The Analogue Dream is dead ...

PaulBrewer

I was reading an interview with Danton Supple (some handle !) in Resolution recently where he was at a demo of the CLASP system

http://www.endlessanalog.com/what-is-clasp

where he was saying that when he heard it he and Mick Glossop agreed they didn't like it as it reminded them both of what they didn't like about tape !

Can we now say digital is better than tape ? I think so.

Is it now down to Good Digital and Bad Digital ?

krd

PaulBrewer wrote: »

Can we now say digital is better than tape ? I think so.

I think the thing with tape, is not so much the sound, but the rituals that go into tracking to tape, over tracking to a DAW.

A band, that sounds good as a band, can be more economically tracked (none of this 76 tracks for the guitar. etc.)

And sometimes maybe being forced to work harder is not such a bad thing.

Is it now down to Good Digital and Bad Digital ?

To avoid aliasing, do I have to record at 96 KHz?

The way the digital signal chain works, you could easily be getting a pseudo 96 kHz....if there's a process or effect that passes data at 44kHz. With plugins, there's no way to be sure one isn't.

splitrmx

And thus began analogue vs digital discussion #2938459630.

PaulBrewer

splitrmx wrote: »

And thus began analogue vs digital discussion #2938459630.

But it's usually the other away around !

Seziertisch

I know a couple of people that have used Clasp and loved it. Horses for courses, I guess.

madtheory

Never liked tape. It hisses, wobbles, dulls the highs, is expensive and machine requires regular calibration and cleaning. A DAW is ridiculously transparent by comparison.

seannash

Can we now say digital is better than tape ? I think so.

Not that i have an opinion on this matter but its not normally like you Paul to take two peoples opinions as gospel.
I'm sure you could easily find engineers out there who prefer analogue

studiorat

I like both.

A well aligned machine is still a thing of beauty.

PaulBrewer

True Sean - however my point is that the tide is turning regarding the romantic view of tape in pro circles.


Professor Glossop, a renouned audiophile, now mixes entirely in the box and has done so for a while - so while I wouldn't necessarily agree with all of his opinions his standing within the industry suggests to me that they're worth examining.

I thought it interesting insofar as its a pretty fair A/B comparison and guys with would have grown up with one technology now prefer the newer 'inferior' one.

madtheory

studiorat wrote: »

I like both.

A well aligned machine is still a thing of beauty.

Yes, it is beautiful. Although I think maybe an original EMT 240 trumps it . Or even better, a purple 1985 Colnago.

flyswatter

Think Foo Fighters recorded their latest album completely to tape didn't they?

studiorat

madtheory wrote: »

Yes, it is beautiful. Although I think maybe an original EMT 240 trumps it . Or even better, a purple 1985 Colnago.

these are a few of my favorite things...

It's blue actually.

Paolo_M

I much prefer using analog gear to using my ears.
At least then I can be sure that it sounds good.

peter05

There's a massive difference between the two in my eye/ear. You can tell straight away from listening to play back which is which.

Now digitial is perfect for amateur work in my eyes/ears, but to have something worked in a pro setting I much perfer tape.

I think the heading is a little misleading saying all analog gear is dead:rolleyes:. What would you be doing without your Manley EQ/ SSL compressor etc.

And to say that using plug-ins of these units emanulating the analog are better is a joke. No plugin can match a big f*ck off transformer in these units. Take the Shadow Hills Compressor for instant, if not been in use most people will bypass it and run the signal through its transformers to add something more to a mix/master. Can't be done with plugin.

My 2cents, analog is far from dead. LONG LIVE ANALOG!!:D

Paolo_M

peter05 wrote: »

There's a massive difference between the two in my eye/ear. You can tell straight away from listening to play back which is which.

Now digitial is perfect for amateur work in my eyes/ears, but to have something worked in a pro setting I much perfer tape.

I think the heading is a little misleading saying all analog gear is dead:rolleyes:. What would you be doing without your Manley EQ/ SSL compressor etc.

And to say that using plug-ins of these units emanulating the analog are better is a joke. No plugin can match a big f*ck off transformer in these units. Take the Shadow Hills Compressor for instant, if not been in use most people will bypass it and run the signal through its transformers to add something more to a mix/master. Can't be done with plugin.

