songs with the individual tracks separate. Let
Zeppelin and Queen to be specific. These are
probably available on line for everybody these days
but back when this was like having nude pictures
of the pope. (As far as rarity, not entertainment
||So take one of these drum tracks of Bonzo and
quantize it. Real simple to do, you hit a button and
suddenly it's exact 16th notes, or dotted 16th
swing notes or whatever it is. Before you did this
you could just visually zoom in on a grid seeing
clearly where he's off a SMPTE based timeline, if at
||Does it feel less... I don't know, John Bonhamy?
||Anyway, take that info and apply it, as necessary,
to what I used to call teletype beats.
||By the way, my dad used to have one of the
machines shown in the link. I'd write "beats" that
imprinted on a strip of paper with holes in it to
play beats. (link)
||//randomize the generation of beats or notes to create an
average bpm rather than strict spacing, although that
random variation would only need to be fractions of one
||I don't know if randomization would help - I think it would
just be subtly annoying. I know nothing about either
drummers or musicians, but I do know that when [a good]
one of them is a 64th of a beat late or early, it's for a
reason. I suspect that knowing instinctively whether you
should be 1/64th early, 1/64th late, or bang on, is what
makes the difference between a competent performer and
a good one.
||NEWS FLASH! Ok, just spoke the my buddy who
gave me all those raw drum tracks years ago and
talked to him about what makes drums "breath" or
"feel natural" and told him my idea to quantize
those raw tracks to see how it was effected.
||So what was the difference? He said, not much.
But he hit me with something that sums it up so
perfectly. What makes John Bonham's drums
amazing? You can listen to it without any other
instruments, and it's beautiful.
||I hate to sound artsy fartsy but he summed it up.
This is art. You can have a computer make a
painting, or a poem, or a song but the appeal of
great art is that.. sorry for this... there's
something appealing about it. It may be the way
singer gasps for breath between notes on a
particulary passionate part, the way a writer
all the character's stories together at the
culmination of the novel, or the way an artist like
Escher or Dali makes you say "How the hell did he
do that? ".
||It's not technical. It's magical. So again, what
makes John Bonham great isn't the mathematics of
where the notes go, it's the way he dances around
the song, the way he listens to Robert Plant
pausing between lines and places punctuating
drum travels that take that line Robert Plant just
sent a chill up your spine with and charges it with
a thousand volts in case you didn't get the
message, and whatever that message is, it's heavy.
I have no idea what they're talking about, but it
hits like a god damn lightening bolt. It's
reaching right down into my soul and pulling
someting out of me I didn't know I had. I don't even
know what the hell it is but I want to feel it again.
||Sorry, got into a bit of a writing frenzy there.
||I can type very fast sometimes.
||ANyioen can tpep faast. Doingo ti accuaratelt is waht coutns.
||OK, somebody made a great video that explains why
Bonham's drumming is better than a drum machine.
(Warning, for the
drumming obsessed only.) See link.
||Having not looked at video and taking all energies into the equation , a computer would be modulating the essential computer essence, abhorent to a human and probably nature herself, and Bonham would be modulating the essence of a star, an aura loved and worshiped by humans.
||Just a left field thought, is digital communication modulating a bad essence of choppy conversation?
||Well, I respect his drumming, not him.
||I started to relate the two stories I had heard first
hand from two different people who saw some
pretty aggressive behavior from this guy but
decided to leave the dead to rest in peace. I'll just
comment on one because it was widely publicized,
his boys jumped some workers at a stadium they
were playing at. My manager (years later
obviously) was the concert promoter at the time
and banned Led Zeppelin from ever playing in
Northern California again after that incident. I just
heard the gory details from somebody who was in
the room. Don't want to say too much negative
about the dead but he was certainly a hard case. I
respect tough guys who fight face to face when
necessary but jumping somebody isn't a tough guy
move. I'll leave it at that.
||It's his talent, not him I respect.
||Hey, the Soviets put the first man in space. Sometimes
when evaluating stuff you have to separate the act from
||Sometimes the full story isn't known and from the outside looks really bad. Hopefully karma was on the case.
||From the video, Bonham sure has that maestro subtle control with respect to drumming. I wonder if there is a physiological ratio of nerve endings to muscle fibres to give a possible unit of subtle motor control with respect to fast action. A muscle puppetry ability score so to speak. And I'm back to sampling and quantization.
||I believe somebody who put the effort and the soul into a
drum machine could make drum parts that could make
you weep, but we tend to just tell it what to do and add
music on top. It's also very easy to do, anybody can do it.
