Upsampling. Truth vs Marketing


Has anyone done a blind AB test of the up sampling capabilities of a player? If so what was the result?

The reason why I ask because all the players and converters that do support up sampling are going to 192 from 44.1. And that is just plane wrong.

This would add huge amount of interpolation errors to the conversion. And should sound like crap, compared.
I understand why MFG don't go the logical 176.4khz, because once again they would have to write more software.

All and all I would like to hear from users who think their player sounds better playing Redbook (44.1) up sampled to 192. I have never come across a sample rate converter chip that does this well sonically and if one exist, then it is truly a silver bullet, then again....44.1 should only be up sample to 88.2 or 176.4 unless you can first go to many GHz and then down sample it 192, even then you will have interpolation errors.
izsakmixer
Non-integer sampling frequency conversion is no big deal. The worst case error is a fraction of the error which would have existed if the interpolated data had not been calculated and output at the higher rate.
This is the one and only experience I have had in listening to a player where you could choose to upsample or not, so take it with a grain of salt...

After we went to a Friday Night Cruise Night in Sommerville, NJ in Bill LeGall's Packard this past autumn, we sat down and listened to his Music Hall Maverick. I asked him if he preferred the upsampling or not, and he told me he didn't know, as he never tried it. So, he asked if we could do some A/B testing, and keep our opinions to ourselves until it was over. Then, we'd see where we both stood. We listened to a few cuts, with and without the upsampling. The other components in the system were his incredibly modded Infinity IRS V speakers (with internal 1500 watt amps for the bass modules), Mesa Baron running 6L6 tubes 2/3 triode - 1/3 pentode power, Audio Research preamp, and his own cabling.

Without the upsampling, the music was clearly more dynamic, open, upfront, and to both of our ears, much more honest. Upsampling REALLLLLLLLLLLLLLLLLY smoothed out the sound, and gave an extreme LP kind of feel to the music. I don't want to say that detail dropped down, but because of such a loss of jump factor in the music, it gave it that kind of vibe. The differences were night and day. We both preferred no upsampling by a mile, but I could clearly see where a lot of audiophiles would feel the opposite, and in many a system, it would probably be a better way to go.

I will say that after this experiment, the Zanden ads that tout its great sound, partly they feel due to the fact that it does not upsample made a lot of sense to me.
Treja, I would have loved to have been there. I have only compared the non-upsampling Audio Note to other upsampling rigs. My preference has always laid with the Audio Note. There is an unfettered feel about the honest read.
Upsampling may or may not sound better depends on overall design and implementation.

Companies use upsampling because it can allow digital filtering designs that do least amount of harm to music signal.

Don't get hung up on the numbers game, listen to the end result - the music.
I own an Ah! 4000 CD Player with the optional upsampler board. It's a neat comparison because you can remove the board and replace with the stock chips. Anyway....

The improvement is DRAMATIC...It's smoother, has better resolution and basically does everything better with the Upsambler installed.
If an when you got an upsamiling player do you know what the actual upsampled rate is? Is it 192khz or 176, none of you listed it. I know of one product that has upsampling and it goes to 192 and it does sound different, but not better. Just looking at the difference in the recorded waveforms (44.1 and 192) clearly shows interpolation errors. If anyone has means to record WAVE files of the output, I would be happy to analyze. All and all I would have to beleive that if your player sounds better upsampled they must be doing 172. Or could it be that their clock is crap and when you use higher sample rate, jitter is less of an issue?..hum..

A<.
An above response having to do with individual preferances rings true. If you do an inventory of some of the most impressive virtual systems on Audiogon you will generally find more upsamplers than those without. I do think that one test of upsampling has to do with amps and speakers. If they work well with upsampling then this is the way to go.

when I first got my Audio Aero Cd player I tried to run it through my receiver and all of the bass and tone controls sounded terrible. I had to set everything at neutral and it sounded better.
I agree with Megasam, specs dont tell you anything about the sound. I cant really comment on the upsampling question, but my experience with amplifiers tells me you dont always get what you would be led to expect. That goes for power ratings, type of technology, price or age. My Totem Ones taught me that numbers can decieve. When a 25 watt ss amp drives a speaker better than a 100 watt ss amp of fairly good quality, you know the only way to tell is by listening. Buy the stuff that sounds good.
Check out what Audio Research has to say about it with their CD3 III machine.
Get an 8.5 x 11 piece of paper and a pencil.