My 2cents, analog is far from dead. LONG LIVE ANALOG!!:D

The a massive difference between different example of the "same" analog gear.
I don't see your point.
Each to their own though.

peter05

My point was, you cant compare chalk and cheese and get the same result from the 2. Both have different sounds it depends what your looking for from your music.

Most people are one or the other on these formats. It's like the OSX vs WINDOWS. a never ending saga.

Hayte

krd wrote: »

To avoid aliasing, do I have to record at 96 KHz?

The way the digital signal chain works, you could easily be getting a pseudo 96 kHz....if there's a process or effect that passes data at 44kHz. With plugins, there's no way to be sure one isn't.

Pretty sure you don't need to worry about aliasing ever since every AD/DA made since the mid 90s has anti aliasing filters before and after sampling. The main reason for designers building to 96khz is that it pushes nyquist far beyond the upper limit of human hearing, which relaxes the burden on the filter (this is a design and engineering problem, not a consumer one).

The main reason for consumers in using higher sampling rates is that you get less latency. 96khz has half the latency as 48khz given that dma buffer size is constant and you don't have any additional safety buffers and what not. The tradeoff is that the files are twice as big so they eat twice as much RAM and use twice as much hard drive space. The computational burden is greater too so you probably need a modern PC otherwise you will be bouncing alot.

krd

Hayte wrote: »

Pretty sure you don't need to worry about aliasing ever since every AD/DA made since the mid 90s has anti aliasing filters before and after sampling. The main reason for designers building to 96khz is that it pushes nyquist far beyond the upper limit of human hearing, which relaxes the burden on the filter (this is a design and engineering problem, not a consumer one).

It's not so much the AD/DA. I'm getting an artefacting problem when I make some stuff in Ableton, and then export the WAV, and begin working on it.

I understand the Nyquist theorem. At 48 kHz, all the signal below 24kHz shouldn't be effected - anything above 24kHz won't be in hearing range so it's unimportant. Though - and this is the though - The artefacting I'm hearing is well below 24kHz...

*When I say Nyquist frequency, what I mean is the frequency needed for human hearing 48kHz, to have all frequencies below 24kHz sampled precisely.

The AD/DA conversion is not the only place the Nyquist frequency is important. It's still important in any Digital to Digital process - that's basically any effect applied to the signal. You can see it in action if you apply a down sampling or bit crushing effect - you're doing a Digital to Digital conversion, but you're going well beneath the Nyquist frequency - and since the digital signal is in steps, when you sample well below the Nyquist frequency the frequency is corrupt (some steps are longer than they should be)

What I'm wondering - if there isn't a brick wall filter of the signal above 24kHz - after a Digital to Digital conversion, or before......Is it possible, the signal corruption in the top 24kHz(outside hearing range) is appearing in the audible bottom 24kHz........And now that I think about it. That's what my problem sounds like - a bitcrush sound that seems to be coming from nowhere.

If you're only concerned about the bottom 24kHz - before and after a DD process at 48kHZ, the top 24kHz should be wiped clear, to stop it creating an artefact in the bottom 24kHz.

I'm going to give it a try - but I don't think I have anything that filters above 24kHz

Has anyone else had a similar experience.

kfoltman

krd wrote: »

It's still important in any Digital to Digital process - that's basically any effect applied to the signal.

Yes and no. A simple gain (multiply by a constant factor) doesn't create additional frequencies, so no extra brickwall filtering is necessary. On the other hand, if the gain is variable, the output signal contains the frequencies that are sums and differences of the frequencies in the original and the modulating signal. Still, for most of the variable gain, the increase in bandwidth is negligible.

A linear filter (lowpass/highpass etc. without any distortion from "tube simulation" etc.) also shouldn't create any additional frequencies - but the shape (frequency response) of the filter depends on the sampling frequency - for example, a 10 kHz lowpass filter has a steeper response when using a 44.1 kHz sampling frequency than when using a 96 kHz sampling frequency (the zero is at 22.5 kHz instead of 48 kHz). It may be audible, but won't be catastrophic in most cases. The higher the sampling frequency, the more the frequency response resembles an ideal analog filter.

