Digital Audio Converters (DACs) Fundamentals

[garbulky]

No, you might have brought it up, but many jumped on, including myself. It is a related subject but filter design is pretty vast all on its on. The problem is trying to keep the threads in this part of the forum a bit more focused so they serve as a reference for the future. When they grow to ten or twenty pages of quasi- or unrelated content they lose value as people no longer want to wade through all the "other" stuff even if it is relevant. Some of these articles assume at least some technical background so it is hard to grasp starting from ground zero. Be like me trying to understand a medical article, even an introductory one; too many assumptions and terms that would be beyond me.

I understand your reasoning. No problems here. I posted something in the Ygdrassil section.

 

[DonH56]

"High quality D/A converter without oversampling and digital filter"

My emphasis. There's no way they will not have put in an anti-alias/reconstruction filter. But it's an analogue filter, not digital.

The objection (by some) to digital filters is those used after the upsampling, in order to further ease the job of the analogue anti-alias/reconstruction filter.

A pure upsampler just inserts zeros between the samples (e.g. 4x upsample would be [sample1, 0, 0, 0, sample2, 0, 0, 0, etc]) and, by the magic of sampling theory, the aliases are still moved out to the new Fs. But you can apply a digital filter to the output of the upsampler, to 'smooth' the output of the upsampler.

Actually, there were a number of designs that did not incorporate an explicit image filter, though in practice there is always a little roll-off. They caused big problems for test instruments since the ultrasonics played havoc with the SNR measurements. And caused problems in some amplifiers that did not handle such high frequency signals nicely.

Upsampling is yet another interesting but large topic... Zero-padding is still used in some applications, but others use a variety of interpolation schemes to "fill in" the samples, from simple linear interpolation to complex equation systems (exponential and sinusoidal among others) to better predict how to get from one sample to the next. Some of them are application-specific, natch, since the type of interpolation that works best depends upon the nature of the signal.

 

[captain paranoia]

Actually, there were a number of designs that did not incorporate an explicit image filter

Thanks. I would never have contemplated doing that. But I guess you could rely on the amp's input circuit doing some filtering, and the naturally limited frequency response of the amplifier itself. Apart from the 'audiophile' amps that feel the need to amplify well into the ultrasonics...

from simple linear interpolation to complex equation systems (exponential and sinusoidal among others)

Yup. Or any of the standard filters, implemented in their digital forms.