Ableton Forum

forgie wrote: There ARE many technical reasons why high sampling frequencies can increase the quality of an audio production - but as some people have said, they mostly apply to the recording and perception of real instruments.

AND THE CALCULATION OF SOFTSYNTHS & EFFECTS
as I explained earlier

Internal calculations of soft synths is only indirectly relevant to this discussion. I don't give a damn if a softsynth samples at 386Khz internally, as long as it sounds good. That's for the synth designers to work out, not me. it's what it comes out as that we're talking about.... aren't we?

err_fatale wrote:nah, when recording real instruments, you want all the resolution you can get. For electronic music and synthetic sounds, lower sample rates and resolution are OK, but definitely not for vocals and real instruments!!! It sound like crap. No way recording an acoustic guitar or nice female vocals in 16/44.1 is going to sound good at all after you throw some effects on it. NO way. Sound like a computer, not real; synthetic, sometimes you want a clear reproduction, plus when using FX and processing sounds you want higher resolution or your sounds just degrade into digital mush/noise.....how can you take your music seriously when it sound shitty like that??

Are you running Live at 96k with Warp on and a fair amount of plugins?

Do you render a mix at 96 then convert to 44.1 for CD's? How? With what?

Geez, Live is barely getting by on my nice rig. How many tracks of 96k audio do you run? What do your productions involve? Computer specs? Do tell.

forgie wrote:I don't give a damn if a softsynth samples at 386Khz internally, as long as it sounds good.

take a guess what sample rate most of them work at internally .
clue - it's the sample rate of your card. Most softsynths dont run every process at 4x sample rate, they just operate at the card frequency. If your card is 44.1 , the synth calculates at 44.1

which leads us to that big image up there.

Angstrom wrote:

forgie wrote:I don't give a damn if a softsynth samples at 386Khz internally, as long as it sounds good.

take a guess what sample rate most of them work at internally .
clue - it's the sample rate of your card. Most softsynths dont run every process at 4x sample rate, they just operate at the card frequency. If your card is 44.1 , the synth calculates at 44.1

which leads us to that big image up there.

Yup.

In my opinion, you should really consider what format you're publishing to for choosing the samplerate to work at- above all else.

If you're going to CD- 44.1 or 88.2- not 96. As Knotkranky stated earlier.


Vinyl- there is no consumer or pro needle that can read frequencies very far above 20 khz. There are no records cut beyond this frequency level either.
The reason for this is easy- since there are no needles that can read that high, a ridge containing data that fast would be run over by the needle as oppossed to tracking. The result is 100% distortion for any frequency thats higher than what the needle can read.

IF there is a benefit to 192- and I don't know if there is or isn't- then its a benefit that can only be reached by correctly using tons of very high level pro equipment. And I just don't know- are there monitors out that have a driver that can push frequencies that high? If you check the specs on most high end mics, they certainly don't record that high. Obviously it exists, so there must be some reason. But I've got to believe its marginal. For a marginal benefit, you would have to have a lot of excess resource because the tradeoff is also seriously increased processing and HD space.

I believe in 24/44.1. It works for me as a good solution between good sound and managment of resources.

http://messageboard.tapeop.com/viewtopic.php?t=32692

these guys know what they are talking about

is it bad internetiquette(is that what you guys call it?) to put a link of another messageboard?
but anyhow really informative, more from the studio perspective.

and a cool free recording magazine(the best in my opinion)real diy, yep free, they pay for it with advertising.

Angstrom wrote:

forgie wrote:I don't give a damn if a softsynth samples at 386Khz internally, as long as it sounds good.

take a guess what sample rate most of them work at internally .
clue - it's the sample rate of your card. Most softsynths dont run every process at 4x sample rate, they just operate at the card frequency. If your card is 44.1 , the synth calculates at 44.1

which leads us to that big image up there.