On the other hand, an effect like a bitcrusher is non-linear, so it creates intermodulation distortion that depends on the signal, but would have infinite bandwidth had it been implemented in analog. In digital domain, the frequencies that go above Nyquist frequency are reflected around it. So, if we have a pure 10 kHz sine wave as the input, the bitcrushing might create extra harmonics at 20 kHz, 30 kHz, 40, 50 and so on ... (the amplitude of these harmonics depends on the level/type of bitcrushing and amplitude of the input signal) The 50 kHz harmonic goes above Nyquist, so it's being "mirrored" as 48 - (50 - 48) = 46 kHz. The 60 kHz partial ends up below Nyquist as 36 kHz, and so on. Ultimately you may get well audible partials at 6 kHz, 4 kHz (mirrored again around 0 Hz), 14 kHz. Note that most of those aren't harmonics of the original frequency - which makes them sound quite ugly (aliasing).

If instead of 10 kHz you use a sine wave with slowly varying frequency (as in, slow sweep from 10 kHz up to 11 kHz ), some of the partials will go up (20 to 22, 30 to 33, 40 to 44) but for others, the frequency goes down - 48 - (55 - 48) = 41 kHz (which is 5 kHz below the original 46 kHz)

You can avoid some of the disaster by brickwall-filtering above the top audible frequency, but as the bitcrushed bandwidth is infinite (theoretically at least), you may still get some aliasing. The only way to avoid is to increase the sampling rate even further, either on the whole project, or by using an oversampling bitcrusher. The latter, by the way, sounds fairly bizarre.

And now that I think about it. That's what my problem sounds like - a bitcrush sound that seems to be coming from nowhere.

Yes, it's a >48 kHz frequency that was "mirrored" back into audible range. Aliasing, basically.

Does it make sense/is it readable? (madtheory?)

kfoltman

TL;DR version:
1. Create this signal chain in any DAW: a sweepable sine wave generator, a bit crusher and a spectrum analyzer
2. Start playing with the frequency of the sine generator
3. Look at the spectrum analyzer.

Robin Ball

A badly tuned drumkit sounds bad whether it is analog or digital. Personally I prefer the convenience over digital and with ever decreasing budgets and time, editing on digital is a blessing.

I think that you need to have a good room, instrument, player and mic before tape becomes a concern.

Analog outboard is certainly not dead and won't be for a long, long time.

ladhrann

peter05 wrote: »

My 2cents, analog is far from dead. LONG LIVE ANALOG!!:D

Remember, in the words of Steve Albini, "the future belongs to the analogue loyalists!"

madtheory

All that stuff above about converters reads like it was copied and pasted from a gearslutz post that someone copied via chinese whisper from wiki. Argh!

peter05

ladhrann wrote: »

Remember, in the words of Steve Albini, "the future belongs to the analogue loyalists!"

you forgot the "f*ck digital" part.:P

Editing in a DAW, am all for that. With converters been so good these days its not much of an issue. I did some tape cutting on a Studer back a few years ago and thought I was going to lose at finger at some stage's. , I was doing work exprience, it wasn't mine.

PaulBrewer

peter05 wrote: »

you forgot the "f*ck digital" part.:P

Editing in a DAW, am all for that. With converters been so good these days its not much of an issue. I did some tape cutting on a Studer back a few years ago and thought I was going to lose at finger at some stage's. , I was doing work exprience, it wasn't mine.

It wasn't your finger ?

kfoltman

madtheory wrote: »

All that stuff above about converters reads like it was copied and pasted from a gearslutz post that someone copied via chinese whisper from wiki. Argh!

Apparently my English is even worse than I previously thought Sorry.

madtheory

kfoltman wrote: »

Apparently my English is even worse than I previously thought Sorry.

The language is fine and I'm not singling you out. It's the facts that are the problem...

kfoltman

I was *trying* to say/show why it's not necessarily true that *all* processing in digital domain needs a 22 kHz brickwall lowpass after each stage of processing when running at 96 kHz, and why krd is getting an unexpected/weird low frequency after passing the signal through the bitcrusher.

As for the latter, It's probably one of these situations where a nice animated GIF would probably do the job much better than a thousand words, especially written in a hurry. But I'm way too lazy to do that.

