Different Kinds Of Noise......
Oversampling can incorporate 'noise shaping' which means altering how quantisation errors are handled, and consequent sonic artifacts.
Pioneer, Sony and Panasonic DA convertors sound dynamically different due to differing oversampling algorithms.
Non OS dacs measure more distorted on sinewave testing but some regard this distortion as more sonically acceptable than OS dacs.
Panasonic 'MASH' oversampling drives me out of the room.
Eric.
Oversampling can incorporate 'noise shaping' which means altering how quantisation errors are handled, and consequent sonic artifacts.
Pioneer, Sony and Panasonic DA convertors sound dynamically different due to differing oversampling algorithms.
Non OS dacs measure more distorted on sinewave testing but some regard this distortion as more sonically acceptable than OS dacs.
Panasonic 'MASH' oversampling drives me out of the room.
Eric.
Re: Re: Re: Please explain OS vs. NOS to me...
There is definitely bigger interest here with simple chips for a simple reason that they are just easier to implement, only few parts are needed and even PCB is not necessary and sounds very good.
One member, jwb if I'm correct, recently came up with a new design for his oversampled DAC. He claims that this will be definitely his last word on the subject, but considering that this is another and seemingly supposedly improved design, one might conclude that he wasn't completely happy with a previous one.
Lyra said:
There is definitely bigger interest here with simple chips for a simple reason that they are just easier to implement, only few parts are needed and even PCB is not necessary and sounds very good.
One member, jwb if I'm correct, recently came up with a new design for his oversampled DAC. He claims that this will be definitely his last word on the subject, but considering that this is another and seemingly supposedly improved design, one might conclude that he wasn't completely happy with a previous one.
Konnichiwa,
Here a likely non-complete list of DAC's (and CDP's) using No Oversampling:
http://www.morgan-audio.co.uk/deva.htm
http://www.audionote.co.uk/dacs/dac_intro.html
AN UK also has CD-Players with NOS.
http://www.audionote.co.jp/digital/index.htm
http://www.ilungo.com/page/ilungo_705II.html
http://www.zandenaudio.com/products/english/5000.html
http://www.wavelengthaudio.com/preamps.html
I am sure that since the NOS principle has been re-popularised (all early CD-Players excluding Philips Chipset derived ones) where no obversampling!!!) many other smaller companies have started making No Oversampling DAC's.
Sayonara
Bricolo said:
Here a likely non-complete list of DAC's (and CDP's) using No Oversampling:
http://www.morgan-audio.co.uk/deva.htm
http://www.audionote.co.uk/dacs/dac_intro.html
AN UK also has CD-Players with NOS.
http://www.audionote.co.jp/digital/index.htm
http://www.ilungo.com/page/ilungo_705II.html
http://www.zandenaudio.com/products/english/5000.html
http://www.wavelengthaudio.com/preamps.html
I am sure that since the NOS principle has been re-popularised (all early CD-Players excluding Philips Chipset derived ones) where no obversampling!!!) many other smaller companies have started making No Oversampling DAC's.
Sayonara
Konnichiwa,
I forgot thios one....
http://www.msbtech.com/platinum_line/products/DacII/
Here Oversampling and Upsampling are defeatable, making the DAC in effect "NOS" if the user so decides. Review in same issue of Stereophool as 47 Labs and is available from their website, liks below....
Sayonara
http://stereophile.com/digitalsourcereviews/800/
http://stereophile.com/digitalsourcereviews/799/
I forgot thios one....
http://www.msbtech.com/platinum_line/products/DacII/
Here Oversampling and Upsampling are defeatable, making the DAC in effect "NOS" if the user so decides. Review in same issue of Stereophool as 47 Labs and is available from their website, liks below....
Sayonara
http://stereophile.com/digitalsourcereviews/800/
http://stereophile.com/digitalsourcereviews/799/
Konnichiwa,
Forgot two more commercial offerings (how rude of me, as I know both purveyors)....
http://ack.dhs.org/
http://www.scott-nixon.com/dac.htm
This brings the manufacturer count up to at least ten. I also know a Manufacturer who is likely to makert some time soon a 24-Bit, non oversampling, discrete differential DAC with switchable oversampling and transformer coupled output, but the stuff from small acorns grows slow....
