Click Here
home features news forums classifieds faqs links search
6050 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
Home
Features
News
Forums
Classifieds
Links
Downloads
Extras
OS4 Zone
IRC Network
AmigaWorld Radio
Newsfeed
Top Members
Amiga Dealers
Information
About Us
FAQs
Advertise
Polls
Terms of Service
Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
79 crawler(s) on-line.
 15 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 pixie:  16 mins ago
 BigD:  23 mins ago
 LaBodilsen:  36 mins ago
 michalsc:  43 mins ago
 MEGA_RJ_MICAL:  1 hr 28 mins ago
 W0IW:  2 hrs 16 mins ago
 Karlos:  4 hrs 2 mins ago
 sibbi:  4 hrs 4 mins ago
 AmigaMac:  5 hrs 15 mins ago
 agami:  6 hrs 48 mins ago

/  Forum Index
   /  General Technology (No Console Threads)
      /  Audio processing, effect simulation, that sort of thing...
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 Next Page )
PosterThread
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 28-Dec-2022 5:57:07
#21 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
Anyway, I think we are in agreement, as much as it doesn't look like it: You want the process things at the highest quality you need resolution.


I think so too. Prior to final mix down sample rates, as high as you can afford will give better results, unless you are all analogue in which case, the better your gear and high grade particle tape, heads, motors pre-amps etc play a more important roll.
Digital has a much better noise to amplitude ratio. Analogue can have a better fidelity and wash/warmth (I hate those terms) simply by adding all those harmonics into the mix(soup of sound) at the cost of colouring with increased noise as well.
Both very much have their uses and place. They operate differently and very much have a sound, that can be captured using the right techniques, even in digital.

Quote:
The reason this doesn't sound awful is that the immediate output goes through a reconstruction filter, in the analogue domain, that effectively produces the analogue sum of the 16 discrete inputs, as well as performing the LPF cutoff to get rid of unwanted artefacts above 22 kHz or so. The end result is a smooth continuous output

Right. Well analogue has no frequency limit other than dither into resistance for amplitude. So, that makes a lot of sense.
Iin the digital domain, you still apply filters to stop aliasing, even AMIGA has a crude analogue filter to stop higher frequency aliasing poking out audio. Another option is oversampling, prior to any filtering.
If you didn't apply those filters, while mixing at those sample limits, that aliasing error, or noise that's created, is applied to the mix, causing it's own harmonics that can, and do often end up in the audible range, adding their own type of colour. Something you may or may not want.

Can you tell the difference between an analogue generated tone at the frequency a digital one did jitter? Yes. One jittered and sounded a certain way then got filtered or cut off. you could check that by going 4 bit Vs 8 bit or 16 I guess that doesn't error at limit and apply a filter. Then again, analogue filters don't reproduce square wave sounds very well. Everything gets rounded off by the analogue domain of amplitude Vs resistance however circuit resonates and delays, material tolerances all play a part in shaping that sound. Maybe that's why people still like transformer noise and consider sound warmer once it goes through it all. DSP technology has come a long way to emulate a lot of this these days, along with oversampling VST plugins etc.

Again, the resonate frequencies and harmonics of a REAL plate reverb are still highly CPU intensive to re-create in the digital realm, if you can call them accurate. A lot of plugin manufactures are out there trying to re-create the old Abby Road analogue gear like Fairchild 660/670 compressors that cost upwards of 20 thousand. Some say they come darn close. Till you do a null test and then, well, they don't but they still sound very good. It's all about how it sounds though, right?

For playback however, I find it very difficult to justify anything over 48k. Mix down plots of waveforms in the overall mix of sound curves, all those harmonics have collided without aliasing in the high-res mixing domain, are now going to be plotted for correct playback of the overall mix against filters that cut off any jitter that no longer needs mixing. it's going out analogue to your ears now.

The only thing left is..........the noise floor.

If you are printing to analogue vinyl, you will be adding in a much higher amplitude noise floor, just by using that medium. So, using a much lower noise to begin with, creates a superior product.

And that's why analogue mastering, can have a benefit of a higher sample digital source.