It may be nitpicky and technical, and mostly incoherent - but I'd rather do that than fill up *another* thread with emotion-based arguments about "analog rulez/digital sucks" or vice versa.

krd

kfoltman wrote: »

Does it make sense/is it readable? (madtheory?)

Yes, that makes sense. I hadn't thought it through before (where am I going to get a brick wall filter that will do over 24kHz?)


The beauty of analog circuits, is they're usually designed with the human hearing range in mind - so the higher frequencies tend to be mushed out - I think the partials might come down into the hearing range, but they may come down in a more warm way - analog circuits will tend to saturate at higher frequencies, that circuit has not been designed to target - digital equations don't. So, that crisp coldness of digital circuits could be emergent partials (in your school physics book look up beats) coming into the hearing range

Still I never thought of brick walling above 24kHz

krd

Robin Ball wrote: »

A badly tuned drumkit sounds bad whether it is analog or digital. Personally I prefer the convenience over digital and with ever decreasing budgets and time, editing on digital is a blessing.

I think that you need to have a good room, instrument, player and mic before tape becomes a concern.

Or, you could ask Robin Ball, for some drum samples, he's prepared earlier.

kfoltman

krd wrote: »

(where am I going to get a brick wall filter that will do over 24kHz?)

No idea, I'm not really familiar with the VST/RTAS scene except for the small set of plugins I tend to use.

I'm not even sure if you really need brickwall characteristics - maybe a steep IIR lowpass is all you need. Say, 8 pole lowpass from Absynth, or a stack of lowpass filters from Reaper's built in EQ plugin. It won't be linear phase or even close - but I think it doesn't matter THAT much, especially if you're using it with deliberate serious quality degradation anyway (bit-crusher and similar effects).

Either way, it may or may not help - the theoretical bandwidth of a bit-crusher is infinite, so you may get plenty of aliasing even when using the filtering. The only way to get very little aliasing is to upsample to gazillion kHz, bit-crush and downsample to 96 kHz - totally pointless IMHO but can be done.

madtheory

Hayte wrote: »

you don't need to worry about aliasing ever since every AD/DA made since the mid 90s has anti aliasing filters before and after sampling.

Wrong. On the AD it's called an anti aliasing filter, on the DA it's a reconstruction filter. They're very similar alright though. Dunno where you got mid nineties from, the filters have always been part of the process- maybe you're confusing it with the emergence of on DAC oversampling techniques? Because to conform with Koletnikov-Shanon-Nyquist, you have to limit the bandwidth first.

Hayte wrote: »

The main reason for designers building to 96khz is that it pushes nyquist far beyond the upper limit of human hearing

96kHz is marketting nonsense based on the fallacy that "more is better". It's a nonsense when all converters are oversampling already.

Hayte wrote: »

The main reason for consumers in using higher sampling rates is that you get less latency.

No, that's not the reason, although it is a somewhat useful side effect in some systems.

Hayte

krd wrote: »

It's not so much the AD/DA. I'm getting an artefacting problem when I make some stuff in Ableton, and then export the WAV, and begin working on it.

I understand the Nyquist theorem. At 48 kHz, all the signal below 24kHz shouldn't be effected - anything above 24kHz won't be in hearing range so it's unimportant. Though - and this is the though - The artefacting I'm hearing is well below 24kHz...

The whole point of inserting an analogue low pass filter before the sampler and the ADC is to stop the aliased components from being sampled in the first place. Once its in there, its in there and it doesn't matter if you jam filter plugins on every channel, after everything that makes a sound. Signal generation (as opposed to sampling) is different but the problem and solution has been well understood for decades. The difficulty is in eliminating aliasing without crushing your CPU. You don't need to worry about that because it is a design and engineering problem for people that make digital signal processors and generators.

All of your signal generator plugins probably have band limited oscillators or they will oversample. Sometimes they might even give you the option to choose how much oversampling happens so you can control how much your CPU is going to get squeezed i.e. uHE ACE. The ones that don't have aliasing as a deliberate effect i.e. ReFX Vanguard and even thats been optional since like 2005. Around the time when they gave you the option to switch between "classic" and "non aliasing" oscillators.

madtheory

You make it sound like the aliasing exists prior to sampling. But it doesn't. What is removed is the harmonics that will cause aliasing during decimation. Then the reconstruction filter fills in the gaps to return the signal to its original form.

