can of worms

[Shadow_7]

I guess that's the advantage of being poor. I buy as I go. When I encounter an issue, I address that issue. If it requires a purchase, I make it. But not until I have an issue that needs to be addressed.

Buy used, sell used, minimal investment. At least for hardware. But I'm software developer (or used to be), so if I need something there, I can kind of widdle my way through making my own stuff. Not exactly a great use of time, reinventing the wheel and all. But I don't have to worry about my jaguar being of no use anymore, because they changed the type of fuel that's available (software compatibility). Plus that whole custom thing. Not that I desire to write the instruction manual when it comes time for someone else to drive my car. But I could, although if I have my way, it'll just drive itself. If I didn't have lazy tendencies and minimal understanding of certain fields of study. But I have lots of time in this economy.

 

[Shadow_7]

i guess i dont see how a computer couldnt perfectly match and wave shape...

It's called samplerate. The computer takes snapshots over time(ADC). It doesn't have the full picture. As in it's picture is a paint by number or connect the dots type image. The hardware would be more of the painter with good eye and a perfect photograph to use for comparison. The computer just connects dot to dot with as straight a line as possible. Which at certain resolutions means that your car has square tires, not round ones. As far as the computer / software is concerned. As in at 44.1kHz you only have one sample for the peak and one for the valley of a 22050 frequency. Which could be a sine wave, a square wave, or sawtooth wav. Based on that resolution all it sees is one sample per peak and one per valley. And just assumes a sine wave in the absence of more information. The hardware has infinite samplerate and can respond accordingly. As long as it stays analog and doesn't do an AD / DA conversion. i.e. Not always the case.

 

[bouldersoundguy]

i know a guy with a studio, and since im a student. he let me put this shit on my computer for educational purposes.

Unless there's something in the EULA that allows for that you're using pirated software.

so i don't really need books...

Look up non sequitur. Nothing about knowing a guy with a studio who let you pirate some software means you don't need books.

i dont know why nobody can give me a straight answer as far as to whats happening to the actual sound wave with analog hardware vs software.... i guess i dont see how a computer couldnt perfectly match and wave shape...

We have. To expand on my previous post: In some cases a plugin uses the same math as the hardware, if the hardware is digital and the manufacturer makes or authorizes a plugin version. But even then the ADDA and cabling can alter the sound of the hardware in ways the plugin won't duplicate. In the case of analog hardware, while it's theoretically possible to simulate it to a very high degree, it may take more processing power than is available or practical. It's like finding the area of a circle. You can use any approximation of pi in your formula depending on how precise you need to be and how much calculation you feel like doing, but you'll never find the actual value for the area because it would take infinite processing power and time.

 

[moresound]

Unless there's something in the EULA that allows for that you're using pirated software.

That's what I thought.

 

[bouldersoundguy]

It's called samplerate. The computer takes snapshots over time(ADC). It doesn't have the full picture. As in it's picture is a paint by number or connect the dots type image. The hardware would be more of the painter with good eye and a perfect photograph to use for comparison. The computer just connects dot to dot with as straight a line as possible. Which at certain resolutions means that your car has square tires, not round ones. As far as the computer / software is concerned. As in at 44.1kHz you only have one sample for the peak and one for the valley of a 22050 frequency. Which could be a sine wave, a square wave, or sawtooth wav. Based on that resolution all it sees is one sample per peak and one per valley. And just assumes a sine wave in the absence of more information. The hardware has infinite samplerate and can respond accordingly. As long as it stays analog and doesn't do an AD / DA conversion. i.e. Not always the case.

I used to think it was something like that until I was corrected by those who actually know what they are talking about.

Audio sampled at 44.1kHz does not contain 22.05kHz. There is at least one filter in the ADC that removes information above 20k. The sampling frequency is always more than double the represented audio frequency, so you always have on average more than two samples per cycle. That is sufficient to avoid the situation you describe.

In the DAC there are filters that reproduce the waveform, imperfectly but much better than you think. The samples are not ever connected by straight lines except on your computer screen.

 

[phriq]

so i don't really need books ...

Not to be a douche, but you can give a guy all the parts needed to build a car, but unless you give him the manual, he will never figure out how to build that car and make it run. Or at least make it run well and how it was supposed to. I used to think that I could just "figure it out" but honestly it just doesn't work that way. That's the honest truth. Regardless if you have all the tools or not, you need to learn how to use them.

i guess i dont see how a computer couldnt perfectly match and wave shape...

