[Niteshade]

When it comes to audio- are there any good reasons for upsampling?

The reason why I ask is this: You cannot get something from nothing.

Am I looking at this wrong?

I am looking at upsampling as a way of filling in spaces and have always believed a 1:1 sampling rate to be the best approach.

[Niteshade]

I've been doing some research on this and it would seem there really is no benefit to upsampling.

Perhaps the best way to upgrade would be to record audio on to DVD's, since they're sampled at 48Khz.

[LostInPA]

Blair,

I am going to disagree with you on this.   I recently got the new input board upgrade for my MSB Technology Platinum DAC III.   This provides the ability for 8X upsampling, which can be engaged from a front panel switch.   Three other pair of good ears were here two days ago.   We selected among four different ways of getting from the CD transport to the analog inputs of the preamp:

1- using the DAC section in a Tascam CD-01U Pro CD player, 16/44.1;

2- using a digital output from the Tascam in to an MSB Technologt Gold Link Dac, 16/44.1;

3- using a digital output from the Tascam in to the MSB Platinum DAC, 16/44.1;

4- as above, engaging the 8X upsampler, 16/352.8.

We switched among the four choices with a variety of music, and it did not take very long for us to stick with #4, and then spend a couple of hours just enjoying some music.   My system is very revealing:  an autoformer-based passive volume control, and a 50 wpc tube stereo amplifier from you (Nightshade Audio) driving a B-G Radia 50" planar-magnetic from 150 Hz and up.   

I do not understand why some people have a problem with upsampling being a way to fill in the spaces.   Would you chose to put tires on your car that are square?  hexagonal?  octagonal?   A circle is a polygon with an infinite number of sides. 

 

[Niteshade]

I don't believe 44.1Khz is anywhere near square or rough. Could it be that your new D/A hardware is better quality?

[denjo]

Blair

I think the issue is whether one would accept the sound as-is (the purist school) or whether one would accept a little bit of help to "fill in the spaces" (as you put it), in order to get a "better" sound. Its much like the analogy of the equalizer. there are some preamplifiers that don't have any tone controls, others that do. Personally, I would not mind a little help to make the music sound more enjoyable, even if this may be at the expense of "adding" or "subtracting" a little from the original. In a subtle way, swapping cables and the like is a form of "tone" control so if one is acceptable, why not the other? Where do you draw the line and say, this is the point where "intervention" ends! Others just want the music the way it is, nothing more and nothing less. One further analogy if I may: I know of some audiophile friends who would prefer an early Beethoven recording (which sounds thin and crappy) instead of a more modern (and technologically sophisticated) recording of the same Symphony!

Best Regards
Dennis

[satfrat]

I'm in full agreement on this one Blair. I've yet to hear any hands-down improvement when comparing upsampled 44.1k recordings from my hard drive to say 48k or 96k. For me, upsampling robs the music of it's life, especially in the highs. But that's just in my system, maybe higher grade upsampling gear can make improvements? In my system, it's strickly 24/44.1k playback.

Cheers,
Robin

[pardales]

One must also keep in mind the distinction between upsampling and oversampling.

I am going to disagree with you on this.   I recently got the new input board upgrade for my MSB Technology Platinum DAC III.   This provides the ability for 8X upsampling, which can be engaged from a front panel switch.   Three other pair of good ears were here two days ago.   We selected among four different ways of getting from the CD transport to the analog inputs of the preamp:
 

It sounds to me like you are describing oversampling here. I am not qualified to elucidate the difference between these two techniques though I know it has been done many times.

[*Scotty*]

My current system outputs the wav file from my hard drive out a USB port to the SPDIF input of a heavily modified Cambridge 740 C that up-samples to 24/384 via a 32-bit Analog Devices Black Fin DSP. This approach yields a very good result that sounds a lot like my vinyl source. I really don't have a digitis problem with this setup. I think if the upsampling is done asynchronously with a powerful enough processor utilizing Anagram Technologies Adaptive Time Filtering algorithm you can achieve a very satisfactory result without the downsides that have been associated with products that have incorporated upsampling in the past. Here is a link to Anagram Technologies website with an explanation suitable for laymen  http://www.anagramtech.com/technology/atf-adaptive-time-filtering/   A paper presented at the 120th AES convention (May 2006) by Thierry Heeb  is available from the AES Store for $20. Cheapskate that I am,I haven't purchased the paper yet. I am content to listen to the results of the applied technology and save my money for more music.
Scotty

[Niteshade]

Thank you for the input folks!