But you're right that krd's problem is unlikely to have anything to do with sampling. But there's no way we can diagnose the problem until we hear these mysterious artefacts he speaks of!

Hayte

madtheory wrote: »

You make it sound like the aliasing exists prior to sampling. But it doesn't. What is removed is the harmonics that will cause aliasing during decimation. Then the reconstruction filter fills in the gaps to return the signal to its original form.

You clearly understand what I'm talking about so why are we even having this argument over semantics?

96kHz is marketting nonsense based on the fallacy that "more is better". It's a nonsense when all converters are oversampling already.

The point is to relax the burden of the analogue filters (for anti aliasing), so they can be made easily and cheaply. It is true that 96khz converters are already surplus to that requirement, which is what led Dan Lavry and others to say that 192khz was marketing nonsense.

So what about 48khz x2 oversampling vs 96 khz no oversampling? This is another engineering problem that really has no relevence to me or you because it has either been addressed by the people that designed our soundcards or it hasn't. Those guys will have to think about the number of decimation stages and how to get back down to the sampling rate you want.

I've heard from folks like Jon Hodgeson that its perfectly possible for one to be as good as the other but there are a different design decisions, costs and cost saving measures involved. Shrugs. If you want to talk about that then go to prosoundweb or other places where they have tech people discussing these sorts of things in detail. But all the time and effort you put into learning the engineering behind music technology won't help you make better music.

No, that's not the reason, although it is a somewhat useful side effect in some systems.

What do you use higher sampling rates for then? The only boon I've found is that you get less latency. Maybe you notice a difference in sound quality but whatever it is for you, it ain't worth double the computer pain for me. You probably have different workflow and a different set of priorities so shrugs. Whatever works for you. (For the record I use 24 bit/48 khz.)

The bottom line is that I'm not a hardware/software engineer. I just make music at home. If you want me to justify technical points on the inner workings of my hardware then all I can do is point you to people like Paul Frindle, Dan Lavry and Jon Hodgeson. Everything I've said is just paraphrased from those guys anyway.

About the only advice I feel qualified to give, is to not concern yourself with engineering problems if what you really want to do is make music. For a start, you begin to train your ears to listen for things that most people don't know or care about. You begin to spend an inordinate amount of time and effort on problems that are at best insignificant to music making and at worst, completely irrelevant. You spend time familiarizing yourself with engineering problems but you lack the engineering knowledge to understand and correct them, thus you may start doing things that make everything worse like putting computationally expensive filters on every channel to fix a problem you don't understand. At best it'll squander computer resources and do nothing. At worst, you are probably using non linear phase, analogue modelled ones with massive Q so all you are doing is phase smearing treble on every channel for no reason.

kfoltman

madtheory wrote: »

But you're right that krd's problem is unlikely to have anything to do with sampling. But there's no way we can diagnose the problem until we hear these mysterious artefacts he speaks of!

As far as I understand it, he *did* describe the signal path already? A signal (that was already in digital domain and had no audible aliasing in first place) was processed entirely in digital domain using a bit crusher, resulting in additional low frequencies not present in the original signal.

A typical bit crusher is a constant, non-linear system, so it's creating new partials due to intermodulation between input frequencies. The frequencies of those new partials are sums of integer multiples of the frequencies in the input signal.

In case the input signal is harmonic, all the new partials are integer multiples of the fundamental frequency. The only reason for the inharmonic partial would be if one of the new high-frequency partials aliased back into audible spectrum. (I'm not talking about any A/D or D/A conversion here - just aliasing due to non-linear processing in digital domain)

In case the input signal *isn't* harmonic, it becomes more complicated: the "unexpected" additional frequencies might be either due to intermodulation distortion itself (so it wouldn't be "digital-related" at all, the same would happen in a theoretical purely analog process), or may be due to aliasing of some of the newly generated partials.

I don't think there's much more to it really.

(trying to write about this is likely a huge waste of time, but at least it's a good excuse for myself to try and poke holes in my understanding of the subject :P)

kfoltman

Hayte wrote: »

What do you use higher sampling rates for then?