The basic reason is simply the way sound, mics, and computers work. Sound is simply pressure variations in the air. Cables and computers don't run with air, SOO through a transducer (like a microphone) the sound energy is converted. A mic converts this energy in the form of electric impulses that mimic those of the air pressure variations. However, if you just sent electrical impulses to a computer, you would fry it like a freshly caught flounder. SOOO once again you convert the line level signal (through A/D, analog to digital, conversion) to a digital (binary) signal. Computers run on binary, simply 1's and 0's. That is why a computer cannot match a wave shape.

The 1's and 0's make up your sample rate and bit rate. If your audio is set to 44.1khz for its sample rate, it means that in a second there are (i am not sure exactly on my math) but i think 44,100 samples for every second of audio. Those are indivdual snapshots of the wave. The computer then "connects the dots" to create an image of a sound wave. Information as far as the attributes of the wave are stored in the bit rate. 16 bit, 24 bit. This is basically the digital length of an audio word or information. a bit is the single numeric value for binary, either a 1 or a 0. so if you have 16bit, thats 65,536 bit combinations possibilities for storing information. 24 bit equals 4,294,967,296. Quite the improvement.

The computer then takes the bit and sample information and then sends it out to a D/A converter (like an interface or soundcard) and creates another electrical impulse signal utilizine that bit and sample info. it then is fed through another transducer (a speaker) which creates sound pressure variations mimicing the electric impulses... Sound.

This was alot to explain, and would be all better explained from a book. Another good reason why you really should look into getting one on recording and sound. Hope this answers your question though.

 

[DucesWild]

That's what I thought.

look, i dont know alot about this stuff yet.... im a music student and am learning, i gave my computer to a guy i know, and now i have a plugin selection called waves in my daw. if i was given stolen software I apologize. if i find out it is stolen i will likely remove it due to the fact that its prolly not even the right thing then but more likely something synister. ive had it on for a little over a week now and my computer seems to be running fine, but i cant say ive even tried to use the program at all.

 

[DucesWild]

Not to be a douche, but you can give a guy all the parts needed to build a car, but unless you give him the manual, he will never figure out how to build that car and make it run. Or at least make it run well and how it was supposed to. I used to think that I could just "figure it out" but honestly it just doesn't work that way. That's the honest truth. Regardless if you have all the tools or not, you need to learn how to use them.



The basic reason is simply the way sound, mics, and computers work. Sound is simply pressure variations in the air. Cables and computers don't run with air, SOO through a transducer (like a microphone) the sound energy is converted. A mic converts this energy in the form of electric impulses that mimic those of the air pressure variations. However, if you just sent electrical impulses to a computer, you would fry it like a freshly caught flounder. SOOO once again you convert the line level signal (through A/D, analog to digital, conversion) to a digital (binary) signal. Computers run on binary, simply 1's and 0's. That is why a computer cannot match a wave shape.

The 1's and 0's make up your sample rate and bit rate. If your audio is set to 44.1khz for its sample rate, it means that in a second there are (i am not sure exactly on my math) but i think 44,100 samples for every second of audio. Those are indivdual snapshots of the wave. The computer then "connects the dots" to create an image of a sound wave. Information as far as the attributes of the wave are stored in the bit rate. 16 bit, 24 bit. This is basically the digital length of an audio word or information. a bit is the single numeric value for binary, either a 1 or a 0. so if you have 16bit, thats 65,536 bit combinations possibilities for storing information. 24 bit equals 4,294,967,296. Quite the improvement.

The computer then takes the bit and sample information and then sends it out to a D/A converter (like an interface or soundcard) and creates another electrical impulse signal utilizine that bit and sample info. it then is fed through another transducer (a speaker) which creates sound pressure variations mimicing the electric impulses... Sound.

This was alot to explain, and would be all better explained from a book. Another good reason why you really should look into getting one on recording and sound. Hope this answers your question though.

when i said i dont need books. its because i have teachers, and would rather learn from them then a book.

and yes that you very much for your detailed answer. all i realy needed to hear was that first paragraph, but the rest was great knowledge...

 

[bouldersoundguy]

However, if you just sent electrical impulses to a computer, you would fry it like a freshly caught flounder. SOOO once again you convert the line level signal (through A/D, analog to digital, conversion) to a digital (binary) signal. Computers run on binary, simply 1's and 0's. That is why a computer cannot match a wave shape.

No, digital computers are inherently digital, so they just can't store or process analog signals. Whether or not the analog signal harms a computer, the real problem is that it's simply an incompatible signal format.