Now draw a straight line horizontally across the page about 1/4 of the way down. Do the same think half-way down the page and then another horizontal line at 3/4 of the way down the page. The top line will be called "A" and the bottom line will be called "B". The center line is simply a dividing line between A & B.

The A & B lines represent zero in these two different lay-outs. The space above the zero line in each case is positive and the space below each zero line is negative. In effect, the upper half of the page has a positive section, the zero line for section A and a negative section. Then you have the dividing line in the middle of the page separating A from B. Below that, we have the positive section of B, the zero line and the negative section of B.

Now start at the left side of the page and draw a series of 20 random "dots" above and below the "zero line" in section A. For demonstration purposes, place 10 dots above and 10 dots below the zero line. Put your first dot in the positive section, move down in into the negative section and place your second dot there. When doing this, start moving gradually from the left side of the page over towards the right. Vary the distance of each dot above and below the zero line and spread them across the entire width of the page.

In other words, some dots should be closer to the zero line, others should be spaced further away from it, etc... Some dots will be closer together and some will be spaced further apart. We are looking for strictly random spacing here, both in terms of above and below the zero line and from left to right.

When you've got all 20 dots filled in on the positive and negative sections of part A, duplicate the placement of those dots down in section B. In other words, the dots in section B should be identical to those in section A, both in terms of the heights above and below the zero line and their left to right spacing.

Now that you have that done, we are going to stay in section B. What you need to do now is to add another 20 dots to what you have in section B. Same rules apply i.e. random vertical and horizontal spacing with ten more above zero in the positive range and ten more below zero in the negative range.

When that is done, set your pencil down and look at the A dots and then look at the B dots. Each dot represents a sampling point of a bunch of musical notes. In case you haven't figured it out, A represents "standard" CD playback and B represents "upsampling".

Pick up the pencil and start in section A. We are going to play "connect the dots" here. Working from left to right, draw a line from zero up to the first positive dot. From that first positive dot, draw a line down to the first negative dot. From that first negative dot, draw a line to the second positive dot. From there, down to the second negative dot. Alternate between positive and negative going from left to right until you've got all twenty dots connected with one jagged line.

Now go down and do section B the same way, but this time, start at zero and draw a single line up to TWO positive dots, then down to the two negative dots, back up to two positive dots and then down to two negative dots, etc... Do this just like you did with section A, working left to right. When you're done, you'll have all fourty dots connected with one jagged line.

Now compare the two "waveforms" that you've just drawn. Section A is a standard CD waveform and section B is an upsampled waveform. As you can see, even though you started off with the same basic amount of data in both sections A & B ( 20 sampling points ), adding the additional data via upsampling ( twice the sampling rate hence twice the data ) drastically alters the waveform and response. Not only are some of the changes in polarity from positive to negative not as abrupt, there's a lot more contouring taking place to fill in the gaps between sample points in section B.

While upsampling hasn't "created more notes" to fill in those gaps, it simply allows us to follow what was already there more accurately. Not only are some of the amplitudes different, but the contours ( rise and fall times ) of each note can be seen more accurately.

By following those contours more precisely, we limit overshoot and ringing. The reduction of overshoot results in a more natural sound albeit less "artificially hyped" in the dynamics department. At the same time, we can also hear how notes decay a little longer, rather than just hitting and fading to the next note so rapidly. This adds a more "lush" albeit "slower" presentation to what we are hearing, much like the natural decay of a plucked string on an upgright bass or cello.

These "extra" samples as seen in the upsampling section of B should not be confused with "oversampling". Oversampling simply looks at the same points on section A and confirms that each reading is in line with the last reading taken. There's no more data recovered, it just keeps verifying that the limited data that it has retrieved in those samples is correct and consistent. It does this repeated sampling of the same points 4x, 8x, 16x, etc... If the samples don't jive, error correction can kick in as needed. The more error correction that kicks in, the more "self interpretation" of the data that the machine itself has to do.

Other than that, the majority of what makes the most audible differences in players is the type and quantity of filtering used and how it is implimented into the circuit. Audio Note eschews much of the filtering and gets rid of the oversampling, which reduces a LOT of the in-band noise and distortion that lesser designs introduce. At the same time, it can introduce out of band noise and distortion into the equation, which isn't good either.