To make a dumb instrument like a bunch of drums sound
like music takes a little skill, and I believe with the skill
to make the trickier instrument make music comes the
||I'm not sure if that makes any sense at all.
||I'll read it tomorrow and see if it does.
||How I read it, At the top end both would take an expert skill it's just that the actual instrument is a longer, steeper initial learning curve which gives more people pushing buttons than competent physical instrument users. And that the extra learning with the subtle physical interaction, musically gives the extra entertainment special source.
||Yes, you said it better than I did.
||So somebody might obviously do something
amazing with any musical tool, including a drum
sequencer, but put it this way.
If you had a robot and said "Robot, write some
amazing music and put my name on it." it might be
compared with somebody who put their heart and
soul into the work and that might come out in the
People hearing the robot written piece might say
"Well, I can tell a Maestrolux 9000 piece when I
hear it, but did you hear what this person did by
singing while tapping on the beat on his chest?
did he make that so interesting?"
||Easier tools might also make music available to
less motivated people, and less motivated people
might not necessarily make better music. Or better
anything for that matter.
||Another good example might be for singing. You
can program robot voices to sing, or you could
listen to the guy in the link doing this with no
music. (see link)
||Humans put emotion, whatever that is, into the
music. We might ask ourselves "What's the human
story behind that?" something we wouldn't ask with
a machine. Unless somebody did something with a
machine to prove that statement wrong which is
entirely possible. How? Well, not by programming
in a beat that everybody's heard a thousand times
before and combining it with video images of a
woman singing scales with generic lyrics for the
sole purpose of selling records.
||Has anybody ever played a drum machine with
drumsticks? Just turned it off and played it as an inert
plastic box being used as a percussion instrument?
||That would be an interesting statement. Guy
pushes play on the drum machine, it starts
churning out some generic beat, he un-plugs it and
whips out a couple of drum sticks and proceeds to
do something amazing with it. The different knobs,
the LED screen, the sides would all probably make
slightly varied percussion sounds. A row of buttons could
be swept across like a guiro for instance.
||Speaking of classic old gear, it just occurred to me the
EMU SP-12, the first commercially available sampling
drum machine either came pre-loaded with John Bonmam
drum hits from "When The Levy Break" or I bought that as
a separate package at the same time I purchased the
||You actually had a robot John Bonham that would play
any beat that you wanted.
||I looked up a picture of this machine and got a rush of
nostalgia that surprised me. I knew I couldn't compete
with the big dogs of the music biz using their equipment
so I made sure I was the first to use this cutting edge stuff
to bust my way in. This thing was like a fucking magic
carpet to me. (link)
||Every month I'd sit and ready my copy of MIX Magazine
from the back to the front to see the ads for the newest
musical equipment. Oddly enough I never read the
interviews with big producers and successful engineers. I
was there for the news on the state of the art from the
technical side. I knew (or thought I knew) nothing they had
to say would apply to me anyway. Plus, looking back, I was
probably jealous and hated them for their big mansions
and expensive sports cars while I had a bed in an old office
building that I had converted into a recording studio.
||OK, that's enough nostalgia for one lifetime. Back to the
||I had a friend that bought the first widely available
commercial sampler that Emu made and I
want to say it was over ten grand.
||OK, can't find the price but I see that it replaced a unit E-
made for $70,000. Not a typo, it was the same price as a
Ferrari at the time.
||Anyway, this guy made the studio rounds with his $10,000
keyboard charging for
its use and he did appear to keep busy. We'd all gather
it like the monkeys in 2001 marveling at the monolith.
||Anyway, I almost felt bad for him when the Emulator 2 and
subsequently much much cheaper samplers came out but
was obviously offset by the fact that I could now afford
||U-law or a-law encoding does something like this
but for individual samples: samples with values
near zero (that might be part of a low volume
sound) are encoded with more precise quantisation
than samples with high magnitude positive or
negative values (that are part of a high volume
sound). They only use 8 bits to store each sample
though so they may still have more quantisation
than you want - probably at least as much as a 16
bit sample without the technique. A similar idea
would be to encode samples as floating point
||One way of doing this is to digitise each sample
with the minimum quantisation you might need
then convert to floating point to reduce storage
space. Another might be to make a direct floating
point analog to digital converter, which I've never
heard of being done.
||Dave Rossum in an interview said that, when they were developing the Emulator (I or II I guess.. or maybe the Drumulator), and had to convert between different clock rates, they had much better results using a steady stream of simple + or - bits at a very high frequency(2MHz?) rather than interpolating or dropping bytes at the comparatively low rates in the 10's of k's.
||Those guys would know about this because they're
the pioneers of cost effective sampling and did this
with 80s computer technology.
||They were located close to me so you could drop a
piece of gear off to be upgraded and spend the day
at the adjacent beach towns waiting for it. What a
friggin' cool company.