The 1's and 0's make up your sample rate and bit rate. If your audio is set to 44.1khz for its sample rate, it means that in a second there are (i am not sure exactly on my math) but i think 44,100 samples for every second of audio. Those are indivdual snapshots of the wave. The computer then "connects the dots" to create an image of a sound wave. Information as far as the attributes of the wave are stored in the bit rate. 16 bit, 24 bit.

No, the attributes of the wave are represented by the combination of the sampling frequency and the word length (bit depth). They are the two dimensions of the wave, sampling frequency for time and word length for amplitude.

 

[phriq]

No, digital computers are inherently digital, so they just can't store or process analog signals. Whether or not the analog signal harms a computer, the real problem is that it's simply an incompatible signal format.



No, the attributes of the wave are represented by the combination of the sampling frequency and the word length (bit depth). They are the two dimensions of the wave, sampling frequency for time and word length for amplitude.

Thanks for clearing that up.

 

[Chili]

i know a guy with a studio, and since im a student. he let me put this shit on my computer for educational purposes....

That's called stealing and we don't offer help to thieves. The Waves bundle requires an iLock which probably means your buddy with the studio also has stolen software if he can load it up on your computer.

 

[Shadow_7]

I used to think it was something like that until I was corrected by those who actually know what they are talking about.

Audio sampled at 44.1kHz does not contain 22.05kHz. There is at least one filter in the ADC that removes information above 20k. The sampling frequency is always more than double the represented audio frequency, so you always have on average more than two samples per cycle. That is sufficient to avoid the situation you describe.

In the DAC there are filters that reproduce the waveform, imperfectly but much better than you think. The samples are not ever connected by straight lines except on your computer screen.

I think you need to take inventory of your samples. Not all ADCs filter out content < 20Hz and > 20kHz. I'd wager that most (at least < $500) don't at all. 16 bits, X thousand(kHz) samples per second. So 2 bytes times number of seconds times samplerate. Will be awfully close to the actual file size. With some extra in there for file headers and meta data.

In terms of digital editing, yes it is DOT to DOT as far as the software is concerned. But there is a lot of fancy math in there to try and treat it otherwise. Now for reproduction over a speaker, it is NOT dot to dot. It's a mechanic device that has mass and momentum. It would be impossible, or at least logistically illogical to put brakes on it to stop at each dot. But if the frequency that is being attempted to be reproduced can't overcome the mass of the mechanical device, it will only be a failed attempt to reproduce that sample. But it'll try, there's not much on the speaker end to tell it not to. At least on the cheaper stuff in use by home recordists. Or your clients who's best stereo is in their car, and it came with the car at the time of purchase.

As I look at the specs of my Korg MR-1000. up to 40kHz at 192kHz. up to 100kHz at DSD. Which is a function of the preamp. My MM-1 preamps are speced 10Hz to 50kHz or something like that. And if those results are resampled digitally, will it have 22050Hz samples? Not everyone captures at the LOWEST common denominator.

 

[bouldersoundguy]

Not all ADCs filter out content < 20Hz and > 20kHz. I'd wager that most (at least < $500) don't at all.

Actually, they do all filter content above 20kHz at 44.1/48kHz. In the old days it was an analog brick wall filter. More recently, with oversampling converters, it's a more gentle slope in analog and additional filters in the digital domain.

In terms of digital editing, yes it is DOT to DOT as far as the software is concerned. But there is a lot of fancy math in there to try and treat it otherwise. Now for reproduction over a speaker, it is NOT dot to dot. It's a mechanic device that has mass and momentum. It would be impossible, or at least logistically illogical to put brakes on it to stop at each dot. But if the frequency that is being attempted to be reproduced can't overcome the mass of the mechanical device, it will only be a failed attempt to reproduce that sample. But it'll try, there's not much on the speaker end to tell it not to. At least on the cheaper stuff in use by home recordists. Or your clients who's best stereo is in their car, and it came with the car at the time of purchase.

A speaker can act as a filter, but that's irrelevant because it wouldn't be an adequate one for DAC. That's why there are filters in the DAC to reconstruct the waveform. Even in DSP, like eq-ing and mixing, I suspect the waveform has to be reconstructed in some sense or the sound would be compromised. Consider a delay that equals some fractional number of samples. If the process simply chose new sample points along the straight lines connecting the original samples the shape of the wave and the resulting sound would be drastically altered. But if you've applied delays before you probably couldn't hear the difference. That suggests to me that the DSP accounts for the shape of the wave as it would be when reconstructed.