Last edited by SHADES on 28-Dec-2022 at 06:27 AM.
Last edited by SHADES on 28-Dec-2022 at 06:11 AM.
Last edited by SHADES on 28-Dec-2022 at 06:09 AM.
Last edited by SHADES on 28-Dec-2022 at 06:07 AM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 28-Dec-2022 10:00:40
#22 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
Right, so, instead of tracking 0 to 180degrees you go 0 to -180 degrees. The sign wave is inverted. or am i missing something here?
If you have two frequencies and one is inverted, in the same time domain, they cancel out right?


They do, but I'm keen to avoid overcomplicating what is being described.

I guess what I was saying is, I don't know what you mean by "inverting the frequency". The inversion of frequency is period, e.g. frequency is per second, period is seconds.

When I said inverting a signal I just mean sign inversion, e.g. flipping the signal upside down. If you mix it with the original, you get silence.

For a sine wave, the inverted signal also looks like a phase shift of 180 degrees (either direction) due to the particular symmetry of the function, but I'm just talking about plain old amplitude inversion.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 28-Dec-2022 11:04:18
#23 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

Quote:
Can you tell the difference between an analogue generated tone at the frequency a digital one did jitter? Yes. One jittered and sounded a certain way then got filtered or cut off


Definitely, depending on what the tone is, the sample rate and so on. For a sine wave however, it's a lot harder. The main reason being that, ideally, all the energy is at one specific frequency. As long as the digital stream has a reasonable sample rate and bit depth and has an appropriate reconstruction filter on the output, and the sine wave itself is well below Nyquist, the reconstructed analogue signal will be an extremely good match for any analogue generated sine.

When it comes to squares you can create digitally perfect waves based on having a period that's an exact, even, number of samples long for the full duty cycle. It will be faithfully generated right up to the DAC stage where the distortion begins. The first thing being phase differences introduced by the latency of the DAC itself but that's likely to be measured in nanoseconds or microseconds at worst and so completely above any subjective measurement technique. However, the analogue signal simply can't just turn on and off, or switch from v to -v because all circuits, even ones constructed from just gold wire have resistance, capacitance and inductance. They will always interact in a way to counter the sudden hard transition from on to off or vice versa. Nothing physical has an instantaneous response, so there will never be a perfectly hard edge. Whether they are enough to produce a subjective difference audibly is hard to quantify for any real circuit with actual components in it. Next, the DAC will have a reconstitution filter on its analogue output, since that's a requirement to prevent aliasing signals and that will specifically aim to remove (or at least strongly attenuate) all the frequencies above half the stream sample rate (if it's properly designed anyway). However our DACs raw analogue output, in this specifically rare case actually carries energy at many frequencies above the Nyquist. Even a significant amount above the input sample rate due to the analogue response to the hard edge switching (because a hard edge can only be perfectly described by a infinite series of harmonics in the frequency domain). Those will all be attenuated, rounding off the corners. The only place such truly hard edges can exist and be processed as hard edges are in the digital side of the system. But eventually you need to convert to analogue and any digital hard edges in the signal will be softened regardless. For normal audio applications this attenuation will only begin at around 20+ kHz by design so it doesn't really matter. You can still blow your speakers and damage your ears with it, even if not perfectly square :D

Last edited by Karlos on 28-Dec-2022 at 02:41 PM.
Last edited by Karlos on 28-Dec-2022 at 02:40 PM.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 28-Dec-2022 15:02:16
#24 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
The only thing left is..........the noise floor.


Nothing beats analogue for noise generation. It's an absolute pain in the bum to programmatically create noise efficiently. As you'll probably know, noise is an important aspect in modular and subtractive synthesis. I've written a reasonably efficient white noise generator that has approximately the right uniform spectral characteristics but only needs to call the RNG once per audio packet.