I always thought of upsampling as something like  blowing up a picture. If you make a small photo into a larger one detail is lost because it has to be. Would this account for added smoothness, especially at the high end?  I can see where this attribute could be easy on the ears.


There's just too many interesting  toys on the market! 

[PhilNYC]

The purpose of upsampling is not to gain resolution.  If it was, you could record music to 8-bit/100hz, then upsample it on the CD player and fit hundreds of hours of high quality music on a CD...

Upsampling is done to help filter out high frequency digital "artifacts" without impacting the audible spectrum.  When a DAC converts a digital signal to analog, slight inaccuracies in this process introduced artifacts at just over the replay frequency; at 44khz sample rate, artifacts are introduced at just above 22khz.  Non-upsampling DACs require an analog filter be placed at just above 22khz to eliminate those artifacts...by being so close to the audible frequency, it is assumed that its possible for some of these filters will impact your listening experience.  By upsampling to a higher frequency, a DAC essentially moves those artifacts to a higher frequency, farther from the audible spectrum, and making them easier to filter out without impacting the audible frequency spectrum.

There are probably a few other reasons for doing it, but this is my understanding of the initial reason for upsampling...

[PhilNYC]

I'm in full agreement on this one Blair. I've yet to hear any hands-down improvement when comparing upsampled 44.1k recordings from my hard drive to say 48k or 96k. For me, upsampling robs the music of it's life, especially in the highs. But that's just in my system, maybe higher grade upsampling gear can make improvements? In my system, it's strickly 24/44.1k playback.

Cheers,
Robin

My Dodson DAC upsamples to 96khz, then oversamples 8x to 768khz...and it's still the best DAC I've ever heard...

(disclaimer; I'm a Dodson dealer...)

[Niteshade]

This is interesting material: http://home.mira.net/~gnb/mac-cdis/cd8.html

[*Scotty*]

Blair,Thanks for posting the link to the MAC articles the first 8 parts were very useful info. It's nice to have a link that you can point someone at that explains the basics of the CD medium. What I wish someone could find is a good discussion of the increasing amount of THD that results as the signal in the digital domain moves away from zero dB which is the point of maximum resolution and lowest THD. An example of what I am talking about is contained in several reviews of DACs in Stereophile. SEE image 
 Fig.4 Linn Klimax DS, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data.
It would be nice to know at what point the distortion reaches 0.3THD which is point at which distortion was considered objectionable in an analogue tape recording system.
To my mind this is a critical flaw in the 16bit system. At what point is the loss of bits objectionable as you move away from 0dB which is the only time where you have 16bits.
To my mind the 16bit system only has perhaps 60dB of usable dynamic range before the distortion levels become objectionable. This problem is the primary argument for a 24bit or higher system
A 24bit waveform at negative 90dB for contrast.
Fig.5 Linn Klimax DS, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data.
Scotty

[JoshK]

The purpose of upsampling is not to gain resolution.  If it was, you could record music to 8-bit/100hz, then upsample it on the CD player and fit hundreds of hours of high quality music on a CD...

Upsampling is done to help filter out high frequency digital "artifacts" without impacting the audible spectrum.  When a DAC converts a digital signal to analog, slight inaccuracies in this process introduced artifacts at just over the replay frequency; at 44khz sample rate, artifacts are introduced at just above 22khz.  Non-upsampling DACs require an analog filter be placed at just above 22khz to eliminate those artifacts...by being so close to the audible frequency, it is assumed that its possible for some of these filters will impact your listening experience.  By upsampling to a higher frequency, a DAC essentially moves those artifacts to a higher frequency, farther from the audible spectrum, and making them easier to filter out without impacting the audible frequency spectrum.