Obviously, the key is to find a way to increase the sampling rates to recover more of the data AND do so in conjunction with well thought out filtering and error correction circuitry. In doing so, we should end up with more of the benefits with less of the drawbacks. Sean
>

PS.... This is kinda - sorta the "quick and dirty" explanation of upsampling. It may not be perfect on all counts in terms of technical accuracy, but you should get the basic idea of what's going on and why it has the potential to be a superior performer.
Sean,
basically your "kinda - sorta the "quick and dirty" explanation" is pretty good for the layman.
I have objections on some of the text you posted:
* Audio Note eschews much of the filtering and gets rid of the oversampling, which reduces a LOT of the in-band noise and distortion
How does getting rid of oversampling & eschewing much of the filtering reduce IN-BAND noise & distortion???
AFAIK, anything in-band cannot be touched. It's sacred as it's THE signal we are looking for. If noise exists in-band or if distortion exists in-band, you basically have to live w/ it OR design better electronics. What you wrote will not do the trick.

* At the same time, it can introduce out of band noise and distortion into the equation
what are you referring to here? i.e. when you write "it can introduce....", what is "it"??

* Obviously, the key is to find a way to increase the sampling rates to recover more of the data
Increasing the sampling rate does NOT recover more data. It, however, allows the discrete-time system to follow the original analog data more truthfully. This is evident from your section A, section B example.

On a historical note, Philips is the co. that is to be credited or discredited with the concept of upsampling. The original idea at Philips Reasearch Labs was to somehow get that analog filter order lower & that transition band less steep. In the original redbook spec, the transition band is 20KHz-22.05KHz. Upsampling was the answer from an engineering perspective & from a cost prespective. They really didn't care about the sonic effects back then.
FWIW. IMHO.

Bombay: Your own description answers the problems that you questioned i.e. "On a historical note, Philips is the co. that is to be credited or discredited with the concept of upsampling. The original idea at Philips Reasearch Labs was to somehow get that analog filter order lower & that transition band less steep. In the original redbook spec, the transition band is 20KHz-22.05KHz. Upsampling was the answer from an engineering perspective & from a cost prespective. They really didn't care about the sonic effects back then."

By playing games with the actual cut-off frequency and Q of the filtering OR by removing the majority of filtering, you reduce the amount of roll-off, phase shift and distortion in the treble region. As far as oversampling and error correction goes, that simply equates to more tampering that the machine itself is doing with the signal and / or noise that it is generating within the power supply and support circuitry.

In effect, error correction is "somewhat" like negative feedback. As such, Audio Note feels that small errors aren't as much of a negative as the problems that result from trying to correct them. Between the lack of oversampling and their approach to filtering, many people seem to agree with the sonic results that they've achieved. As a side note, Moncrieff covered error correction in IAR many years ago. Sean
>

Mathematically, there are no differences between upsampling and oversampling. Upsampling is basically a marketing term and it is NOT coincidental that it was conjured up during the redbook lull prior to DVD-A format agreements. Really, what is so special about 96kHz or 192kHz?? Why not 88.2kHz or 176.4kHz? For that matter, why not 352.8kHz or 705.6kHz? The choice of resampling a 44.1kHz signal to 96kHz or 192kHz is entirely about piggy-backing on the new high rez formats for marketing purposes. In fact, there is potential for loss of information by resampling assymetrically rather than by integer multiples.

Please refer to Charles Hansen (Ayre) or Madrigal, or Jeff Kalt (Resolution Audio), or Wadia, or Theta. All have made multiple statements that upsampling is nothing more than a marketing tool. Maybe it's good for high end in this sense...certainly high end redbook CD sales jumped after the "upsampling" boom. Magazine reviewers seemed eager to turn a blind eye since their livelihood depended on a healthy high-end market. Waiting 2-5 years for decent universal players certainly wasn't attractive, nor would reviewing the latest $20k redbook CD player when the general consensus at the time was that even bad high rez would blow away great redbook.
Phillips used 4 times oversampling in their first CD players so that they could achieve 16 bit accuracy from a 14 bit D/A. At that time, 16 bit D/A, as used by Sony, were lousy, but the 14 bit units that Phillips used were good. The really cool part of the story is that Phillips didn't tell Sony what they were up to until it was too late for Sony to respond, and the Phillips players ran circles around the Sony ones.

In Sean's explanation the second set of 20 dots in set B should not be random. Those dots should lie somewhere between the two dots adjacent to them.

Here is my explanation.