FM synths. FM may create a lot of high frequencies, that may end up as aliasing.

On the other hand, the world's most famous FM synth, Yamaha DX7, used sampling rate of 32 kHz (AFAIK) and very few people were complaining about it.

Actually, it was 57 kHz. So maybe using higher sample rate isn't such a stupid thing in this particular case.

madtheory

Hayte wrote: »

But all the time and effort you put into learning the engineering behind music technology won't help you make better music.

Who said we're learning this in order to make better music? I learn it because I'm fascinated by it. I also happen to make music. Learning is fun, and I don't believe in being goal oriented. I do believe that one should understand one's tools though.

Hayte wrote: »

What do you use higher sampling rates for then?

I don't- although a lot of the plugins we use have upsampling, which has many benefits for audio processing. Some plugins let you switch it on and off, such as Slate VCC. It's interesting to hear how sound and processor load are affected.

Hayte wrote: »

The bottom line is that I'm not a hardware/software engineer. I just make music at home.

Thanks for the clarification.

madtheory

kfoltman wrote: »

As far as I understand it, he *did* describe the signal path already? A signal (that was already in digital domain and had no audible aliasing in first place) was processed entirely in digital domain using a bit crusher, resulting in additional low frequencies not present in the original signal.

So let me get this straight- he's using a plugin that is meant to create aliasing distortion, and is wondering why there are "artefacts"? That's mad- what does he expect? Surely there's more to it than that? Is the issue that the alias tones are LF? Anyway- my point is that if we could hear these "artefacts" we'd probably recognise them straight away.

Edit: he doesn't mention bitcrushing at all?? He says "I'm getting an artefacting problem when I make some stuff in Ableton, and then export the WAV, and begin working on it."

So again, we need to hear it in order to diagnose it.

kfoltman wrote: »

In case the input signal *isn't* harmonic,

Why wouldn't it be "hamonic"? What does that even mean?

kfoltman wrote: »

FM synths. FM may create a lot of high frequencies, that may end up as aliasing.

By "may" do you mean- it's up to the designer to sort those out?

kfoltman wrote: »

(trying to write about this is likely a huge waste of time, but at least it's a good excuse for myself to try and poke holes in my understanding of the subject :P)

Agreed! I always learn something new this way. It's great to see someone else's POV.

BTW there's a very good thread on GS where they analyse various plugins for aliasing etc.

rcaz

ladhrann wrote: »

Remember, in the words of Steve Albini, "the future belongs to the analogue loyalists!"

Remember, Steve Albini is a bit of a ****... :pac:

kfoltman

madtheory wrote: »

So let me get this straight- he's using a plugin that is meant to create aliasing distortion, and is wondering why there are "artefacts"?

Well, that depends on what is meant by "bit crushing". I.e. if it's a quantization of sample values (something like: out[n] = ((int)in[n]) & ~0xFFF) or if it's a sort of sample-and-hold (something like out[n] = in[n & ~3]).

The latter will of course alias like hell, because it's effectively sampling at 4x lower frequency without bandlimiting. Yikes. But the former shouldn't be THAT bad. I mean - it will alias, sure, but not necessarily so much that it appears at LF. Depends on the input signal, of course.

That's mad- what does he expect? Surely there's more to it than that? Is the issue that the alias tones are LF?

Probably - at least a LF artefact in HF signal stands out pretty badly. And there's no guarantee about the input signal and the artefact being in tune - which probably makes things worse.

Anyway- my point is that if we could hear these "artefacts" we'd probably recognise them straight away.

No doubt.

Edit: he doesn't mention bitcrushing at all?? He says "I'm getting an artefacting problem when I make some stuff in Ableton, and then export the WAV, and begin working on it."

See this:

krd wrote:

You can see it in action if you apply a down sampling or bit crushing effect - you're doing a Digital to Digital conversion, but you're going well beneath the Nyquist frequency - and since the digital signal is in steps, when you sample well below the Nyquist frequency the frequency is corrupt (some steps are longer than they should be)

I overlooked the downsampling part - the "some steps are longer" probably means sample&hold which will produce loads of aliasing.

