There is a point that is not clear to me, since I did not have the opportunity to listen or test to any oversampling DAC without a digital filter.
If, we say, the main drawback of "oversampling" (vs. "NOS") is the use of the digital filter (for ringing and artifacts problems), is it possible to create an "oversampling" DAC without a digital filter, but still including a smooth analog filter?
If someone prefers the NOS sound, is the "oversampled DAC but without digital filter" sound competitive (still natural and musical, like NOS BUT much more detailed, live OS)?
Or, even tricky, if an analog reconstruction output filter design works good and enough with a NOS, would it sound better in a oversampled DAC (but without digital filter)?
If, we say, the main drawback of "oversampling" (vs. "NOS") is the use of the digital filter (for ringing and artifacts problems), is it possible to create an "oversampling" DAC without a digital filter, but still including a smooth analog filter?
If someone prefers the NOS sound, is the "oversampled DAC but without digital filter" sound competitive (still natural and musical, like NOS BUT much more detailed, live OS)?
Or, even tricky, if an analog reconstruction output filter design works good and enough with a NOS, would it sound better in a oversampled DAC (but without digital filter)?
http://lavryengineering.com/pdfs/lavry-sampling-oversampling-imaging-aliasing.pdf
OS means some sort of interpolation (zero stuffing or more intelligent) - in there lies the challenge!?
//
OS means some sort of interpolation (zero stuffing or more intelligent) - in there lies the challenge!?
//
Yes, but my original question was: can I do this filter (in an oversampled DAC) 100% in analog domain and not partially digital and partially analog like the most of oversampled DAC? How this sounds in comparison to NOS (with analog filter)?
Last edited:
If you oversample but use no digital filtering at all, the primary advantage of oversampling (the ability to use a simpler analog filter) is lost. With no digital filter, oversampling must simply repeat the previous sample for the interpolated value(s) (zero order hold) and that's equivalent to no oversampling.
But if I oversample n times without the digital filter, I can use a gentle analog filter cut much higher (with a cut off frequency n times higher than NOS). Or not?
Say you decided to repeat the previous sample 3 times, giving you an update rate of 4X the original. You've still got images which start at 24.1kHz because you've not removed those through use of digital filtering. Meaning you've got to build a very sharp analog filter to eliminate them, just as for NOS.
I don't understand this. If your new sample rate is 4x why would you have such low images? Is this not the same as upsampling with respect to images?
As Abraxalito already said, it is only "oversampling" or "upsampling" if the digital filter process (interpolation process) that removes the previously existing images around the original sampling frequency (and it´s multiples) is included.
Unfortunately (i hope it does not add confusio mentioning this) there is a inconsistency in the nomenclature, as in system theory the block that does increase the sampling rate is also called quite frequently a "upsampler" followed by an so-called "interpolator" block.
But on a device level (DACs, digital players and so on) "upsampling" and "oversampling" always means the complete process of digital low pass filtering and raising the sampling rate.
Unfortunately (i hope it does not add confusio mentioning this) there is a inconsistency in the nomenclature, as in system theory the block that does increase the sampling rate is also called quite frequently a "upsampler" followed by an so-called "interpolator" block.
But on a device level (DACs, digital players and so on) "upsampling" and "oversampling" always means the complete process of digital low pass filtering and raising the sampling rate.
It depends on your definition of 'digital filter'. I built a 2X OS DAC without a filter implemented by means of digital multiplier and digital accumulator - it used analog multipliers (a resistor network) and an analog accumulator (summing currents). But I doubt if such a DAC exists commercially so you're probably right.
Of course difficult to ensure for each and every amplifier in the world, but imo generally chances are low that amplifiers/speaker can handle it.
Let´s assume that in the original content a spectral component around 1 kHz with full scale level exists; if no filtering is given (even not an analog filter) it follows that at the Fsample - 1 kHz and Fsample + 1kHz (means at 43.1kHz and 45.1 kHz, in case of a 44.1 kHz sample frequency) full scale signals exist and at the muliples of the sampling frequency the same happens,i.e. NxFsample - 1 kHz and NxFsample + 1 kHz (N = 2,3,4,5....).
And that is just an example for one spectral component, usually there are a lot more especially so for records suffering from "loudness war" .
@ Ygg-it,
yes, or iow any player that does not filter the spectral images should not be denoted as an "oversampling" player to keep the meaning of the terms.
If one uses for example shorter impulses to lessen the impact of the amplitude level weighting introduced by the zero-order-hold function in a 44.1 kHz system, it is not oversampling and should be denoted by another term.
Last edited:
I very much doubt images will trouble speakers. I do think amps in general distort more due to excess HF. Its not typically heard as distortion though, the subjective effect is either exaggerated leading edges or perhaps reduced dynamics, or both.
The most important is the upsampling in the reccording
Is it correct to assume the ringings & artifacts are already pushed for the biggest amount above the ears fequencies during the analog to digital reccording. And it is where it is mostly important?
Seems all the reccordings are upsampled at 88/98 K Hz or 171/192 Khz to remove at the best phantom images and sub harmonics artifacts digitals filters and mixings add and then just print on CDs a 44.1 Hz "extrapolation" of the reccording. And it doesn't mean the reccording engineers do not use some others artifacts which can be heard too : dynamic compressing, lot of mixing, etc to glue to the limit of Redbook. It may be a better subjective hearing of micro dynamics which is "loved" by NOS enthusiats?
Question is whether our hearing system gives a good enough "natural" low pass filter with all those "unatural" reccording. And materials with upsample above the Redbook stay rare.
Do we really need a steep analog filter to help the hears when not using digital filters? Sorry if mx the concepts, my understanding of it is low.
Is it correct to assume the ringings & artifacts are already pushed for the biggest amount above the ears fequencies during the analog to digital reccording. And it is where it is mostly important?
Seems all the reccordings are upsampled at 88/98 K Hz or 171/192 Khz to remove at the best phantom images and sub harmonics artifacts digitals filters and mixings add and then just print on CDs a 44.1 Hz "extrapolation" of the reccording. And it doesn't mean the reccording engineers do not use some others artifacts which can be heard too : dynamic compressing, lot of mixing, etc to glue to the limit of Redbook. It may be a better subjective hearing of micro dynamics which is "loved" by NOS enthusiats?
Question is whether our hearing system gives a good enough "natural" low pass filter with all those "unatural" reccording. And materials with upsample above the Redbook stay rare.
Do we really need a steep analog filter to help the hears when not using digital filters? Sorry if mx the concepts, my understanding of it is low.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.