It's not only about generating noise for sculpting into sound. The digital implementation of filters tends to be around modelling poles. As long as you are feeding them a signal everything is fine, but protracted periods of truly silent input can result in "collapse" where, for example, state values within the filter get stuck (you can imagine the poles as attractors the complex plane in the analogue domain, with amplitude and phase as your real and imaginary axes). Digital implementations with their finite numerical precision can simply stop working without being "reset". This is almost never a problem due to the fact they are sitting directly on the output of some signal generator. For standalone cases though, adding a noise injection into the input at levels below the final output precision prevents the issue nicely without the need to add a bunch of branchy conditional logic within the filter itself which is suboptimal. And you can just reuse a short section of pregenerated noise to achieve this so it can be done in a computationally efficient manner.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 22:25:24
#25 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
They do, but I'm keen to avoid overcomplicating what is being described. I guess what I was saying is, I don't know what you mean by "inverting the frequency".
The inversion of frequency is period, e.g. frequency is per second, period is seconds.
When I said inverting a signal I just mean sign inversion, e.g. flipping the signal upside down.
If you mix it with the original, you get silence.
For a sine wave, the inverted signal also looks like a phase shift of 180 degrees (either direction) due to the particular symmetry of the function, but I'm just talking about plain old amplitude inversion.


Very big difference in inverting a pure sign wave Vs a complex signal source.
I think i was talking about an analogue snare.
If you think about he top of a drum skin getting pushed down by the hit of a stick inside the drum barrel, it then pushes the bottom skin out (-180) as that pressure in the drum hits the bottom skin.
There is also the tone that the two skins are not tuned the same.

If we are talking mixing as a whole, once everything is mixed together, you will often find that a kick drum (also has the same issues, with beater Vs boom sides, or inside/out skins) plays in the same range as your Bass guitar. (Quite often tom and snare range too) When mixing, these sounds, they sound lifeless(hate these terms) it's just phases of similar frequencies, cancelling each other out when mixed together.
You will find that, simply flipping or inverting a bass guitar signal or mixed kick (after phase correction) fixes that problem entirely, or you may shift a few milliseconds left or right, or you can try side chaining compression where you use the input of the kick channel to tell a compressor when to compress a bass signal, to make it poke through in amplitude. Or a combination of all three, or you may try mid/side eq. Where the sides are your spatial stereo source and your mid is the mono energy.
Each drum in a drum kit fights with others, unless you use a single microphone to capture the overall kit, however, that almost never sounds as good as each drum bring treated as a separate musical instrument and captured and then mixed as a drum kit bus track, mixed into the other sources.

When sounds are miking, a lot of them overlap, and we use these techniques to get past these issues.
Then of course there is EQ and generated harmonics, like adding in a sub-bass, generated from the original source, as well. works great in EDM

Last edited by SHADES on 29-Dec-2022 at 10:30 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 22:45:59
#26 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Ok, there's a lot to unpack here. I'm talking about digital signal mixing. Even in the analogue domain, within some tolerances, this should still be true. If you invert a signal and mix it with itself, if you do so at the same intensity ans phase offset, then, poof, it's gone. X + -X = 0. It doesn't matter how complex the signal is, as long as it's the same signal.

The drum example seems very different, and I'm not sure how it is comparable. For one thing, we are dealing with two skins, and assuming the thing has microphones for top and bottom, there's two different (even if related) signals. There's the small delay from the propagation of the pressure wave through the drum which would presumably upset the phase offset a bit (wasn't it about 1 sample per 7mm or so at 330m/s 48000Hz? I think that's what I worked out). There's also other losses to factor here. Air is compressible so even if you tuned the top and bottom skins perfectly, made of the same material and had the drum in some temperature and humidity controlled environment, when you hit the drum head, not all the energy is transferred to the lower skin, some is just lost to compressive heating (I'm sure it's probably miniscule) and other vibrations that result in a less than perfect response from the lower skin than the upper one. I'm sure that it's possible to tune and mic up the whole instrument "just right" to maximise the loudness, timbre, etc. as well as an infinite number of ways to do it badly.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 22:48:36
#27 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
For a sine wave however, it's a lot harder. The main reason being that, ideally, all the energy is at one specific frequency. As long as the digital stream has a reasonable sample rate and bit depth and has an appropriate reconstruction filter on the output, and the sine wave itself is well below Nyquist, the reconstructed analogue signal will be an extremely good match for any analogue generated sine.


That's the point. Music is very rarely pure sign waves. Analogue doesn't make pure anything very well.

