AKAI "SynthStation" looks fun
- Thread starter SocialNumb
- Start date
Kazrog
Kazrog, Inc.
Definitely impressive. Nobody's going to be nailing the Moog ladder filter on the native audio hardware though. And square waves at 44.1 aren't very square...
drew_drummer
Dancefap
Mutant
I hate that supercow !
Actually a real analog synth can't produce a perfect square wave just like a cab can't go from -100 to +100 instantly.
And this imperfection actually sounds very good.
And this imperfection actually sounds very good.
drew_drummer
Dancefap
Yeah. That's why I was baffed. Because digital can produce perfect square waves, regardless of the sample rate. Sample rate doesn't really have much to do with how accurate a waveform can be plotted. Simplistically speaking, it's better to think of sample rate as affecting the a/d stage, and introducing aliasing when synths and samplers are not designed with oversampling in mind.
Kazrog
Kazrog, Inc.
Guys, digital synths cannot produce nearly as perfect of square waves as analog synths, because digital audio is a summation of sinewaves. While nothing is truly perfect, digital synths are "consistently" imperfect, because instead of a square wave you end up with a very steep sinewave, or more accurately a summation of several sinewaves to account for the harmonic overtones present in a square wave. The higher the resolution, the steeper and more accurate it gets.
Compare a square wave sound in a typical soft synth at 44.1k to a the same square wave at 96k. Bounce wave files out so you can a/b them. There are a variety of approaches/algorithms, so some softsynths may do this in a more audibly pleasing way, but there's no beating physics.
The good news is, at 96k or higher, you get a square wave that is audibly indistinguishable from analog on all but the most revealing/expensive systems. But, so far Apple has yet to put out a mobile device capable of 96k. iPhones, iPods, and iPads max out at 48k as far as I know. Even if they were to support 96k or higher, the cost-effective DACs they'd be likely to utilize (after all, these are mass-consumer goods geared at folks who think mp3s sound great) would almost certainly lose clocking accuracy at higher resolutions.
Further reading:
http://en.wikipedia.org/wiki/Square_wave
http://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem
Not trying to derail the thread or spoil the fun. It looks like a fun app, and really sounds good for what it is. I especially like the X/Y pad automation, how it resembles Pong. Totally awesome concept.
Compare a square wave sound in a typical soft synth at 44.1k to a the same square wave at 96k. Bounce wave files out so you can a/b them. There are a variety of approaches/algorithms, so some softsynths may do this in a more audibly pleasing way, but there's no beating physics.
The good news is, at 96k or higher, you get a square wave that is audibly indistinguishable from analog on all but the most revealing/expensive systems. But, so far Apple has yet to put out a mobile device capable of 96k. iPhones, iPods, and iPads max out at 48k as far as I know. Even if they were to support 96k or higher, the cost-effective DACs they'd be likely to utilize (after all, these are mass-consumer goods geared at folks who think mp3s sound great) would almost certainly lose clocking accuracy at higher resolutions.
Further reading:
http://en.wikipedia.org/wiki/Square_wave
http://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem
Not trying to derail the thread or spoil the fun. It looks like a fun app, and really sounds good for what it is. I especially like the X/Y pad automation, how it resembles Pong. Totally awesome concept.
Mutant
I hate that supercow !
While you are generally right on many points here (especially about the sine waves creating digital audio), you are still wrong about analog square waves being more perfect than digital ones.
Maybe it is because of different definitions of "perfect".
My definition of a perfect square wave is that first it goes straight up at exactly 90 degrees and in a straight line, then it turns 90 degrees, then another 90 degrees etc. - if you put a geometrical rectangle on it, it will fit perfectly.
Analog components can't do it, because it would require instant changes in voltages to go straight up, then instant stabilization of voltage to form the flat top, then another instant change in voltage and so on and of course they can't switch and stabilize voltages instantly.
In a digital world we can make a perfect square wave, but still the digital to analog converter in our sound cards will not render them perfectly (while still much closer to perfection than any analog synth would do).
Maybe it is because of different definitions of "perfect".
My definition of a perfect square wave is that first it goes straight up at exactly 90 degrees and in a straight line, then it turns 90 degrees, then another 90 degrees etc. - if you put a geometrical rectangle on it, it will fit perfectly.
Analog components can't do it, because it would require instant changes in voltages to go straight up, then instant stabilization of voltage to form the flat top, then another instant change in voltage and so on and of course they can't switch and stabilize voltages instantly.
