ArrayArrayArray
BrainModularBrainModular Users Forum2017-01-20T19:57:03+02:00https://brainmodular.com/forums/app.php/feed/topic/56572017-01-20T19:57:03+02:002017-01-20T19:57:03+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36635#p36635Statistics: Posted by 23fx23 — 20 Jan 2017, 18:57
]]>
2017-01-20T16:31:58+02:002017-01-20T16:31:58+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36634#p36634 from the wiki ( https://en.wikipedia.org/wiki/Digital-t ... _converter ): Instead of impulses, a conventional practical DAC updates the analog voltage at uniform sampling intervals, which is then interpolated via a reconstruction filter to continuously varied levels.
And what you had to have in mind human ear is something like high pass filter on 20Hz and lowpass filter on 20kHz. And the thing is that your brain sees sound like "spectrogram". Sound goes to the drum as mplitude, but behind drum there is a snake which does "FFT". so, practically if you have 15kHz square signal, there is for our brain in fact no difference between 15kHz square/triangle/sine because our brain is not able to see in this frequency higher harmonics - we hear just the first sine (second harmonics is on 2*15=30kHz which is far behing the "low pass" filter (hence Nyquist/Shannon-Kotelnik teorem) and because of this higher frequencies there is as well high pass filter on 20Hz on DACs - because higher harmonics can mirror into low frequency signals.
What i am trying to say, is you don't need to know exact signal behing the DAC because our hearing apparatus is doing some "postprocessing" as well
]]>2017-01-19T10:06:07+02:002017-01-19T10:06:07+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36630#p36630 ie here the samples true values are the binary sequences dots, we can see the resulting wavform will be kind of highly different.
for exemple, vu meters mostly consider sample values and will indicate -6db, whereas in reality it can reach over 0dbfs spikes intersamples, or if i try to normalise that it will takes the sample values as computation. id like to have a control of this and perform operations in an usine module/script based on the 'real' output estimation, not the samples themselfs.
so to sumup: what is formula, given an array of samples values (ie dots) ,to generate a higher length one giving the real analog output curve converters representation
it seems something called spline or cubic interpolation but wonder if anyone could know/point me to the exact formula i should use to compute this? senso if you read this :p (or anyone that knows of course!)
]]>BrainModularBrainModular Users Forum2017-01-20T19:57:03+02:00https://brainmodular.com/forums/app.php/feed/topic/56572017-01-20T19:57:03+02:002017-01-20T19:57:03+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36635#p36635Statistics: Posted by 23fx23 — 20 Jan 2017, 18:57
]]>2017-01-20T16:31:58+02:002017-01-20T16:31:58+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36634#p36634 from the wiki ( https://en.wikipedia.org/wiki/Digital-t ... _converter ): Instead of impulses, a conventional practical DAC updates the analog voltage at uniform sampling intervals, which is then interpolated via a reconstruction filter to continuously varied levels.
And what you had to have in mind human ear is something like high pass filter on 20Hz and lowpass filter on 20kHz. And the thing is that your brain sees sound like "spectrogram". Sound goes to the drum as mplitude, but behind drum there is a snake which does "FFT". so, practically if you have 15kHz square signal, there is for our brain in fact no difference between 15kHz square/triangle/sine because our brain is not able to see in this frequency higher harmonics - we hear just the first sine (second harmonics is on 2*15=30kHz which is far behing the "low pass" filter (hence Nyquist/Shannon-Kotelnik teorem) and because of this higher frequencies there is as well high pass filter on 20Hz on DACs - because higher harmonics can mirror into low frequency signals.
What i am trying to say, is you don't need to know exact signal behing the DAC because our hearing apparatus is doing some "postprocessing" as well
]]>2017-01-19T10:06:07+02:002017-01-19T10:06:07+02:00https://brainmodular.com/forums/viewtopic.php?t=5657&p=36630#p36630 ie here the samples true values are the binary sequences dots, we can see the resulting wavform will be kind of highly different.
for exemple, vu meters mostly consider sample values and will indicate -6db, whereas in reality it can reach over 0dbfs spikes intersamples, or if i try to normalise that it will takes the sample values as computation. id like to have a control of this and perform operations in an usine module/script based on the 'real' output estimation, not the samples themselfs.
so to sumup: what is formula, given an array of samples values (ie dots) ,to generate a higher length one giving the real analog output curve converters representation
it seems something called spline or cubic interpolation but wonder if anyone could know/point me to the exact formula i should use to compute this? senso if you read this :p (or anyone that knows of course!)