A single note off a piano or sung is full of harmonics. These harmonics are a combination on the instrument, the capture device, the room etc.
If you are taking electronically generated, digital, you get better results, converting to analogue and mixing after resample.
I've never seen pre-generated source, digital mixing done. Everything normally is in signasaudal conversion first. If adding in square wave sounds into a mix, high sample rates are a must due to the harmonics required to make the square wave mix in the sound source.
A square wave is actually an infinite series of sine wave harmonics added together, when created in the sound domain. The more harmonics it has, the more "square" the edges become, with noise.
In nature, square waves don't exist because sound is 360 degrees. Spherical. So, square waves are lots and lots of tightly packed frequencies that stretch out towards a centre. That can play havoc inside low res mixes as chirping.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 22:55:31
#28 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:

Karlos wrote:
@SHADES

Ok, there's a lot to unpack here. I'm talking about digital signal mixing. Even in the analogue domain, within some tolerances, this should still be true. If you invert a signal and mix it with itself, if you do so at the same intensity ans phase offset, then, poof, it's gone. X + -X = 0. It doesn't matter how complex the signal is, as long as it's the same signal.

The drum example seems very different, and I'm not sure how it is comparable. For one thing, we are dealing with two skins, and assuming the thing has microphones for top and bottom, there's two different (even if related) signals. There's the small delay from the propagation of the pressure wave through the drum which would presumably upset the phase offset a bit (wasn't it about 1 sample per 7mm or so at 330m/s 48000Hz? I think that's what I worked out). There's also other losses to factor here. Air is compressible so even if you tuned the top and bottom skins perfectly, made of the same material and had the drum in some temperature and humidity controlled environment, when you hit the drum head, not all the energy is transferred to the lower skin, some is just lost to compressive heating (I'm sure it's probably miniscule) and other vibrations that result in a less than perfect response from the lower skin than the upper one. I'm sure that it's possible to tune and mic up the whole instrument "just right" to maximise the loudness, timbre, etc. as well as an infinite number of ways to do it badly.



Right. No it's not possible to tune both drum heads to be exact. They are analogue. You may get close however, one head has a snare on it. Even if it didn't, the skins aren't exact, the microphones membranes aren't exact or the magnets the coils travel over etc. It's never exact.

Digital, pre mixing, source generated in real time, digital mixing. No idea.
Never seen that done.
Everything does get converted to analogue at some stage. Has to.
If you were to generate two square waves in digital and mix them in digital, I assume you could use an Or gate or something like that. I assume they would cancel in hearing. Everything you hear is analogue.
The sound domain is 360 degrees so, to hear it, means harmonics in sign.

Last edited by SHADES on 29-Dec-2022 at 10:58 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 23:01:43
#29 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
Music is very rarely pure sign waves


Music? Well, no, but electronic audio generation in general uses them for all sorts. Just consider traditional FM synthesis. Having said that, of course, FM synthesis is rarely done using analogue methods outside of modular synth setups.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 23:07:40
#30 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
The digital implementation of filters tends to be around modelling poles. As long as you are feeding them a signal everything is fine, but protracted periods of truly silent input can result in "collapse" where, for example, state values within the filter get stuck (you can imagine the poles as attractors the complex plane in the analogue domain, with amplitude and phase as your real and imaginary axes). Digital implementations with their finite numerical precision can simply stop working without being "reset".


Sounds like "ringing" vibration or oscillation being described. if the filter isn't getting a value, the output of the effect wouldn't be heard correctly, as there is no values being output to calculate change from.

Yes, even downsampling, adding in dither or noise helps it sound better, I assume because plotting of the waves has different intersection points in sampling due to being noisy, and correct frequencies poke through nyquist limitations when they phase. With the crap noise along with them.
Maybe that was the whole idea on making the 1-bit pc beep speaker work with real audio. I guess if you could switch it on and off fast enough.....

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 23:09:34
#31 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
Just consider traditional FM synthesis. Having said that, of course, FM synthesis is rarely done using analogue methods outside of modular synth setups


Even that has to go analogue to be heard. Mixing wouldn't be done digitally. It would be done after generation. I often mix in FM synth stuff and i'm using waveforms, not digital signals.

Here lies the problem in definitions. Sound is angular in form and offset in floating point definition. To define in digital requires accuracy, samples and bit depths. To reproduce requires, filtering and force of vibration. Unless being beamed into your brain, even then, the brain is cooked to look for patterns in vibration, so, this is where analogue and digital naming conventions collide as either something or nothing.
In nature, sound must be thought of as a wave of 360 degrees. No matter how it is represented in bits.
Filters can help shape curves but alter sound, by doing so, maybe in the correct arc, maybe not depending on what is left out by defining them in accuracy.