Why wouldn't it be "hamonic"? What does that even mean?

I mean something like a sampled periodic signal. A signal with a bunch of frequencies that are all multiples of a single, common fundamental. A signal that, passed through a perfect reconstruction filter, will produce a periodic signal.

Let me try and reformulate the point.

Case 1: Before bit-crusher, you have a signal that consists of two sine waves at F and 2F. If you pass this signal through a bit crusher / any other waveshaping, the intermodulation will produce a large number of sine waves at different integer multiples of F. The only frequencies that will *not* be multiples of F will be due to aliasing (when n*F > Nyquist). So, in this case, you can point at the "offending" LF artefact and say "this frequency is here because of aliasing" and noone in their right mind will question that.

Case 2: You have a different signal, also two sine waves, but the ratio of frequencies is not a small integer or even a rational number with small numerator and denominator. Say, 1000 Hz and 1050 Hz. This time, the intermodulation will produce frequencies like 50 Hz, 100 Hz - LF content also perceived as artifacts. In this case, the 50 Hz artifact is not due to aliasing, and increasing sample rate won't help. There may also be aliasing, but at least aliasing is not the only reason why the artifacts happen.

In other words - I tried to say that if you have a simple periodic signal going into bitcrusher, you have enough knowledge to say that the LF artifact must be aliasing. If you have a more complex input signal, then you probably need to take a closer look.

(ok, technically speaking, 1000 Hz+1050 Hz mix is also periodic but the fundamental is 50 Hz)

By "may" do you mean- it's up to the designer to sort those out?

AFAIK, theoretically, the bandwidth in FM is infinite (amplitudes follow some Bessel function that I have no clue about ) - but the amount of HF content depends on:
- the waveform shape of carrier and modulator
- the ratio between modulation frequency and carrier frequency
- and the modulation index (amount of the modulator signal controlling the carrier frequency or phase).

Hey, *you're* an expert on this, aren't ya? :P

I don't think it's possible to simulate FM in a way that prevents audible aliasing from occurring in first place. There's no FM counterpart of BLITs and minBLEPs, as far as I know. The only way to deal with it is by increasing the sampling rate.

However, a bad design/cutting corners may add more aliasing - see the difference between Csound's foscil / foscili / foscil3 (sp? never used the last one).

BTW there's a very good thread on GS where they analyse various plugins for aliasing etc.

Must read it - haven't seen it before, as I'm pretty much allergic to GS. Thanks!

splitrmx

See what you did Paul!

krd

madtheory wrote: »

But you're right that krd's problem is unlikely to have anything to do with sampling. But there's no way we can diagnose the problem until we hear these mysterious artefacts he speaks of!

http://soundcloud.com/krdnoobian/volatile-proton-noobian

You can hear the artefacting through out most of the first have of the track.

Between 2.24 - 2.30, there's synth playing an arpeggio. You can her this ugly sounding distorted crush sound. The arp is meant to be very dry and sharp. There's not meant to be any noisy fuzz sound.

If you skip, to 4.40, you can hear the same arp again, except this time I put it through a reverb, and the artefacting is not there.

kfoltman

krd wrote: »

Between 2.24 - 2.30, there's synth playing an arpeggio. You can her this ugly sounding distorted crush sound. The arp is meant to be very dry and sharp. There's not meant to be any noisy fuzz sound.

You mean the harsh noisy stuff on the 3 lowest notes of the arpeggio?

Any chance it's an artifact from mp3 encoding? (probably not, you mentioned it's in the original WAV as well)

kfoltman

Honestly, it's hard to tell what's damaged by mp3 and what's damaged by the plugin. But, to me, it doesn't look like a "victim" of bitcrushing (no "teeth" or "steps" in waveform) or sample rate reduction (the harmonic peaks are all evenly spaced until 16 kHz). It sounds somewhat familiar though, lo-fi quality resembling the Subtractor synth in Reason. The lower end of the spectrum is indeed full of nasty noise.

With bitcrushing or sample'n'hold I would expect the higher notes to be distorted more.

No idea what it is, ringmod with white/brown noise?

krd

kfoltman wrote: »

You mean the harsh noisy stuff on the 3 lowest notes of the arpeggio?