As I look at the specs of my Korg MR-1000. up to 40kHz at 192kHz. up to 100kHz at DSD. Which is a function of the preamp. My MM-1 preamps are speced 10Hz to 50kHz or something like that. And if those results are resampled digitally, will it have 22050Hz samples? Not everyone captures at the LOWEST common denominator.

Notice that at 192kHz it doesn't capture anything near 98kHz (the Nyquist frequency). While 192kHz can represent such high frequencies, that's not really the benefit. The benefit is that the ADC LPF can be much gentler. For the converter to be flat the 40kHz it has to give up some of that advantage and use a steeper filter from 40kHz up rather than a gentler one from 20kHz. And if you down sample to 44.1kHz then it has to be filtered to remove everything from somewhere below the Nyquist frequency up. When your Nyquist frequency is 22.05kHz there's not much room for discretion without going below 20kHz.

 

[Supercreep]

Well, that went swimmingly.

 

[moresound]

Well, that went swimmingly.

Like a lead weight!

 

[Shadow_7]

Actually, they do all filter content above 20kHz at 44.1/48kHz. In the old days it was an analog brick wall filter. More recently, with oversampling converters, it's a more gentle slope in analog and additional filters in the digital domain.

Prove it? Practical tests. IF you have monitors capable of generating the pitch, and mics capturing that range of sounds. TEST what you're advocating. SOME software might brickwall filter, not NOT ALL will. Also note that the limits I stated are limits of the PREAMPS, not the samplerate, bits, or anything else in the chain. But it should be simple enough to TEST. If you have the right gear. I don't currently have monitors capable of those frequencies so I can't currently do that sort of test. Although my PC speakers can go up to 17kHz. So I guess I could record at 22050Hz to test 10000 to 11025Hz. But that doesn't test the 20kHz limit you seem to be advocating. A limit of AVERAGE human hearing, not the gear used to record.

 

[Shadow_7]

I just proved it to myself.

limit_20kHz_test.sh

A little script to generate content with 20kHz, 21kHz, and 22kHz sine waves and others. Using mostly sox. If it was a limit, it couldn't have generated said content. It does seem to break up near 22kHz at the point of generation, but 21kHz is NO PROBLEM. So I generated at 48kHz to have a clean 22kHz generation. I took that over to my Korg MR-1000. Played it from the Korg while Recording that playback on my Mobile Pre. And what do you know 21kHz recorded just fine. Came in at 21013 according to audacity's plot spectrum analysis thing. So I'm certainly not seeing this "LIMIT" you keep hinting at. It does break up a bit past 20kHz in terms of gain levels and oddities, but IT IS THERE. Certainly not a limit of my ADC and DAC abilities. And per factory specs my preamps have no problem according to the manufacturer exceeding that 20Hz to 20kHz range. 10Hz - 50kHz for my MM-1's. And at least 21kHz will reach and be sampled by my interface and recording software at 44.1kHz. A test proved it.

Not to say that some fancy mic that converts to digital and broadcasts wirelessly doesn't do what you were hinting at. But IME most gear, at least on the low end, don't discriminate on things outside the range of human hearing, and within the range of the sample rate. The extreme end is always degraded, but not to the 10% or more of 44.1kHz that you seem to believe. There is certainly no brickwall filter on my current gear.

 

[bouldersoundguy]

Okay, let me rephrase and refine my point. At and above the Nyquist frequency (half the sampling frequency) the signal must be sufficiently attenuated to prevent aliasing. Anti-aliasing LPFs are essential for audio ADCs. You can't omit them and have good sound. If you don't trust me look it up for yourself.

Where the filter starts is largely a matter of what's technically possible. Early on the filters were analog and there were limits to the slope, so starting as low as possible (e.g. 20k) was unavoidable. Newer oversampling converters allow most of the filtering to be done in the digital domain. Much steeper slopes are possible, allowing them to start closer to the Nyquist frequency.

Down-sampling requires anti-aliasing filters whether the signal was converted from analog or synthesized, so I did it with a synthesized signal. First I created a file at 96k and generated a 15-30k sweep. Then I down-sampled it to 48k and 44.1k, up-sampling it back to 96k so the display would show the higher frequencies for clarity. I left the anti-aliasing filter option checked because that's how ADCs do it.

15-30kHz sweep at 44.1kHz


15-30kHz sweep at 48kHz