You could say, digital isn't perfect at capturing analogue sound and analogue sound is not perfect at reproducing digital form. Each has it's own pro's and cons. Which was my main point and why I use both.

That noise floor when being taken to a vinyl presser, only gets amplified by the printing to vinyl, and why starting with a lower noise floor, can be of benefit if the vinyl gear is half decent.

Last edited by SHADES on 29-Dec-2022 at 11:23 PM.
Last edited by SHADES on 29-Dec-2022 at 11:10 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 23:37:41
#32 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

I don't follow. Mixing is done digitally all the time. All the digital synth equipment I've ever owned is entirely digital right up to the final output. Synths like the DX7 were exceptional in that they mixed using the aforementioned multiplexed digital output stages for each voice at 800kHz into a filter that summed in the analogue domain but that's the exception rather than the rule. Today, many digital synths include purely digital output. When playing complex multitimbral, high polyphony music, all the voices are being generated, processed and mixed internally without ever being analogue. And of course there are entire software DAW these days.

Even the last bit of musical equipment I own, a rompler I have an obturate attachment to (MU100R) has 64 voice polyphony, 32 part multitimbral, with filters, EQ, pan, effect send etc per part, and absolutely all of it, is digital, and that's from 1998 or something. Only the final, digital mixdown of all that goes to the internal DAC for conversion to analogue.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 29-Dec-2022 23:44:27
#33 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
each voice at 800kHz into a filter that summed in the analogue domain

Analoge.
Quote:
Today, many digital synths include purely digital output

Pure digital isn't sound.

Terms. lol
Sound by definition is analogue or 360 degrees.
Trying to mix in pure digital without plotting to waveforms first would be very CPU intensive.

A speaker is analogue.

But yeah, I guess it depends on how far you take it.

That's all i meant by that. I digressed.

Last edited by SHADES on 29-Dec-2022 at 11:45 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 0:08:10
#34 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
Only the final, digital mixdown of all that goes to the internal DAC for conversion to analogue.


Yep, so you can plot curves that sit in the analogue domain to make "sound" it's still converted as those arcs to create that sound, in analogue.

As for trying to do that in low resolution sample rates for analogue, it would sound bad/different to high resolution.

If you are talking complex waveforms like an analogue vocal voice being reproduced into digital, synthetic reproduction is extremely difficult and expensive due to it not being a clean sine wave or just a complication of square wave or other harmonics. It can be done, it's not very quick and highly expensive.
Once you get into harmonic vibration representations, how much bandwidth is enough?
44.1 isn't going to cut it when you are mixing music these days, you get chirping and other noises in your mixes due to harmonics that interact with the sounds, not being processed and reflecting back at incorrect angles, unless, you are sending that sound, to analogue gear. Like a real plate reverb for example, that doesn't have frequency limitations as such, and re-sampling the result, after the audio passed through it. Mind you, you have also, just added noise by doing so. The effect of going out to the analogue domain.

That's why I suggested the video on oversampling. So you can see how these effects occurred and why we mix in higher sample rate domains. The math hasn't changed.

In the analogue domain, you don't have the same issues because, it just gets noisier as frequency increases. Those digital reflections don't occur. They just interact in with the mix into resistance getting less energy to affect other waveforms. As opposed to, hitting Nyquist and reflecting back, as there is no where to go. If your sample rate is higher, you can just add a filter above the range of hearing frequencies so that as those harmonics start to reflect back down, you clip them prior. Not before they are/get generated. They are needed. And not in the wrong angle, as they sound weird because they shouldn't exist as reflections to begin with.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 0:21:31
#35 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

"Trying to mix in pure digital without plotting to waveforms"

This is a non sequitur. Digital simply means that the waveform is represented digitally. A stream of binary numbers, where each number represents the instantaneous intensity of the waveform at a discrete moment in time. To evenly mix two such streams you just add their respective discrete samples together.