How can soft synths sound good at 44.1Khz then? I know that a lot of synths I use sound good, and DON'T have the sort of artifacts that would be generated by internally calculating everything at 44.1Khz. They must use some other mathematical methods to calculate what effect the unheard transients between each time quantum are. If they do use some sort of mathematical function internally that doesn't give you aliasing artefacts, then drops it down to 44.1Khz, then I don't see what your point it. If however they seriously do calculate each sample without using some sort of tricky calculus to determine what effect the 'gaps' have on the output, then you are correct.

I don't know much about synths, but I know a bit about maths! I'm eager to understand how softsynths are normally programmed.

knotkranky wrote:

err_fatale wrote:nah, when recording real instruments, you want all the resolution you can get. For electronic music and synthetic sounds, lower sample rates and resolution are OK, but definitely not for vocals and real instruments!!! It sound like crap. No way recording an acoustic guitar or nice female vocals in 16/44.1 is going to sound good at all after you throw some effects on it. NO way. Sound like a computer, not real; synthetic, sometimes you want a clear reproduction, plus when using FX and processing sounds you want higher resolution or your sounds just degrade into digital mush/noise.....how can you take your music seriously when it sound shitty like that??

Are you running Live at 96k with Warp on and a fair amount of plugins?

Do you render a mix at 96 then convert to 44.1 for CD's? How? With what?

Geez, Live is barely getting by on my nice rig. How many tracks of 96k audio do you run? What do your productions involve? Computer specs? Do tell.

no i use 24/48 usually......certain sounds I render at 96 if i want to get all obsessive about them but not often, mostly guitar or vocal sounds .....i'm no engineer here, i'm just fukkin around......I posed the question because someone was telling me that i needed a digi 002 with 192 if I wanted to recored live instruments or it would sound shitty and i started worrying, I think 96 is fine for me. I don't have the disk space/processor power and it would take so damn long to render too.....if i was a professional recording engineer i would worry about it but for me it's not practical or important.....i do notice the difference in a new audio interface over the built in 16 bit 44,100 khz in my notebook, totally.....

Angstrom wrote:tdespite no one reading it and continuing to post "I dare you to hear 52khz" type posts and completely miss the point.

Don't sweat it, not all posts get read, sometimes if a sucka has to read a post more than once they'll skip it, and any notion of math can put people off. There are misconceptions, opinions and TeraBytes of audiogeek flame wars on the web.

How do you make your synths? I've looked into .vst and .dxi development but for the most part when I get home I want to play and not get into that level of DSP/math/programming/thinking/web page reading. Can you lay out the basics on how to get started, what tools you download, how the writing and compiling of a plug-in goes?

forgie wrote:I'm eager to understand how softsynths are normally programmed.

well here's a quick runthrough of softsynth and effect writing:
normally what you would do is use a wavetable for the oscillators other than sine, which you get methematically. To produces a saw mathematically you either use a counter which will produce aliased harmonics.or additive sines in harmonic seiries which you need a hell of a lot of! So most people go wavetable, the alternative is bandlimiting - which is cpu intensive.

for filters you can oversample to calculate, but it's generally considered optional - because unless you modulate the frequency at audio rates you will be free from the worst artefacts. That said, there is still an audible difference in every day use between (for example) a basic 44.1 ladder filter and an oversampled ladder (moog type) . Most softsynths let you activate some sort of 'high quality mode' and eat more CPU if you want.

where it goes wrong - modulation at sampling rate
Even if you bandlimit your oscilators and oversample your filters you will still get nasty artefacts from modulation processes. Frequency multiplication (FM) often shows you some aliasing. In fact a good example is Operator - which actually advertises it's aliasing as a bonus ! In FM - thats how you get white noise, there is so much aliasing that it turns totally into noise.

That is a extreme example, but the same also appear across the board, distortion, ring mod, etc. can all produce aliasing pretty damn easily. Unfortunately they are CPU costly to remove so often they are left in. Lets face it many users cant tell in normal operation. It's like JPEG compression though... if you can't hear it - great! ,but once you hear it, you hear it everywhere.

regarding common effects such as EQ, Robert Henke has posted elsewhere on this board stating how EQ4 operates better at higher sample rates. I wont bore you further with why, though ... its another marathon post !