In a digital world we can make a perfect square wave, but still the digital to analog converter in our sound cards will not render them perfectly (while still much closer to perfection than any analog synth would do).
So Kazrog, because I do a bunch of synth type music, I should be working @ 96k or above? Or am I way off here?
I've been asking a lot of electronic musicians and most of them work @ 44.1 or 48k. Very few of them work any higher because they claim it's a waste.
I've been asking a lot of electronic musicians and most of them work @ 44.1 or 48k. Very few of them work any higher because they claim it's a waste.
^That's what I'm thinking. If I do hear a tiny difference I can't help but toss it up to the placebo effect.
That's a nice free synth. I'm in love with Sylenth1 and U-he ACE. I'm not sure if either are oversampled internally but they sound great. ACE is a little cpu intensive but well worth it IMO.
That's a nice free synth. I'm in love with Sylenth1 and U-he ACE. I'm not sure if either are oversampled internally but they sound great. ACE is a little cpu intensive but well worth it IMO.
drew_drummer
Dancefap
You don't need to be running your host at a higher sample rate if the synths themselves have internal oversampling. The DCAM Synth Squad plugins for instance have internal oversampling up to x32 - user sets it in the preferences.
Yes... it does make a big difference. But not really to the shape of the waveform, which is what we're discussing here. AFAIK, oversampling is indeed to combat nyquist. But nyquist again doesn't relate to the shape of waveforms - only the frequencies contained within. Oversampling prevents frequencies wrapping around the spectrum, resulting in aliasing.
But the shape of the waveform is dictated by floating point mathematics, which on a computer can easily draw a simple square wave - it's a series of 1's and 0's... that's literally all a square wave is.
Analog has voltage to take account of, and cannot go from 0 to 1 as quick as a digital system can. Look at an analog waveform in an audio editor. You'll see there is a rise and fall at the start of the square wave - this is because of the physics of analog audio. Hence a digital system produces more perfect waveform shapes than analog.
It just so happens that the more imperfect shapes are what are more pleasing to hear.
This is the way I understand it anyway. I could be wrong, and I'm open to friendly discussion on it.
drew_drummer
Dancefap
http://commons.wikimedia.org/wiki/File:Square_wave_closeup.png
There is an unfiltered digitally generated square wave.
There is an unfiltered digitally generated square wave.
drew_drummer
Dancefap
This would imply that no square wave is ever perfect. The question is... which is more perfect... analog or digital?? I'd say digital.
Kazrog
Kazrog, Inc.
Oversampling isn't going to make 44.1 capable of things it can't do. It will just filter the sound to make it more pleasing. The best any oversampling routine can hope to achieve is the effect of recording an analog synth to a higher res, then downsampling it with a nice algorithm. It can sound great, really, but it still isn't as square as a pure analog source.
Analog square waves are far more square than digital ones. This is measurable on an oscilloscope. Here's an article with some pictures of square waves at different resolutions, including DSD resolutions (MHz, not KHz, used in the controversial SACD format, much faster sampling rates than our interfaces are capable of, but with some accuracy loss in the top end.)
http://www.craigmandigital.com/education/PCM_vs_DSD.aspx
Analog square waves are far more square than digital ones. This is measurable on an oscilloscope. Here's an article with some pictures of square waves at different resolutions, including DSD resolutions (MHz, not KHz, used in the controversial SACD format, much faster sampling rates than our interfaces are capable of, but with some accuracy loss in the top end.)
http://www.craigmandigital.com/education/PCM_vs_DSD.aspx
Kazrog
Kazrog, Inc.
Might as well work at 96khz. If you're doing electronic music, most of your tracks will be MIDI and you won't need to bounce very often I wouldn't think, so disk space won't be an issue as much as if you were tracking a full band.
I still work at 44.1 most of the time, because with all the plugins and VSTis I need to run, it's really impractical/slow running at 96khz, even on my quad-core setup, and 48khz isn't enough of a positive difference to matter to me.
My goal in my next upgrade (in 2 years or so) is to be able to work at 96khz or even 192khz fulltime. To do the latter, I'd have to switch from Cubase to Nuendo, not a big problem but kind of a hassle.
drew_drummer
Dancefap
Don't mean to labour the point, but:
That is a digitally generated square wave in Logic at 20hz. Sample rate was 44.1khz. Looks square to me. Not a whole lot of rise time. Certainly less than any analog images I've seen.
Oversampling is the equivalent of running a plugin internally at a higher sample-rate, which is then downsampled at the output stage. This is why most oversampling features are multiples of 2 - because the math is easier. Theoretically you could oversample by 1.234 or something ridiculous... but the floating point calculations would rape your CPU.
