Autofilter

[Guest]

No getting around it two samples per waveform (even if that wave form is a defacto sine) is not even in the ballpark. He was correct only in that with one sample you have ziltch. With two the hope is that each sample rate would hit a different part of the wave and we would eventually get something happening.

the shannon-nyquist-frequency in this case describes the highest possible SINE WAVE of all the SINE WAVES in a pcm-sampled audio signal. the a/d needs to "hit" the sine wave at least TWICE in one period to accurately descibe it's frequency, amplitude and hopefully phase also. if it hits a frequency only once per period you will in fact get a resulting sine wave which has a lower frequency than the original (because it's running out of phase with the a/d's clock source; the point where the a/d measures the input signal shifts along the time axis of the input sine wave). and i hope i need not tell you that every audio signal is composed of multiple sine waves.

-mike

[forge]

Aaargh, it's the Chewbacca Defense. It just Doesn't Make Sense.

none of it does mate, it's all gone to shit, if they don't get us this beta soon the forums gonna start freying at the edges and come apart...

[montrealbreaks]

I had the same problem with the autofilter and also would prefer a wider range. Maybe with some dynamic resonance level correlation. So that at the top range of the filter you get maybe no resonance at all, but have acoustic transparency. However ...the Autofilter is a very nice plug allready as it is.

I originally posted this in tips n' tricks, but what the hey, here it is again. It's a 'auto dynamic resonance correction' for the EQ 4 in filter mode. They are VERY transparent when the filters are open.

________________________________________________

1. Use the EQ 4 as a resonant high pass and low pass filter, with no resonance when fully open:
The EQ 4 makes a great resonant filter, but the problem is when fully open, as high pass or low pass, you may hear a problem: with resonance, the Low pass filter creates a resonating hiss at the highest frequency. The high pass filter, at the bottom, creates a booming muddy sound that clips. There is a way to get the filters, when open, to reduce their resonance at the highest and lowest points...

The solution? well, either bypass when not in use (requiring a keystroke or midi controller), or assign two more knobs to the resonance values, but these are not user friendly solutions.

Try using band two and three to drop the signal at the top and bottom frequencies of the EQ4. Set band two to the absolute minimum, and band 3 to the maximum frequencies. Then, set the gain reduction, and Q value of band 2 so that when band 1 is at the bottom, band 2 exactly negates band 1's resonant gain. Also, set band 3 so when band 4 is at the highest freq, the same thing happens.

Thus, you'll get a pretty much unfiltered, unresonant signal when the filters are open, effectively bypassing them. The curve diagram on the EQ 4 will help you set this up. Compare the wet sound (with the filters open) and the dry sound, and you should not be able to hear a difference if you set things up right.

Try assigning the band 1 and band 4 frequencies to knobs... You can now perform very efficient filter sweeps, and when returned to the fully opened position, the filter is effectively bypassed. So, when you grab a knob and turn it, the resonance only comes in once it passes away from the high or low point. As you move the high pass filter upwards, the resonance slowly increases, until I get around 150 Hz when it becomes noticeable. Same for the low pass filter, the resonance slowly increases as you drop the freq, and kicks in around 10,000 Hz.
Try these settings for a virtually "dry" sound when the filters are open, and increasing resonance as you sweep the filters:

Band 1: High Pass
Freq: 40 hz
Q: 2.5
Gain: 0

Band 2:
Freq: 40 hz
Gain: -8.57
Q: 1.23

Band 3:
Freq: 18.5 Khz
Gain: -8.57
Q: 1.23

Band 4: Low Pass
Freq: 18.5 Khz
Q: 2.5
Gain: 0

These settings can be adjusted to your taste, but give a VERY transparent sound when the filters are totally open...

L8er
Montrealbreaks

[montrealbreaks]

Nyquist is a Dope ! Think it through gentlemen. If you sample a signal twice per wavelength you get a very poor representation of it. This is why we are now seeing much higher sample rates, so that the highest frequencies are still sampled several times thus rendering a smoother and clearer, but still incomplete representation of the signal.

I would bet a month's salary that Nyquest was not only smarter, but more educated and successful than both you and I loophead...

You have a point, but take what Nyquest said in context. At 22.05 KHz (for a 44.1 KHz sample) you get a square wave, true. That creates harsh harmonics at even higher frequencies - yes, out of audible range, but there and enough to psycho-acoustically alter the sound. That's why A/D converters have steep filters up at the top of their range.

That's also why Nyquest stated that it was the MINIMUM frequency, not preferred or best frequency to sample at.

L8er
Montrealbreaks

[noisetonepause]

"Foundation of digital audio" or not he's still a dope. No getting around it two samples per waveform (even if that wave form is a defacto sine) is not even in the ballpark. He was correct only in that with one sample you have ziltch. With two the hope is that each sample rate would hit a different part of the wave and we would eventually get something happening.

I think you're forgetting an important thing: the smoothing filter.

Nyquist was not a 'dope', whatever that means.

-Paws

[Guest]

Try these settings for a virtually "dry" sound when the filters are open, and increasing resonance as you sweep the filters:

Band 1: High Pass
Freq: 40 hz
Q: 2.5
Gain: 0

Band 2:
Freq: 40 hz
Gain: -8.57
Q: 1.23

Band 3:
Freq: 18.5 Khz
Gain: -8.57
Q: 1.23

Band 4: Low Pass
Freq: 18.5 Khz
Q: 2.5
Gain: 0

These settings can be adjusted to your taste, but give a VERY transparent sound when the filters are totally open...

L8er
Montrealbreaks

Since the clip envelopes are relative not absolute changes, you can use a clip envelope to automate an LFO on these too. Just make it subtle, or you'll lose control.

L8er
Montrealbreaks

[Vercengetorex]

The Nyquist Theorem is only used to describe the ABSOLUTE MINIMUM REQUIREMENTS for the capture and reproduction of digital audio. It does not say that two samples per waveform is adaquate for digital capture and reproduction ad infinitum. Besides, it is not stated as requiring two samples per waveform, it is stated as two samples per Delta. Two samples per CHANGE. This is infact why higher sample rates allow for higher quality sound; A waveform does not have to cross the zero volt line before a change in vector.

I do agree that in many cases cheaper DACs do benifit from moving the anti-aliasing filter out of the audible range, but this is an issue of design quality and nothing more.

[loophead]

Well of course its the "minimum" and of course delta and of course with the minimum one needs to filter the piss out of it. And I'm also sure that Mr. Nyquist is NOT frequenting a software forum, with sweaty hands, heart palpitations and baited breath waiting for the F-ing betaaaaaaaaaa !

Nice to be amongst friends in these trying times............

[TheAnimal]

Nice tip! And I thougt it could not be automated ... stupid me