More usually, each stream will have it's own volume level in the mix. To apply volume, you multiply the samples by the volume, before you add them. This is why most modern audio software uses floating point because the all the maths is neater when digital signals are in the range -1.0 to +1.0 and you have FPUs that eat multiply accumulate operations for breakfast. It's no sweat for modern CPU's, I mean I'm able to do it for a few dozen voices in php which is an interpreted scripting language! Mixing (i.e. summing of streams) is the least CPU intensive part of it.

Again, I suspect we are not quite talking about the same thing. Mixing signals digitally is is definitely not a problem. It's only ever a ball ache if the signals being mixed are at different sampling rates since you will need to resample one or more of them to some common (usually higher) rate.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 3:45:35
#36 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
It's only ever a ball ache if the signals being mixed are at different sampling rates since you will need to resample one or more of them to some common (usually higher) rate.


Or, adding effects, harmonics etc that end up being created beyond nyquist, noise, mixing with other instruments that create other higher frequencies when combined. fast transients etc

You can mix in different sample rates together in higher resolutions, going lower has more limitations with regards to perceived artefacts.

Harmonics will get you every time. Unless you don't mind that effect.

Maybe if digitally all generated, it's not as noticeable? I hear it in most amiga mods but that could also be sample bit depth and they are generally analogue sampled, so, nasty.

Basically, any type of frequency that is generated/created that has the ability to pass the 44.1 nyquist limit, has the potential of creating unwanted audible sound when mixing. If it's going directly out to analogue instead, like an amplifier, the analogue realm will take care of that, it's just during mixing.

Last edited by SHADES on 30-Dec-2022 at 06:07 AM.
Last edited by SHADES on 30-Dec-2022 at 04:27 AM.
Last edited by SHADES on 30-Dec-2022 at 03:49 AM.
Last edited by SHADES on 30-Dec-2022 at 03:48 AM.
Last edited by SHADES on 30-Dec-2022 at 03:46 AM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 10:24:55
#37 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
Or, adding effects, harmonics etc that end up being created beyond nyquist, noise, mixing with other instruments that create other higher frequencies when combined


Again I think you are using mixing in a different, perhaps more studio specific context. I'm using it in the "combine two or more streams of audio into one" sense. No effects, no EQ. Just mixing with control over each streams volume only.

To reiterate, you can't create harmonics above half the sample rate in the digital domain. When you have a 48kHz stream your absolute limit is 24kHz. Two successive samples is the shortest possible period for a wave and one sample is the shortest possible phase difference. Nothing processed at 48kHz rate can produce a frequency above 24kHz since that's the limit of resolution. You can resample that stream to 96kHz. You can now do stuff to it in order to generate harmonics up to 48kHz in the resampled data, but those will be lost when you downsample back to 48kHz. If you don't downsample properly (using a filter) you'll get some alias frequencies that may contaminate the audio region up to 24kHz.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
SHADES 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 11:38:35
#38 ]
Cult Member
Joined: 13-Nov-2003
Posts: 841
From: Melbourne

@Karlos

Quote:
Again I think you are using mixing in a different, perhaps more studio specific context. I'm using it in the "combine two or more streams of audio into one" sense. No effects, no EQ. Just mixing with control over each streams volume only. To reiterate, you can't create harmonics above half the sample rate in the digital domain.


That seems fair enough then. You can create harmonics that go beyond nyquist in studio digital domains, when looking at 6th order or more harmonics with digital mixing.
If you aren't adding in anything, you are limited to nyquist.
If you have high-ordered harmonics, things get out of hand quickly.

https://www.izotope.com/en/learn/4-ways-to-add-augment-or-excite-upper-harmonics.html

if you are sending generated high-ordered, combined digital generated sound, to an output, say analogue amplifier, it doesn't matter as sound is shaped by doing so and it goes analogue.

If you start studio mixing in these sorts of distortions and harmonics in studio daw scenarios, sample rates, matter.
If you are mixing down back to 48, nyquist cuts off anything above 24k freq. Above what we notice, as long as it wasn't introducing returned, harmonics by mixing. Direct generated / summed digital out to analogue output, would be, ok.