Assume there is a smoothly varying analog waveform with values at uniform time spacing, as follows. (Actually there are an infinite number of in-between points).

..0.. 1.. 2.. 3.. 4.. 5.. 6.. 7.. 8.. 9.. etc

If the waveform is sampled at a frequency 1/4 that of the example, (44.1 KHz perhaps) the data will look like the following:

..0.......... 3.......... 6...........9..... THIS IS ALL THERE IS ON THE DISC.

A D/A reading this data, at however high a frequency, will output an analog "staircase" voltage as follows:

..000000000000333333333333666666666666999999999

But suppose we read the digital data four times faster than it is really changing, add the four values up,
and divide by 4.

First point……..(0+0+0+3)/4 = 0.75
Second point…. (0+0+3+3)/4 = 1.5
Third point…… (0+3+3+3)/4 = 2.25
Fourth point….. (3+3+3+3)/4 = 3.0
Fifth point……. (3+3+3+6)/4 = 3.75
Sixth point……. (3+3+6+6)/4 = 4.5
Seventh point…. (3+6+6+6)/4 = 5.25
Eighth point…… (6+6+6+6)/4 = 6
….And so on

Again we have a staircase that only approximates the instantaneous analog voltage gererated by the microphone when the music was recorded and digitized, but the steps of this staircase are much smaller than the staircase obtained when the digital data stream from the disc is only processed at the same rate that it was digitized at. The smaller steps mean that the staircase stays closer to the original analog ramping signal.

Note also that we are now quantized at 0.25, instead of 1, which is the quantization of the data stream obtained from the disc. A factor of 4. That’s like 2 bits of additional resolution. That’s how Phillips got 16 bit performance from a 14 bit D/A.
The term "Error Correction" applies to a scheme where redundant data is combined with the information in such a way that a decoding algorithm can recover the original information WITHOUT ANY LOSS, provided that the number of transmission errors, and their distribution in time, does not exceed what the encoding algorithm is designed to deal with. This is not a "bandaid" for poor transmission. It is a way to make it possible to run the hardware at much higher bandwidth because errors can be alowed to occur.

"Interpolation" is not "Error Correction". Interpolation is what you can do if the errors do exceed what your algorithm is designed to deal with. Depending on what the signal is doing at the time that transmission glitches occur interpolation may or may not result in significant error in recovery of the information.
Thanks for the feedback Sean. Putting your orig. & 2nd post together, I see what you were trying to say.
El said: "In Sean's explanation the second set of 20 dots in set B should not be random. Those dots should lie somewhere between the two dots adjacent to them".

By placing the "extra" dots ( sampling points ) "mid-point" between the previously adjoined dots, the end result would look MUCH smoother and far more predictable. While this "could" be the case if playing back sine waves of varying amplitude and duration, music is anything but "sinusoidal" by nature. There are very rapid peaks and dips that take place, sometimes completely changing the direction that the signal was previously headed just a split second previous. These peaks and dips can can switch randomly back and forth across the "zero line" or they can remain above or below the "zero line" for extended periods of time. On top of that, these waveforms may not be symmetrical at all i.e. much bigger peaks on the positive side than there are dips on the negative side or vice-versa. It is for this reason that "industry standard test tones" aren't quite as revealing as we would like as far as revealing how a component performs during normal use reproducing musical waveforms. This is why several different types of tests have to be used in order to obtain any type of meaningful relationship between test bench performance and real world performance.

If music was more like a sine wave i.e. with predictable amplitudes, polarities and durations, error correction algorithms could be much simpler and far more accurate. However, musical notes are anything but predictable in terms of amplitudes, polarities, durations or patterns. As such, the potential to read an error from anything but a perfect disc is not only high, but the potential for further errors to take place when data is lost and the machine is trying to "fill in the blanks" becomes even higher.

Somewhere in one of the old IAR's ( International Audio Review ), Moncrieff covered quite a bit on the flaws of how "Redbook" cd was designed and how their "error correction" and / or "interpolation" techniques were far from all-encompassing. Then again, this was all newer technology at the time, so they were kind of winging it as they went along. As such, the potential for a newer, much better digitally based format is definitely there, especially if we learn from past mistakes and take advantage of the more recent technology that we have.

Germanboxer: As far as certain manufacturers supporting / slagging specific design attributes, did anyone ever expect a manufacturer to support a design / type of product that they themselves didn't already take advantage of? Would you expect a company that didn't use upsampling to say that upsampling was superior or a company that did use upsampling to say that the technology that they were using was a poor choice?