There are probably a few other reasons for doing it, but this is my understanding of the initial reason for upsampling...

I get mixed up which is upsampling and which is oversampling.  I thought the 44khz part was oversampling and the bit was upsampling, but I could have them backwards.  The khz is the part affecting the filter.

Well said!  I am not an expert on the subject but the above is consistent with my understanding.  The problem comes down to the supersonic filter (above 22khz).  With the just above 22khz problem, if you use a shallow slope filter it will affect the upper audible range, if you use a steep filter it introduces group delay with can bleed down into the audible range.  So oversampling is one way of extending the artifacts up high enough to use a steep filter without the bleed down. 

AFAIK, oversampling is all about the filter process.  Non oversampling DACs sometimes don't have a filter since it is hard to make them inaudible (given the above problem).  This can result in supersonic energy that some amps do not like and can affect the stability of the amp.   

[*Scotty*]

Josh, NOS Dacs have another problem when they are operated without a filter the ultrasonic components downward intermodulate with the signal with objectionable consequences all the way down into the midrange. A cure for the phase shift due to the filter is to push the noise high enough in frequency through over/upsampling so that a simple Bessel filter can be used which does not have the inband phase shift problems. Here is a link to an explanation of oversampling and upsampling.  http://www.audioholics.com/education/audio-formats-technology/upsampling-vs-oversampling-for-digital-audio
Scotty

[PhilNYC]

I get mixed up which is upsampling and which is oversampling.  I thought the 44khz part was oversampling and the bit was upsampling, but I could have them backwards.  The khz is the part affecting the filter.

My understanding is that "upsampling" is the process of completely re-calculating data points to raise the sample rate to a higher number (eg. convert 44.1khz to 96khz) using Fourier transformations or other mathematical algorithms, whereas "oversampling" converts a set of data points at one rate to a whole-number/even-integer multiple (eg. 2x, 4x, 8x, etc), leaving the original data intact, but adding points of data between the original data points.  The number of bits is increased in both cases (eg. from 16-bit to 24-bit) in order to capture the increased precision (significant digits) needed for the new datapoints.

"Oversampling" came first, because digital clocks were easy to tune to even-integer multiples.  "Upsampling" became an option when DVD audio (not the DVD-A format necessarily, but the audio tracks on movie DVDs) decided to use 96khz as its sample-rate standard...and someone had the bright idea of using a DAC chip designed for DVD playback in a CD player...

[satfrat]

Another explanation of upsampling/oversampling from Steve Nugent in a recent thread located here;

upsampling != oversampling.

It might very well be true that it's also non-oversampling, but then I'm confused about the role of the superclock.

Oversampling is synchronous and does add more sample points.  The purpose is to enable digital filtering.  I think the algorithm is fairly simple. In the Overdrive, this is selectable, so you can push the digital filtering up in frequency until it has very little effect.  Looking at the waveform on the scope reveals that the classic "stairsteps" of the D/A waveform are there, just like they would be on a NOS DAC.

Upsampling is asynchronous usually and also adds more sample points.  The algorithms can be complicated and take a lot of compute power.  Usually the sample-rate is changed to a higher standard such as 44.1kHz to 96kHz.

The Superclock is used both in the USB interfaces and the Pace-Car to reclock the data without changing the frequency, in order to reduce jitter.  Clocks are simply used to move the data from point A to point B.  Digital data must be clocked in order for it to move from one point to another.  The clock that is used to move the data is changed from the computer clock to the Pace-Car Superclock or Off-Ramp Superclock.

Steve N.

Cheers,
Robin

[*Scotty*]

Steve Nugents explanation is very concise, the computing power necessary to asynchronously upsample used to be expensive which limited it to products which were also expensive. As computing power has become less costly upsampling will probably find its way into an ever widening number of products. One of the early examples of upsampling was the NagraDac back in 2004 priced at $12,000, it also upsampled to 192kHz and used a SHARC processor.The Dacmagic at $400 is a good example of trickle-down and also has upsampling to192kHz.
Scotty