To re iterate, if the highest harmonic frequency is well below half the sampling rate (nyquist), and/or if the harmonic amplitudes decrease rapidly, with harmonic number, as normally happens in real world natural occurring waveforms, you get clean, high quality results. If not, you get artificial artefacts that add their own colour as they bounce back from nyquist limits and create further ordered hamonic series, that can effect your overall sound, prior to mix down, so, they effect the sample down.

This is a heavier read but explains it nicely. I find the maths insane though. Head to page 6
https://ccrma.stanford.edu/~stilti/papers/blit.pdf

In the studio, any type of dither, noise distortion, reverb, compression all have an effect of adding in harmonics that can often generate on a trajectory that exceeds twice the sample rate of 44.1 or 48.
I use 96 on just about everything, prior to dump at 48k for this very reason however, oversampling is a better way, if your plugin supports it. It's less work to do it on input and sample down to workspace mix domain after. Artefacts get filtered before coming back to audible range, that output gets mixed in with the rest of your tracks.
The other way is sending out to analogue with no limits and re-sample back that content. added noise from analogue is introduced however, on resample. So, can be used last in chain, or during if you want that smear to add "warmth", it also helps out in lower sample rates that may exceed nyquist, like 8 bit etc harsh digital tones, get smoothed by analogue manipulation just by being analogue.

As for phase and being milliseconds out making huge differences, in our previous discussions, huge variations in tone from micing up speaker cabinets, just by shifting the mic away or closer to the cabinet or instrument, going off axis to center of cone or side of cone. Actually a twin mic technique of capturing a guitar cab amp where one mic is offset by another, just centimetres apart, and then summing the two, tracks, in the box, can give a really strong sound. A well known trick in making sounds seem more weighted. The time domain change is tiny. Some even go as far as using two different mics on the one stand, pointed to the same speaker, slightly different angles

Last edited by SHADES on 30-Dec-2022 at 12:17 PM.
Last edited by SHADES on 30-Dec-2022 at 12:13 PM.
Last edited by SHADES on 30-Dec-2022 at 12:04 PM.
Last edited by SHADES on 30-Dec-2022 at 11:54 AM.
Last edited by SHADES on 30-Dec-2022 at 11:48 AM.
Last edited by SHADES on 30-Dec-2022 at 11:47 AM.
Last edited by SHADES on 30-Dec-2022 at 11:44 AM.
Last edited by SHADES on 30-Dec-2022 at 11:43 AM.
Last edited by SHADES on 30-Dec-2022 at 11:41 AM.

_________________
It's not the question that's the problem, it's the problem that's the question.

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 30-Dec-2022 14:03:25
#39 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
That seems fair enough then. You can create harmonics that go beyond nyquist in studio digital domains, when looking at 6th order or more harmonics with digital mixing


Presumably only as a consequence of oversampling. A 48kHz stream literally has no way to convey athing above 24kHz (other than unwanted aliasing frequencies below it). If your mixing process takes multiple streams, oversamples them to some multiple of their original sample rate and allows them to be mixed, boost, cut, EQ, compressed, phase adjusted etc. at that new sample rate then of course you can generate frequencies that weren't present in your originals.

Perhaps one place where this gets... interesting... is reconstruction to analogue. If I decide to omit the reconstruction filter for my 48kHz sample, my analogue signal will probably contain frequencies all over the place due to unfiltered chair step changes in voltage that the circuit can't cope with. This can make some things sound brighter but usually it will make things sound worse.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
Karlos 
Re: Audio processing, effect simulation, that sort of thing...
Posted on 31-Dec-2022 10:23:36
#40 ]
Elite Member
Joined: 24-Aug-2003
Posts: 3511
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@SHADES

Quote:
Actually a twin mic technique of capturing a guitar cab amp where one mic is offset by another, just centimetres apart


I remember reading about various cool things done using gated mics at different distances from the source, adding various chorus or other effects to vocals that only happen when the source volume increases above the more distant mics gate level. I think some of David Bowie's tracks used this technique on the vocals. Or the happy accidental discovery that led to the gated reverb on percussion made famous by Phil Collins.

Aesthetically, one problem I think exists with electronic/digital only production (by which I mean sound generation, mixing and processing done entirely electronically) is the lack of opportunity for similar happy accidents. Everything becomes by intent and if you want some cool new sound you have to think about it carefully before you capture it, or find a way to do it afterwards.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle