Yes that is what I mean of course, samples/lolliepops shaped like a perfect saw. A perfect saw probably doen't even exist in classical mechanics/ physicscron wrote:There is no such thing as a perfect saw in the digital domain. Such a thing would require an infinite sampling rate, as a perfect saw contains frequencies all the way up to infinity. Sure, you could draw samples in the shape of a perfect saw, but the D/A reconstruction of those samples doesn't simply 'join the dots' in linear fashion.Dúnedain wrote:thank you
But even after Monty I still do not understand it: why does a perfect saw already in digital domain alias?
How to get a smooth saw in Spire
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
Dúnedain
- KVRAF
- 3054 posts since 25 Apr, 2011
In other words; the sample rate, or resolution of "parts per second", or slices of data per second, isn't enough to make "a perfect" saw, or pulse, or square. This why the amount of aliasing is going down if you use higher sample rates (192KHz for instance). Technically, there is still aliasing, so it isn't perfect even at those high sample rates, but we can't hear it, because the aliasing is way to high in the high register for us to notice.cron wrote:There is no such thing as a perfect saw in the digital domain. Such a thing would require an infinite sampling rate, as a perfect saw contains frequencies all the way up to infinity. Sure, you could draw samples in the shape of a perfect saw, but the D/A reconstruction of those samples doesn't simply 'join the dots' in linear fashion.Dúnedain wrote:thank you
But even after Monty I still do not understand it: why does a perfect saw already in digital domain alias?
- KVRAF
- 3054 posts since 25 Apr, 2011
what do you mean with that sentence?? "why would a "perfect" saw already...." <--Dúnedain wrote:Whatever the sample rate: why would a "perfect" saw already in digital domain alias?????
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.exmatproton wrote:what do you mean with that sentence?? "why would a "perfect" saw already...." <--Dúnedain wrote:Whatever the sample rate: why would a "perfect" saw already in digital domain alias?????
Dúnedain
-
- KVRAF
- 3477 posts since 27 Dec, 2002 from North East England
You can not have a perfect saw in the digital domain. The sampling theorem only works properly when signals are bandlimited to half the sampling rate (i.e. nothing above 22.05 kHz at a 44.1 kHz sampling rate). You can't look at the sample points in a digital file and immediately understand what the reconstructed analog signal will look like from those points. It's not as simple as 'joining the dots'. Sure, it's intuitive up to a point, but try playing a saw wave in the 10kHz plus range if you want to take this to extremes. After reconstruction it'll look like a sine wave. The actual sample points (the 'lollipops' in Monty's terminology) will look like noise.Dúnedain wrote:Whatever the sample rate: why would a "perfect" saw already in digital domain alias?????
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
This is what I mean by a "perfect" saw:
So yes: you can have a "perfect" saw in digital domain. And Spire claims it can generate it, but it can notDúnedain wrote:
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.
Dúnedain
-
- KVRAF
- 3477 posts since 27 Dec, 2002 from North East England
Good luck listening to that without converting it to analog first.Dúnedain wrote:This is what I mean by a "perfect" saw:
So yes: you can have a "perfect" saw in digital domain.Dúnedain wrote:
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
And that is one of my questions: what will go wrong when converting to analog?cron wrote:Good luck listening to that without converting it to analog first.Dúnedain wrote:This is what I mean by a "perfect" saw:
So yes: you can have a "perfect" saw in digital domain.Dúnedain wrote:
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.
Dúnedain
- KVRAF
- 3054 posts since 25 Apr, 2011
Spire claims nothing??Dúnedain wrote:This is what I mean by a "perfect" saw:
So yes: you can have a "perfect" saw in digital domain. And Spire claims it can generate it, but it can notDúnedain wrote:
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.
It shows something in a very small, low res screen. It is just a 'global' representation, not exact science.
You know, Spire is an awesome synth to make some music with I suggest you should use it that way.
If you want some proper science equipment to do exact science, you'll need different gear
-
- KVRAF
- 3477 posts since 27 Dec, 2002 from North East England
Because D/A conversion isn't just 'joining the dots' with straight lines. Digital audio would sound pretty terrible if that was the case! It may 'look' like you're joining the dots when you're dealing with lower frequency signals, but that swiftly gets blown out of the water once you get into high frequencies. Take the below image of a very high frequency sine wave approaching Nyquist. The top is what is *actually* reconstructed from the sample points, whereas the bottom is what you'd get if you just joined the dots with straight lines.Dúnedain wrote:And that is one of my questions: what will go wrong when converting to analog?cron wrote:Good luck listening to that without converting it to analog first.Dúnedain wrote:This is what I mean by a "perfect" saw:
So yes: you can have a "perfect" saw in digital domain.Dúnedain wrote:
after the highest value the next sample is the lowest value of the saw. All other values have linear incremention.
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
Ah come on. Whenever you turn one of the knobs the display immediately updates. But when you choose a clean saw the pictures displays the wrong saw.exmatproton wrote:
Spire claims nothing??
It shows something in a very small, low res screen. It is just a 'global' representation, not exact science.
What do you think the output of the oscillator will be? The output is in fact a "picture" (no not a freq spectrum diagram).
Last edited by Dúnedain on Sun Dec 04, 2016 9:23 pm, edited 1 time in total.
Dúnedain
-
- KVRist
- Topic Starter
- 446 posts since 14 Dec, 2014
Thanks, I already knew that. But again: what goes "wrong" when going from a "perfect" saw in digital domain to analog? What aliasing will be introduced?cron wrote:
Because D/A conversion isn't just 'joining the dots' with straight lines. Digital audio would sound pretty terrible if that was the case! It may 'look' like you're joining the dots when you're dealing with lower frequency signals, but that swiftly gets blown out of the water once you get into high frequencies. Take the below image of a very high frequency sine wave approaching Nyquist. The top is what is *actually* reconstructed from the sample points, whereas the bottom is what you'd get if you just joined the dots with straight lines.
- KVRAF
- 3054 posts since 25 Apr, 2011
Listen. Spire never claims a perfect saw. It is 'their' saw. Compare it to hardware devices. Often you can see drawings of the waveforms on hardware synths. Those are "perfect" representations of theoretical waveforms. If one would record a saw from an hardware synth, changes are huge those aren't "perfect" as well.Dúnedain wrote:Ah come on. Whenever you turn one of the knobs the display immediately updates. But when you choose a clean saw the pictures displays the wrong saw.exmatproton wrote:
Spire claims nothing??
It shows something in a very small, low res screen. It is just a 'global' representation, not exact science.
What do you think the output of the oscillator will be? The output is in fact a "picture" (no not a freq spectrum diagram).
Besides, the main output is also different when a filter is added for instance.
"Thanks, I already knew that. But again: what goes "wrong" when going from a "perfect" saw in digital domain to analog? What aliasing will be introduced?"
Aliasing is when a soundsource has higher frequency content then nyquist. Because of the steep filter at nyquist, content that is higher pitched then that, will be trown downwards. Mirrored in some sense. This will create a-harmonic content, usually heard in the higher regimes.
Last edited by exmatproton on Sun Dec 04, 2016 9:43 pm, edited 2 times in total.
-
- KVRAF
- 3477 posts since 27 Dec, 2002 from North East England
Right, apologies. I'm going to have to leave that to someone with a bit more knowledge than me. I can only really tell you what it looks like in the frequency domain (mirrored at nyquist) rather than the time domain.Thanks, I already knew that. But again: what goes "wrong" when going from a "perfect" saw in digital domain to analog? What aliasing will be introduced?