Bombay: Glad that you were able to see where i was coming from after further explanation. Hopefully, others could follow along here too.

As a side note, read the description of this DAC as listed on Agon. You'll see that the designer not only played with various types of filtering, but gave the end user the option to accomodate their personal preferences / system compatibility at the flip of a switch. Bare in mind that this unit was out long before Philips came out with their SACD 1000, which also gave users the options of various filter shaping and cut-off frequencies, etc... Sean
>
Eldartford's sentence: "In Sean's explanation the second set of 20 dots in set B should not be random. Those dots should lie somewhere between the two dots adjacent to them".

is exactly correct. One possible location of "somewhere between" could be legitimately the midpoint. There is no problem with that at all. If the waveform looks smooth then what's the issue with that??? How, in the world, do you know that the waveform at this point in the CD is not supposed to be smooth?? There could be a consistently low volume passage or a consistently loud volume passage of 1 particular instrument that creates a smooth area. Entirely possible.

Anyway, the thing to remember in your 2nd example is that when you placed that "random" set of points, you were looking at the output of the digital estimation filter. The output of digital estimation filter is very deterministic & it is designer created. The o/p simply cannot be random - no way!! It lies "somewhere between" the actual sampled data points off the CD along a line determined by the algorithm of the digital estimation filter. This is that (digital) filter that creates all those signature sounds (like Wadia's house sound, Sim Audio's, dCS's, etc, etc) that many love & equally many hate.

In Eldartford's example, I think, that he used a smooth waveform only to illustrate the point. This is the way that it is usually introduced in DSP 101 classes. His particular example is pertains to oversampling. When he shows the repeating of numbers, he has considered a 12X oversampling & when he does the div-by-4, he is considering 4X oversampling. The div-by-4 most probably represents the digital FIR that follows any over (or up) sampling operation.
My only question here is why did the example consider an oversampling of 12X then later decimate to 4X?? Should have just started of with a 4X DAC. Anyway.....

You mentioned "error correction" for the 2nd time. Error correction in redbook CD playback has nothing to do w/ upsampling or oversampling. Error correction is NOT designed to correct the music written on the CD. It is designed to compensate for high-speed read & transmission of the bits where read errors will occur (owing to the high speed read operation). I think Eldartford's succinct explanation is exactly what error correction is all about. Any other idea of it is a mistaken impression.

I have read the recent upsampling verbose text by Moncrieff on IAR. IMHO, I have not read more bull**** anywhere that filled up so many pages. Very little of what he has written is correct. AFAIK, Moncrieff is very lost when it comes to up & oversampling. If you are taking your lessons from him, then I can see why you are mistaken too. Get hold of a DSP text (like Oppenheim & Schaeffer or Rabiner & Gold) & read that. You'll get the correct explanation of upsampling & oversampling.

Sean...The sampling (your first set of dots) is at 44.1KHz. The highest audio information that exists at this sampling rate is around 20KHz, and at this frequency the music signal amplitude is very small. Therefore, unless the signal is momentarily a constant (two adjacent points the same) the in-between points will lie between adjacent points. Of course this is all overlaid with random noise that will blur the quantization staircase.

The CD recording protocol has been cited as an everyday example of the application of CRC error correcting technology, and I have seen descriptions of the CD protocol as having interpolation as a "fall back" procedure when the CRC error correction fails. Of course the second "fall back" is to abort playing the disc, and this ought to be the only time that the process is easily heard.

To tell the truth I have never actually read this infamous "Red Book" which defines the CD spec, and so am relying on what others have reported. How would I get a copy?
Bombaywalla...Your check is in the mail :-)
Sean....Homework is to read.. http://en.wikipedia.org/wiki/Reed-Solomon_error_correction
Test on Monday :-)
Sean, in fact, Jeff Kalt of Resolution Audio was marketing and using "upsampling" in his players at the time, so yes I do think some manufacturers will be honest when asked directly.

But more to the point, what do you think is so magical about 96kHz or 192kHz? Why not 88.2 or 176.4 or 352.8? I think the obvious answer is that the high rez format in DVD-A is either 96kHz or 192kHz...marketing anyone??