Yep, that sound.

Any chance it's an artifact from mp3 encoding? (probably not, you mentioned it's in the original WAV as well)

It's in the original wave - I'm not sure if it's happening when it's rendered in Ableton, or when I put it in my sound editor (I do a little hard limiting in Goldwave - and make the mp3 from there). The original sound in Ableton doesn't have the artefact.

Sound cloud make it sound a little worse. They're putting the file through their own compression/limiting - and I've not control over what they're doing. It's annoying - I wanted the first half of the track at lower volume, and they've made nearly the entire track the same volume.

i57dwun4yb1pt8

for phucks sake Brewer ,
you wouldnt let it lye would ya :pac::pac::pac:


i do however like that clasp yoke

Hayte

krd wrote: »

Yep, that sound.



It's in the original wave - I'm not sure if it's happening when it's rendered in Ableton, or when I put it in my sound editor (I do a little hard limiting in Goldwave - and make the mp3 from there). The original sound in Ableton doesn't have the artefact.

Sound cloud make it sound a little worse. They're putting the file through their own compression/limiting - and I've not control over what they're doing. It's annoying - I wanted the first half of the track at lower volume, and they've made nearly the entire track the same volume.

Its hard to make any comparison when theres lots of different sounds playing at the same time because of masking and other auditory phenomena that change the way you listen to sound.

Can I also ask why you think this is aliasing? One way you can tell is by dialling in only 1 oscillator, muting everything else so only that synth plays. If its a sawtooth wave then it will have all odd and even partials. To get the amplitude of the partials you need to do the fourier transform, or if you suck at math you can bag an FFT analyser and just read it off the graph because it does the math for you in realtime. But generally the higher the frequency, the lower the amplitude.

So if you play a 5khz sawtooth, the 1st partial is 5khz, the 2nd is 10khz, 3rd is 15khz, 4th is 20khz and 5th is 25khz and the amplitude is generally going down the higher up you go. With a fundamental that high you will be playing roughly 6 octaves up from middle c (c4) which is miles higher than the highest note of your arp.

At 48khz sampling rate the 5th partial will alias at 23khz if we consider that there is no oversampling occuring, the oscillator is not band limited (or band limited enough?), and the converter's anti aliasing isn't effective. And god knows what else but thats all I can think of right now.

You can't hear that 5th partial (even if it aliases) but it will show up on an FFT. I still don't see it though.

Until you mute everything except the synth and do a proper comparison, I'd be more inclined to believe its white noise or oscillator cross modulation or something like that. You only notice it at certain points in the track due to the effect auditory masking. But I don't know what synth you used or how it was programmed (or if it even has a noise generator, cross mod capability or really fast LFOs). So...

madtheory wrote: »

Who said we're learning this in order to make better music?

Its obvious because everyone here is asking questions in a musical context, using words. If you want to get under the engineering principles then you need to start talking in math and the thread will start to look like something on physicsforum.com where you set out the problem, all known data and the appropriate fourier series equations. Then you go crunch numbers and furnish your proofs in...math!

Seziertisch

http://www.neilyoung.com/

krd

Hayte wrote: »

Its hard to make any comparison when theres lots of different sounds playing at the same time because of masking and other auditory phenomena that change the way you listen to sound.

The problem is the whole way throughout the track, but between 2.24 - 2.30, it's just the dry synth playing. There shouldn't be any noisy smush in the arpeggio. There aren't any other sounds to mask it. It should be nearly irritatingly dry and crisp. There's a kind of squishy mush noise that occurred somewhere in rendering the track.

http://soundcloud.com/krdnoobian/vol...proton-noobian

Its obvious because everyone here is asking questions in a musical context, using words. If you want to get under the engineering principles then you need to start talking in math and the thread will start to look like something on physicsforum.com where you set out the problem, all known data and the appropriate fourier series equations. Then you go crunch numbers and furnish your proofs in...math!

There's no way of seeing the precise math a process is using or not. I'm playing back something on my DAW this morning, and one of the synths is doing something different every time it's played.

danjokill

i just creamed myself when I watched that neilyoung site intro ............ SWEET JESUS!!!!!!!!!!!!!!!!!!!!!!!