Run at 44.1khz. If you reeeeaaally need to run at a higher rate, run at 88.2khz. The math conversion is easier, and will be kinder on your CPU.
I have to say, I find recording guitar based music MUCH easier on the CPU and RAM side of things than electronic music. Not sure what you're doing Shane for it to be the opposite way round, but that hasn't been my experience at all.
Anyway... it's nearly 1am and I need to be up in eight hours, so I'm not gonna be around to get into this. But practically speaking... I don't think it really matters, because square waves are fundamentally NEVER perfect - there is always a certain amount of rise and fall.
That is a digitally generated square wave in Logic at 20hz. Sample rate was 44.1khz. Looks square to me. Not a whole lot of rise time. Certainly less than any analog images I've seen.
Oversampling is the equivalent of running a plugin internally at a higher sample-rate, which is then downsampled at the output stage. This is why most oversampling features are multiples of 2 - because the math is easier. Theoretically you could oversample by 1.234 or something ridiculous... but the floating point calculations would rape your CPU.
Run at 44.1khz. If you reeeeaaally need to run at a higher rate, run at 88.2khz. The math conversion is easier, and will be kinder on your CPU.
I have to say, I find recording guitar based music MUCH easier on the CPU and RAM side of things than electronic music. Not sure what you're doing Shane for it to be the opposite way round, but that hasn't been my experience at all.
Anyway... it's nearly 1am and I need to be up in eight hours, so I'm not gonna be around to get into this. But practically speaking... I don't think it really matters, because square waves are fundamentally NEVER perfect - there is always a certain amount of rise and fall.
Kazrog
Kazrog, Inc.
Right, a 20hz square wave has plenty of resolution to be described adequately in your example. But once you get to higher frequencies, there is far less information to describe square waves, and the accuracy plummets.
TheWinterSnow
Den Mørke Natt
rise times and fall times in both digital and analog audio are ALL dependent on the maximum slew rate of the op-amps and transistors of the summing amplifier. Both analog and digital logic use summing amplifiers, but in order to generate a wave they are done differently
Older Analog: Generates Square wave by adding the fundamental frequency and the the first 4 harmonics (the 5th Harmonic). The maximum rise time is slightly lower that the maximum slew rate of the amplifying components.
Newer Analog: Generates Square wave by using a comparator to convert a sine wave to a square wave. The sine wave is railed to the maximum slew rate that the amplifying component.
Digital Logic: Generates Square wave by adding digital logic bits to a DAC, which is a summing amplifier. The maximum rise time is the addition of the logic slew rate pulse the summing device's slew rate.
High frequencies don't matter in terms of quality, the rise and fall time frequency will be the same on all square waves. The actual frequency of a square wave is only determined by the wave's pulse width which is determined by how long the wave holds its voltage. Regardless of digital or analog, a square wave will become a triangle wave if the frequency of the pulse width becomes half the frequency of the slew rate. And where that frequency is depends on the quality of the components used. The reality is, neither analog or digital has a perfect wave, and neither one is better than the other, since the quality of components changes how perfect the wave is. That means that a digital can be better than some analog and vice versa. Analog will sound different also because they bring other noise and distortions along with them and filtering particularity in vintage gear with a lower bandwidth.
Older Analog: Generates Square wave by adding the fundamental frequency and the the first 4 harmonics (the 5th Harmonic). The maximum rise time is slightly lower that the maximum slew rate of the amplifying components.
Newer Analog: Generates Square wave by using a comparator to convert a sine wave to a square wave. The sine wave is railed to the maximum slew rate that the amplifying component.
Digital Logic: Generates Square wave by adding digital logic bits to a DAC, which is a summing amplifier. The maximum rise time is the addition of the logic slew rate pulse the summing device's slew rate.
High frequencies don't matter in terms of quality, the rise and fall time frequency will be the same on all square waves. The actual frequency of a square wave is only determined by the wave's pulse width which is determined by how long the wave holds its voltage. Regardless of digital or analog, a square wave will become a triangle wave if the frequency of the pulse width becomes half the frequency of the slew rate. And where that frequency is depends on the quality of the components used. The reality is, neither analog or digital has a perfect wave, and neither one is better than the other, since the quality of components changes how perfect the wave is. That means that a digital can be better than some analog and vice versa. Analog will sound different also because they bring other noise and distortions along with them and filtering particularity in vintage gear with a lower bandwidth.