If you could, would you please contrast your upsampling dot graph with the equivalent oversampling dot graph? Remember that to get to 96kHz from 44.1kHz in your example you have to increase the number of dots from 20 to 43.5 dots. What you described is essentially a 2x oversampling routine with linear interpolation. The graph cannot get any smoother than the original unless you use something other than linear interpolation. Yyou are just connecting a series of dots in a line between samples, otherwise.

The main reason for adding points between the original samples is ultimately allow a more gentle analog filter. The original (really bad souding) CD players used no oversampling and analog brick wall filters to avoid the problems associated with the Nyquist limit for 44.1kHz sampling (22.05kHz) and the spurious images that get reflected back in band. These sounded horrible and led to 2X, 4X, 8X etc oversampling moving these images well beyond the audio band and allowing more gentle (better sounding) analog filters.

To paraphrase Charles Hansen, adding another digital filter (upsampling) to the chain will affect the sound; however it is certainly possible to design a single digital filter with exactly the same composite characteristics as the two cascaded filters, usually justs costs a little more money.

Anyway, not trying to be a pain in the a$$, I just think the marketing component of the choice for 96kHz or 192kHz needs to be pointed out.
The connect-the-dots metaphor is really unfortunate, because a lot of audiophiles buy into the idea that that's what a DAC does. But reconstructing an analog wave is nothing like connecting dots. More dots DOES make it easier to DRAW a wave. But as long as you have enough samples for the bandwidth, a DAC can reconstruct that wave without more information.

(BTW, the example given above didn't have enough information to do so, because it called for only two samples per cycle. You need fractionally more than two to reconstruct the wave properly.)

Imagine that, instead of a wave, you were trying to trace a straight line. The more dots you had, the easier it would be to do this freehand. But a graphing calculator would only need two points.
Pabelson...According to Nyquist, just two (error free) samples per cycle will perfectly recover a sine wave. But, in this error-prone nonsinusoidal world, where I played with digital data stream representations of analog waveforms (non audio), experience taught me that four samples per cycle was worth the trouble. That's why the 96 KHz PCM of DVDA (or 192KHz for stereo) solves the bigest problem with redbook CDs. 24 Bits is nice too.
You guys are all correct. Where i "fell down" on this one was that i was thinking in the analogue realm rather than in the digital realm. When i was thinking of how to explain "upsampling", i was trying to demonstrate exactly how "non symmetrical" a musical waveform really is. This is why i suggested that the second set of 20 dots / samples be placed randomly rather than in a neat and orderly fashion between the other dots / sample spaces.

What i forgot to take into account was that we weren't dealing with analogue here at all. We are dealing with analogue that has been hacked to bits ( literally ), completely butchered as it was converted into another format and is now trying to be re-assembled as best possible back to what it was originally. Kind of like taking a fish, throwing it into a blender and hoping to re-build the fish once it comes out of the blender. Good luck.

If you doubt this, try looking at some of the waveforms that Stereophile tries to reproduce on various digital devices. Given that these are symmetrical test tones, you can only imagine how poorly some of designs / devices would do with more dynamically complex musical signals fed into them.

As such, there is little resemblance to what the original analogue waveform looks like after digital processing due to a LOT of various factors, some of which have been more than amply pointed out above. I apologize for the mistake and would like to say "Thank You" to those that corrected my mistakes.

Having said that, i'm glad that at least part of what i was trying to convey was understood and not completely lost. To be specific, i'm talking about the various types of filtering and cut-off frequencies used, why this area of operation affects what we hear inside the audible bandwidth, etc...

As to Germanboxer's comments, i agree that the majority of upsampling is based on parts that are already commercially available products. This not only makes things easier to design, it is also cheaper to produce. Otherwise, manufacturers would have to build "one off" devices for each product manufactured, which would make every upsampling DAC a custom built piece. While this would probably result in better quality as everything would be designed from the ground up rather than just using what was already available, it would also be horrendously expensive to produce, especially in very small quantities. By relying on parts / circuitry that is already in production, at least a portion of the benefits of such an approach can be had and prices kept within the "working man's" budget. Even then, some "working men" may still have a problem with the price on some of these units. Sean
>
Not quite, Eldartford. A digital system cannot accurately reconstruct a wave that is exactly half the sampling frequency. That's why I said the sampling rate had to be fractionally higher (granted, a very tiny fraction) than twice the highest frequency. In the example given, there were exactly two samples per cycle, and that wouldn't work.

