Sampling Theorem/Nyquist

[ctoptrophobe]

Originally Posted by BobbyDazzler

No, basically you sample the audio at many times the desired sampling rate so you don't have to use such a nasty Low Pass Filter on the analog side, (The filter avoids the fold-down problems outlined in Nyquist theory). Oversampling lets you use a gentle analog filter, and then filter digitally to avoid the phase issues of high order analog filters.

So the sample rate stays the same and it leaves the Nyquist frequency at the same level, lets just say 44.1 kHz so 22.05 kHz Nyquist, by oversampling (lets say x2) does it set the filter for an 88.1 kHz sample rate and allow up to 44.1 kHz (Nyquist?) to be heard without alias in the original 44.1 kHz sample because it speeds up and allows for more "space" to be filled by the original sound?

[BobbyDazzler]

Oversampling is mainly a bonus on the Analog to Digital convertor side of things. And with oversampling you're generally talking about being 100+ times the sampling rate (I remember 256 times oversampling being common a few years back).
Everything above half the sample rate has to be filtered out, oversampling doesn't circumvent this, it does however allow a more gentle (and cheaper) Low Pass Filter to be used prior to the A to D Convertor.
So there is no more air above the Nyquist freq, it just means you not living with the phase issues and ringing of ultra brickwall analog filters.

[ctoptrophobe]

To accomplish this, is this only for a recording straight into the DAW like with an MBox or with a prerecorded file that needs to be sized down? As in import it at an oversampled rate and bounce it back to the neccesary rate?

[BobbyDazzler]

I'm assuming when you do a conversion it will be filtered by the program that bounces/converts.

[engr.fawad]

A noiseless 3-kHz channel cannot transmit binary (i.e. L=2) signals at a rate exceeding 6000 bps. Note that the signal is passed through the low-pass filter in order to avoid Aliasing. In 1948, Shannon carried Nyquist's work further and wanted to derive an equation for the case of a noisy channel and he was successful. The amount of noise in the channel /medium is measured by a ratio know as signal to noise ratio in which signal's power and noise present is taken into consideration. The Signal's power is represented by "S" whereas noise power is represented by "N". So, the ratio is represented by S/N. If we represent the ratio by SNR, then the value of signal to noise ratio is taken in decibels i.e SNR= 10(log10)S/N. For example SNR with S/N of 100 is 20db.

Nyquist's theorem


www.daenotes.com

[aka21stCentury]

Ask about the CD player/audio ref. system you are listening on. Setup a scope in the lab. Play a test tone - square wave. Is it filtered (rounded off?). The play a true sine wave? What does it look like? Many have patented these hardware filters, it is doubtful what you are listening to is an accurate reproduction of what was recorded. 22.05 or 44.1k

The Brickwall filter (Digital Waterfall) causes damage also to the high frequency phase and can adversely effect the conversion to math (binary) in the A-D process.

The further you magnify it (192k, 32 bit etc) the smaller the square waves become, still not perfectly portraying the analogue waveform and this is why so many are mentioning oversampling. EQ, Brickwall filters on both ends, cause phase distortion, although Harvey Holmes (father of DSP) mentioned once he had invented a phase coherent EQ, but I believe this was DSP.

Art Benade would've asked you: what compression waves are actually created by the loudspeaker's movement and what 'sound' is added by your brain? (me now: not ears... but brain). Art understood very well how the inner ear worked. It was the brain's ability to add sub-harmonics, and upper partials that puzzled him, and how compression waves somehow turn into enjoyable or disagreeable music in the brain. One can't always be certain of what you think you hear...

Arthur C. Clarke mentions this also in his 'Colors of Infinity' video on Fractals: Mandlebrot and Julius Sets.

The Pythagorian series. Do you understand this subdivision in nature? Read Keplar's Harmony Mundi for a more in depth explanation.

From experience. Chopping bits off destroys the hall ambience. While snares, bowed strings require a very fine sample rate to differentiate between the sine wave, the sawtooth and square wave high frequency component produced by the resin coated horse hair sawing away at a vibrating string whose fundamental is essentially a sine wave.

A one of a kind recording to resample at different rates: Cleveland Orchestra direct to Telefunken tube lathe. No tape. No transistors. No EQ. Recorded in an excellent hall, Masonic Auditorium. Ken Hamann's son Paul at Suma Recording, Ohio.

This is an extremely rare vinyl one off release. But if anyone has it, or the masters, it would be Paul. Two Telefunken Tube lathes recording simultaneously. Each run just a single pressing from the masters. Ken's last word to NAB & AES. Ken is no longer living.

The idea behind this is to remove enharmonic distortion producing (transistors) as well as tape compression. A more dynamically accurate recording. Previous to this a Telefunken expander was used to cut acetate masters from tape.

edit: Can you tell what type of recording this is? http://en.wikipedia.org/wiki/File:Spectrogram.png

Well Bobby Dazzler said it all in almost one paragraph. Let's see what I can do in the 7th inning stretch. Good analogue DI boxes for bass must pass 10 hz with no phase damage. Even more critical with digital. You hear so much about clocks. Well it takes 3 cycles for a human to detect the pitch of a pure sine wave. That is way too slow for groove oriented music. So what really happens? I believe we train our ear and brain to seek out the 1st and 3rd partials, and if the mix is phase coherent and these partials are in tune and audible, our brains calculate the fundamental without this delay. Theoretical acousticians don't understand this. What Art Benade or Arthur C. Clarke are suggesting and puzzled their entire careers about this sub-aural type of additive brain activity. Clarke calls them near lightspeed calculations (or near)... even going so far as to call them Mandlebrot or Julius equations. Musicians then 'hear' a sub-aural bass not mechanically possible. So phase and bandwidth (the less damage caused by brickwall filters and the range that audio may pass is key) IMO. I've seen mixers at work, moving their tiny aurotones, 4 ft spacing, across the console (actually on top of the console knobs), with no talking, no A-2, head tucked in between those phase coherent speakers until a third hand is needed for faders going to 1/2 track. Then one gets invited in.