Sayonara
Forgot two more commercial offerings (how rude of me, as I know both purveyors)....
http://ack.dhs.org/
http://www.scott-nixon.com/dac.htm
This brings the manufacturer count up to at least ten. I also know a Manufacturer who is likely to makert some time soon a 24-Bit, non oversampling, discrete differential DAC with switchable oversampling and transformer coupled output, but the stuff from small acorns grows slow....
Sayonara
Seems like your original question was never answered.......
Oversampling is a mathematical process to raise the effective sample rate. This is generally done to make the analog low pass anti-aliasing filter simpler and of lower order. The fixed quantization noise is also spread over a wider bandwidth, so this effectively lowers the noise floor.
First: Oversampling is mathematically perfect within the limits of the resolution of the math. If you do your oversampling with enough bits of mathematical resolution, then there will be essentially no error. This is not debatable (though many hear may).
Second: Oversampling and noise shaping are different. That being said, many semiconductor DACs incorporate oversampling (sigma\delta), and as part of that process, they do noise shaping.
So we come to the problem of implementation. There are some potential implementation problems and perceived implementation problems.
- Digital filtering: Most oversampled implementations use FIR filters as these provide linear phase and are relatively simple to implement in the digital domain. They don't require as high of mathematical resolution as IIR filters. However, there is no reason why you could not implement an IIR filter, i.e. butterworth, bessel, or whatever to give a similar to analog filter response.
- DACs. In order to truly get the advantage of oversampling, you need to implement the DAC with higher resolution than you started, at least to limit the mathematical error. However, almost all high resolution DACS these days are sigma-delta with their own oversampling\noise shaping. TI\BB still make some high resolution ones that are not.
- OS Chips (i.e. AD1896, etc). Are based on FIR type filters. You may or may not like this.
Some of the complaints you will hear about OS DACs [FIR based] are things such as "pre-ringing", etc. Pre-ringing is not really so much ringing as the accurate playback of a bandwidth limited square wave. Is it right or wrong? Good question. Some of that may be determined by how it was recorded. Many recording systems use sigma-delta A/D convertors by the way.
What sounds better? I have heard very good implemetations of both? Age may actually play a part. As you get older, you lose your high frequency hearing, so some NON-OS artifacts may get naturally filtered out.
Hope this helps, but maybe it just confused you more.
Oversampling is a mathematical process to raise the effective sample rate. This is generally done to make the analog low pass anti-aliasing filter simpler and of lower order. The fixed quantization noise is also spread over a wider bandwidth, so this effectively lowers the noise floor.
First: Oversampling is mathematically perfect within the limits of the resolution of the math. If you do your oversampling with enough bits of mathematical resolution, then there will be essentially no error. This is not debatable (though many hear may).
Second: Oversampling and noise shaping are different. That being said, many semiconductor DACs incorporate oversampling (sigma\delta), and as part of that process, they do noise shaping.
So we come to the problem of implementation. There are some potential implementation problems and perceived implementation problems.
- Digital filtering: Most oversampled implementations use FIR filters as these provide linear phase and are relatively simple to implement in the digital domain. They don't require as high of mathematical resolution as IIR filters. However, there is no reason why you could not implement an IIR filter, i.e. butterworth, bessel, or whatever to give a similar to analog filter response.
- DACs. In order to truly get the advantage of oversampling, you need to implement the DAC with higher resolution than you started, at least to limit the mathematical error. However, almost all high resolution DACS these days are sigma-delta with their own oversampling\noise shaping. TI\BB still make some high resolution ones that are not.
- OS Chips (i.e. AD1896, etc). Are based on FIR type filters. You may or may not like this.
Some of the complaints you will hear about OS DACs [FIR based] are things such as "pre-ringing", etc. Pre-ringing is not really so much ringing as the accurate playback of a bandwidth limited square wave. Is it right or wrong? Good question. Some of that may be determined by how it was recorded. Many recording systems use sigma-delta A/D convertors by the way.
What sounds better? I have heard very good implemetations of both? Age may actually play a part. As you get older, you lose your high frequency hearing, so some NON-OS artifacts may get naturally filtered out.
Hope this helps, but maybe it just confused you more.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.