And just what is it you think is "the biggest problem with redbook CDs"?
Germanboxers,
basically, you are correct in pointing out that 96K & 192K were selected owing to other hi-res audio formats (namely DVD-A). Here the electronics is a multi-rate system wherein it up/oversamples by 160 & then decimates by 147 to change the sampling rate to 48K from 44.1K.

However, if you buy any of SimAudio's products, then you'll find that they oversample at exactly 8X, which is 352.8KHz!!! So, here is one commercial co. that doesn't use 96K, 192K or 384K. There must be others too but I cannot think of them right now.
FWIW.
Pabelson...I guess you mean that if the sine wave frequency is EXACTLY one half the sampling frequency, a sync situation exists. OK. Change the sampling frequency enough so that the phasing of the sine wave drifts across the sampling interval. Picky, Picky :-)

I personally don't have much of a gripe about CDs, but then my ears are 67 years old, and don't have the HF sensitivity of some of our golden eared friends. Based on my experience, which led me to believe that Nyquist was an optimist, I can believe that HF is a lot better with 96KHz sampling.

Sean...I disagree about the effect on quality of "off the shelf" parts. In the military electronics business, we used to design all our own chips, even microprocessors. However, even at great expense we could never match the research and development effort, propriatary skill, and quantity production, typical of commercial products that were functionally equivalent to our designs. A mature "off the shelf" product has had all its bugs weeded out.
Since Sean has confessed his error, I will do the same. My explanation actually showed a ramping signal of 3 units in four samples. While this was not incorrect, it is not consistent with the analog signal that I assumed at the beginning. The following is an updated version of my explanation, for posterity.

Phillips used 4 times oversampling in their first CD players so that they could achieve 16 bit accuracy from a 14 bit D/A. At that time, 16 bit D/A, as used by Sony, were lousy, but the 14 bit units that Phillips used were good. The really cool part of the story is that Phillips didn't tell Sony what they were up to until it was too late for Sony to respond, and the Phillips players ran circles around the Sony ones.

In Sean's explanation the second set of 20 dots in set B should not be random. Those dots should lie somewhere between the two dots adjacent to them.

Here is my explanation.

Assume there is a smoothly varying analog waveform with values at uniform time spacing, as follows. (Actually there are an infinite number of in-between points).

..0.. 1.. 2.. 3.. 4.. 5.. 6.. 7.. 8.. 9.. 10. 11. 12 etc.

If the waveform is sampled at a frequency 1/4 that of uniform time spacing of the example, (44.1 KHz perhaps) the data will look like the following:

..0............... 4.............. 8...............12..
THIS IS ALL THERE IS ON THE DISC.

A D/A reading this data, at however high a frequency, will output an analog "staircase" voltage as follows:

..000000000000000004444444444444444488888888888888812

But suppose we read the digital data just four times faster than it is really changing, add the four values up,
and divide by 4.

First point……..(0+0+0+4)/4 = 1
Second point....(0+0+4+4)/4 = 2
Third point.....(0+4+4+4)/4 = 3
Fourth point....(4+4+4+4)/4 = 4
Fifth point.....(4+4+4+8)/4 = 5
Sixth point.....(4+4+8+8)/4 = 6
Seventh point...(4+8+8+8)/4 = 7
Eighth point....(8+8+8+8)/4 = 8
....And so on

Again we have a staircase that only approximates the instantaneous analog voltage gererated by the microphone when the music was recorded and digitized, but the steps of this staircase are much smaller than the staircase obtained when the digital data stream from the disc is only processed at the same rate that it was digitized at. The smaller steps mean that the staircase stays closer to the original analog continuously ramping signal.

Note also that we are now quantized at 1, instead of 4, which is the quantization of the raw data stream obtained from the disc. A factor of 4. That’s like 2 bits of additional resolution. That’s how Phillips got 16 bit performance from a 14 bit D/A.
The Esoteric DV-50 is another player that does not use 96K, 192K, or 384K. The first upsampling point on the DV-50 is 352.8K, and the higher selections continue to follow that pattern.
Hmmm... I'm surprised that nobody jumped all over me for stating the obvious. That is, digital is a poor replication of what is originally an analogue source.

I'm also glad to see that nobody contradicts the fact that having more sampling points can only improve the linearity of a system which is less than linear to begin with. After all, if digital was linear, we could linearly reproduce standardized test tones. The fact that we can't do that, at least not as of yet with current standards, would only lead one to believe that analogue is still a more accurate means of reproducing even more complex waveforms.