DeadlyKungFu wrote: I've looked into .vst and .dxi development but for the most part when I get home I want to play and not get into that level of DSP/math/programming/thinking/web page reading. Can you lay out the basics on how to get started, what tools you download, how the writing and compiling of a plug-in goes?

www.synthmaker.com , very useful code (c style and ASM) windows in a graphical IDE. Win only right now. Very helpful people on the forums

a basic ladder filter not oversampled

knotkranky wrote:192,000 slices a second is gonna sound better than 96,000 or 44,100 for that matter.

not nesecarilly.

The problem ith digital sampling frequency is that you get an extinction frequency exactly half of the sampling frequency.

example - if you record a 1khz test tone with a sample rate of 2khz, you will hear nothing, because the point of sample will fall on 0 every time, unless the phase is moved by 90 degres in which only a tri wave will be heard.

Thats why CD's have taken 44.1khz, because the extinction freq. will be 22.05khz which is just outside human hearing,

However this will alter the phase, so the high frequencies will sound more spatial and out of phase.

to correct this, you can simply add frequencies from 48khz, which will keep the phaze in line.

To do this you will need a sample rate of 96khz minimum.

If you want to record and hear the differences between 96khz and 192khz,you will need monitors capable of reproducing more than 50khz - example Tannoy precision range, and you will need microphones capable of produing more than 50khz & samplers/synths capable of produing more than 50khz.

in a proffessional studio, Yes, in your bedroom - very unlikely.

the point here is this:

if you cannot prduce the sounds to take advantage of 192khz, and you dont have the equipment to present 80khz (monitors),

then there is no need to use 196khz.

especially when making dance music, which is big on bass and mid frequency ranges.


I havnt read the whole thread BTW, sorry if all of this has already been said.

Angstrom wrote:www.synthmaker.com , very useful code (c style and ASM) windows in a graphical IDE. Win only right now. Very helpful people on the forums

Very cool, reminds me of Max MSP, is that a fair comparison?

What's your background on this stuff? I'm just curious how you got to your level of understanding. I have a bachelors in electrical engineering, I'm a hardware guy, I didn't focus on DSP, but I had some classes in DSP theory, MatLab, Communications theory, numerous flavors of Laplace and Fourier, crap like that. I've been out of school for almost 10 years, I get the gist of the filter code but anymore I couldn't get to those numbers without cracking some old textbooks. Any advice? Looks like a good forum, 304 posts heh? You do know your shit.

I used MaxMsp a little and SM is indeed similar, but I gave up on max for one reason and another.
I'm probably not the best person to advise on starting DSP as I came in sideways from stuff like reaktor and a little bit of doing it in C++ before I realise I couldn't be bothered writing classes for absolutely everything I needed before getting onto the good stuff. I'm no-where near as skilled at this stuff as some of the people on either kvr or the SM forums.
sources of info are mainly
http://www.musicdsp.org/
and google books , for pilfering code.
like this..
http://books.google.com/books?ie=UTF-8& ... 2Dixivede0
as long as you have an idea what you are looking for you can get clues from there.

No useful college education on this stuff for me, I trained as a designer before learning all my sound info on the job in studios, sitting on the couches waiting for things, reading tape-op magazine.

btw my high postcount merely indicates a life spent too close to a computer

I was working with 96 for awhile. I have not noticed a difference with 44.1, and I'm working with much more ease. I have a decent room, $1000 monitors and converters worth $2000. Bedroom quality stuff.

Make it easy for yourself in the bedroom environment. Stick with 44.1, unless you notice a difference between that and 88.2. I see no advantage to 48 if you're going to put it onto a CD and have to resample. If you have a high end studio, by all means... go with 192.

Alright, maybe I'll play around with 88.2 again now that I have an external reverb/effects processor and a quicker computer.