Gibb’s phenomenon

[Mulder]

I have some thoughts about gibb's phenomenon and how it is handled in today's DACs. When it comes to gibbs phenomenon and how the "ripples" can be handled, ie reduced, there is some written about windowing. But on the other hand, this means that the transition band becomes broader. Is that why many DACs have a stop band only at 24 khz rather than at 22 khz? Is it to reduce the effect of gibbs phenomenon?

 

[JohnYang1997]

Gibb's phenomenon is just a phenomenon. It's neither good nor bad.
In a band limited system it's the desired result.

To the question in the end, no. Relaxing cut off frequency allows deeper attenuation and flatter 20khzBW frequency response without sacrificing anything below 20khz while using realistic number of orders. The imaging are above 20khz so not an audible issue.

 

[Mulder]

Another issue with Gibb’s phenomenom, as I understand it, is the possibillity of distortion due to a rise in amplitude, ie if this rise are clipped. Maybe not audible, but still.

 

[JohnYang1997]

Another issue with Gibb’s phenomenom, as I understand it, is the possibillity of distortion due to a rise in amplitude, ie if this rise are clipped. Maybe not audible, but still.

If there's no clipping with multitone, it's not an issue in practice. If there is clipping with multitone, then there is a chance it can be clipping with real music.

 

[pjug]

Good design should allow some headroom for intersample overs. So this would handle minor Gibbs ripple.

 

[Mulder]

Good design should allow some headroom for intersample overs. So this would handle minor Gibbs ripple.

Yes. I have understood it so to. But most DAC:s do to my knowledge not have ”headroom”. RME and Benchmark have some headroom, but with others I don’t know.

 

[Mulder]

The real reason I started thinking about Gibb's phenomenon is that I was trying to understand what MQA really does. Bob Stuart is talking about ringing, and I was wondering if that's the effect of Gibb's phenomenon he's talking about. If so, I was thinking about whether or not MQA is mostly a way to cross the river for water.

 

[JohnYang1997]

The real reason I started thinking about Gibb's phenomenon is that I was trying to understand what MQA really does. Bob Stuart is talking about ringing, and I was wondering if that's the effect of Gibb's phenomenon he's talking about. If so, I was thinking about whether or not MQA is mostly a way to cross the river for water.

It's bull crap.

 

[pjug]

Yes. I have understood it so to. But most DAC:s do to my knowledge not have ”headroom”. RME and Benchmark have some headroom, but with others I don’t know.

If you have digital volume control this also gives you headroom to allow for these things.

 

[JohnYang1997]

1, Gibb's phenomenon is desired result in a band limited system.
2, Real music does not excite more intersample overs than multitone signal.
Nothing else matters.

 

[pjug]

Bob Stuart is talking.

These days the best thing to do is stop listening immediately

 

[Mulder]

It's bull crap.

Who is ? What is? Mine posting or MQA?

 

[NTK]

Is that why many DACs have a stop band only at 24 khz rather than at 22 khz? Is it to reduce the effect of gibbs phenomenon?

The target upper frequency range of audio DAC has always been flat to 20 kHz. With 44.1 kHz sampling rate, Nyquist frequency is 22.05 kHz, therefore the aliasing artifacts above 22.05 kHz are "reflected back" at this frequency. The aliased component reflected to 20 kHz is thus 22.05 + (22.05 - 20) = 24.1 kHz. This is the reason stop band is specified at 24.1 kHz, because signals above this frequency are aliased into the pass band.

For example, here is the page from the AK4499 datasheet. (I am only using AKM as an example here. Last I checked, every other audio DAC IC manufacturer does the same.) At fs = 44.1 kHz, passband (PB) is 0.4535 × fs = 19.99935 kHz, and stop band (SB) = 0.546 × fs = 24.0786 kHz.

 

[DonH56]

DAC fundamentals: https://www.audiosciencereview.com/...ital-audio-converters-dacs-fundamentals.1927/

See link in my sig for other articles where this is discussed.

 

[Blumlein 88]

Who is ? What is? Mine posting or MQA?

Bob Stuart is spreading crap ideas spouting fud regarding "ringing filters". The Gibbs is due to bandwidth limiting. It is what you should see in a properly band limited system.

Watch this video for some real examples of how things work and stop worrying about things Bob Stuart is talking about.

 

[mhardy6647]

as opposed, I suppose, to Gibbs' phenomenom...

at any rate, when I think of Gibbs, I usually think of this guy...

 

[Mulder]

Bob Stuart is spreading crap ideas spouting fud regarding "ringing filters". The Gibbs is due to bandwidth limiting. It is what you should see in a properly band limited system.

Watch this video for some real examples of how things work and stop worrying about things Bob Stuart is talking about.

#Blumlein. I am not worried. I am just trying to understand Gibb’s phenomenom. I am not trying to understand Stuart.
The origin of my question was admittedly MQA, but more precisely it was about a discussion in another forum where I am also a member. Someone there said that Stuart confuses things when he talks about ringing, and said that it is in fact about Gibb's phenomenon. This in turn is handled by providing some headroom, he said. He said that most DACs do not handle this, but that it is doubtful whether one can hear the effect of Gibb's phenomenon regardless. This was the reason why I started thinking about this.

 

[Mulder]

But then When trying to find an explanation I found this text.
”The Gibbs phenomenon ripples can be reduced by windowing finite impulse response filters. For example, one can use a Blackman window, Hann window, Hamming window, Tukey window, Gaussian window, Kaiser window, and many others. Windowing, although reducing the ripples and usually improving the stop band attenuation of the filter, comes at the expense of wider transition bands.” So, therefore my initial question about stopband. https://www.recordingblogs.com/wiki/gibbs-phenomenon

 

[pjug]

The target upper frequency range of audio DAC has always been flat to 20 kHz. With 44.1 kHz sampling rate, Nyquist frequency is 22.05 kHz, therefore the aliasing artifacts above 22.05 kHz are "reflected back" at this frequency. The aliased component reflected to 20 kHz is thus 22.05 + (22.05 - 20) = 24.1 kHz. This is the reason stop band is specified at 24.1 kHz, because signals above this frequency are aliased into the pass band.

This does seem to be the basis for the design of this filter, and this would make sense as you describe it for an ADC or for downsampling. Shouldn't design goals be different for a DAC where audibility is not the issue, but rather the effect of ultrasonic images on the equipment fed by the DAC? So then Amir is correct when he praises "correct" filters that are better than just -6dB at 22KHz. I realize in practice it might not matter so much, but still it seems strange that these are the filters that seem to have become the standard sharp DAC filter.

 

[Mulder]

So then Amir is correct when he praises "correct" filters that are better than just -6dB at 22KHz. I realize in practice it might not matter so much, but still it seems strange that these are the filters that seem to have become the standard sharp DAC filter.

This is a position I have encountered in another context, ie the claim that today's DAC chips are defective as they do not fully attenuate at 22.05. It has been argued that the underlying mathematics is not correct.