Converting analogue to digital back to analogue again only lends itself to potential signal degradation and a loss of information. One would think that by sampling as much of the data as possible ( via upsampling above the normal sampling rate ), that one would have the greatest chances for better performance with a reduction the amount of non-linearities that already exist in the format. Evidently, there are those that see things differently. Sean
>
Sean, thank you for all the diagrams, and patient tutoring. It all makes logical sense, sure.

If I had never heard the Audio Note, I would be looking at one of the top up sampling players on the market. Oh well.
More corrections! They don't affect the basic idea, but could easily confuse people. Sorry about that. Hopefully this is it.

If the waveform is sampled at a frequency four times that which corresponds to the uniform time spacing of the example, (44.1 KHz perhaps) the data will look like the following:

Note also that we are now quantized at 1/4, (0+0+0+1)/4 ,instead of 1, which is the quantization of the raw data stream obtained from the disc. A factor of 4. That’s like 2 bits of additional resolution. That’s how Phillips got 16 bit performance from a 14 bit D/A.

OK Sean...Sorry you felt left out because no one jumped all over you. The following is my modification of your statement.

Some digital representations of analog (analogue in England) waveforms are a poor replication of the analog source because they lack the resolution (bits) and sampling rate appropriate for the bandwidth of the signal. Inaccuracy is not inherent to the digital format, but represents a design decision regarding what level of error is acceptable.
As long as human beings are analog, the initial & final music will always be analog. What's in between can be digital. Digital is a compromise for an analog signal - no doubt. How good or bad it is depends on how well the digital system is engineered for the 20Hz-20KHz bandwidth. Digital is chosen mostly for its cost effectiveness (scalability of the DSP engines with shrinking CMOS technology) & what Carver Mead once pointed out - its tremendous noise immunity. Corrupting a stream of digital data to the point of making it useless is very difficult as it requires a lot energy to flip a bit. Some bits do get flipped but the overall context of the message is very much retrievable by using various error correction algorithms. This is hardly the case with a purely analog music signal.

Having more sampling points with an estimation filter allows the digital to better track the analog waveform. Whatever benefits one accured with over/upsampling could be lost by distortions in the analog reconstruction filter. Hence, the above mention of implementation. Having somebody engineer a good re-produced sound CDP solution is priceless (for everything else there's MasterCard!).

Yes, if one converts from analog->digital->analog, one does degrade the original sound. That is to be expected as we take only a finite # of samples (hence the term "quantization"). BTW, if we had infinite # of samples, it would analog! In the redbook CD format, the powers-that-were decided in all their infinite wisdom to Nyquist sample the data onto the CD disc . Thus, no matter how much one oversamples, one can never undo this. Hence the rise of "hi-res" music formats. In fact, if the over/upsampling was A1 perfect, you'd get *exactly* what was on the CD, which is Nyquist sampled!! How good is taking just 2 samples of a dynamically waveform music signal? Not very good I'm afraid!
Eldartford cited his experience: 4 samples was worth every effort. I've found 5-8 samples is worth the effort. The difference is that my work is voice-related. Not hi-res by any standards but when people hear another voice at the other end, they do want to recognize it. Need more samples for this.
FWIW. IMHO.
El: Thanks for correcting my previous errors, your previous errors and then confirming my last statements.

The bottom line is that, as good as digital is and can be, it is still trailing behind analogue as we know both formats today. It is too bad that the decisions foisted upon the audio industry when selecting these design parameters were made by those that don't really listen to the products that they produce. Otherwise, we would have started off with wider bandwidth designs and higher sampling rates to begin with, making conversations like this moot. Then again, hind-sight is almost always 20/20 : ) Sean
>
Bombaywalla...It's an interesting question about whether the world is, at its heart, digital or analog. The electrical signal, regarded as analog, actually consists of discrete packets of charge, called electrons. The human sense of hearing is implemented by the "firing" of discrete cells called neurons.

Of course this digital condition exists at so fine a level that it is entirely reasonable to consider the process as being continuous. But it does point out the fact that at some point, and we can debate where that point is, digital audio becomes, for all practical purposes, the same as analog.
Eldartford,
Good observation indeed! It would be like splitting hair, IMHO, to find out where that analog-digital demarkation is!! I'll let the others "battle" it out while I listen to my music.