I will be upsampling CDs to 24/192 on the cheap: $299! (1 Viewer)

[Wes Stover]

The TX-DS989 does upsampling also as well as the Integra DTR-9.1. No surprise...

 

[KeithH]

Wes, what is no surprise? The fact that a company's flagship receivers upsample? As was discussed earlier in this thread, Onkyo's 'DS797 and 'DS898 receivers also upsample. I am surprised to see home-theater receivers that upsample and still wonder how good they are at doing it.

 

[Chip E]

Hmmm, i'll have to look and see if my Denon does it Keith. I think i read in the manual it does?... not positive. Hell, i have no experience with this either!

 

[KeithH]

Chip, in a cursory read of the features of the '5803 on Crutchfield's web site, I saw no mention of upsampling. However, the Onkyo 'DS989 upsamples, and Crutchfield says nothing about it.

 

[Saurav]

However, I have little firsthand experience with upsampling, as I said, so I cannot say one way or the other whether upsampling is beneficial.

It is also entirely possible that it will work better with some systems/components/tastes than with others.

 

[KeithH]

Saurav, I figured you were some sort of engineer, or something.
You said:

It is also entirely possible that it will work better with some systems/components/tastes than with others.

Absolutely. As I said in other posts, I wonder how well the Onkyo receivers implement upsampling. Furthermore, I would expect the Musical Fidelity A324 DAC at $1200 to handle upsampling better than the $299 upsampling board for the Ah! Njoe Tjoeb 4000 player. As I said, I will be doing upsampling on the cheap!

 

[Justin Lane]

Saurav,

From what I know (not saying I am an expert), the only reason to increase sampling rates is to eliminate aliasing effects created by overlapping frequency sprectra. These overlapping spectra change the original signal, creating what could be called distortion.

Examining the Fourier Transform of such a distorted signal you could imagine it as a series of sawteeth where overlapping regions create a raised horizontally flat region between the teeth distorting the signal. An undistorted signal could be described as a series of sawteeth where one tooth ends the next begins.

CDs are already sampled at 44.1 kHz ensuring the entire audible range will be acurately represented due to the Nyquist theory (sampling rate should be double the highest frequency desired or more). Upsampling will not hurt anything, but I cannot see where it will help in the case of a CD which is sampled at a set frequency and does not contain anymore information. In the original mastering process upsampling could have an effect though, but not after the fact.

What does this prove? Nothing really, until Keith receives his upgrade. If it ends up making his CDs sound better, then it will be worth the cost. I do not think the upsampling will be the reason for any improvement (better DACs most likely), but he will have to ultimately be the judge.

J

 

[KeithH]

Justin, I will certainly give a report when I receive the upsampling board for the Ah! Njoe Tjoeb 4000. Someday I'll try a quality DAC like the Musical Fidelity A324 to compare 16/44.1, 24/96, and 24/192. Perhaps you should do the same thing.

 

[Mark Tranchant]

(Justin,) Upsampling only doesn't hurt anything in the "old studio" meaning of the word: changing the sample rate from x to mx where m is an integer. Changing to an unrelated sample rate (e.g. the 9:4 conversion rate that 44.1 to 192 is) will introduce distortion to the original signal.

There is nothing that can be achieved by upsampling then passing through the higher rate DAC that a good low rate DAC cannot do. The job of the DAC is to interpolate analogue signals between the digital samples it is given to best re-create the original analogue waveform. The quality of an "ideal DAC" (no noise or distortion) is solely down to the interpolation algorithms. Higher resolutions and sample rates in the original data make accurate interpolation easier.

The upsampler/DAC combo does some of this interpolation in the digital domain: the upsampler adds some more "faked up" data samples, and the higher resolution / sample rate DAC then handles the conversion to analogue. For a multiplied sample rate (88.2 or 176.4) or higher resolution (20, 24 bits), all of the original sample points are retained, with digitally interpolated samples in between them. For a 44.1 to 196 conversion, the output needs to have nine equally-spaced samples in the space of four original samples, so most of the original data will be discarded.

I'd guess that a good engineering reason for doing the upsampling is that you can then build your DAC more simply, to just process 24/96 and 24/192 signals, but have the unit as a whole compatible with 16/44.1 as well.

Perhaps you should write a letter to Stereophile voicing your objection.

Yeah, like that'd work. They must get hundreds of letters a week pointing out how daft some of the stuff they write is.

 

[Justin Lane]

Mark,
Thanks for some clarification on how exactly these upsamplers work.
I guess "sound quality" comes down to the algorithm implemented by the upsampler/DAC. The more expensive upsamplers may use a better/different algorithm, but then again that may not be the case.
Keith,
I will let you be the lab rat as far as upsampling is concerned I will be waiting to see what you think once you get your upsampler before taking the plunge.
J

 

[Saurav]

Mark,
That's a good point. If your design budget allows you to only use a DAC of a certain quality/price-point, you might get better sound by upsampling and using a higher resolution DAC, assuming you cannot afford to use a 16/44.1 DAC of higher quality.
Justin,

Quote:

Quite possible. Based on your knowledge and understanding, you will try some devices and reject others that don't make sense. I do the same, everyone does. And like you said, at the end, all that matters is if it sounds better or not.

 

[Mark Tranchant]

Mark seems to have a much better understanding of the issues involved, so maybe he can help us out here.

...or maybe I'm just a better bluffer.
I'll see if I can make my point a bit clearer. Imagine, if you will, a line of wooden posts of various heights spaced 1m apart, holding up a gently curving metal beam.
OK. The beam is the original analogue waveform, and the posts are the digital samples. The CD stores the post heights only - the exact shape of the curved beam is lost. This is what determines the maximum frequency response of a data stream - very abrupt changes in the beam's shape cannot be reconstructed given only the post heights.
Now, once you have your CD with post heights stored on it, you cannot do anything to learn any more about the beam's original shape than use these post heights to guess the "best fit" shape. That is what the DAC does, using tricks like cubic interpolation to try to fit a curve to all the points.
Simple upsampling to a multiple of the original frequency (such as 44.1 to 88.2) involves building the posts from the CD, then working out intermediate post heights halfway between the originals, then trying to fit a better curve. As the DAC's curve-fitting algorithms are fairly complex, it can be easier to perform some of it digitally. Think about it like an artist - the upsampling draws in extra, guessed construction lines before the freehand curve is drawn. This can be easier but not necessarily more accurate.
Upsampling to a non-integer multiple (such as 44.1 to 192) involves building the original posts from the CD, then putting in an adjacent line of posts at a different spacing (nine new within the space of four old for 44.1 to 192) whilst trying to calculate the appropriate heights. Then the old posts are removed and the curve fitted to the new, mostly guessed posts.
This must be worse than (or at best, equal to) the first case, which at least retains the original data. The question is whether the approximations involved in building the new line of posts (the resampling + high-rate DAC) are greater than the approximations the freehand artist makes trying to fit a curve to the more widely-spaced original posts (the original low-rate DAC).
I hope that helps clarify the issues involved with sampling rate changes - I haven't touched on bit resolution here. I am not saying that it is not possible for a 16/44.1 signal to sound better when up/resampled to 24/192 before playing when comparing two specific systems; I am stating that the method is not inherently capable of superior D-A conversion.
I still think the main reason would be for DAC simplicity - the design and engineering work can go into making a top notch 24/192 DAC without having to patch in legacy support for 16/44.1. It's really a piece of value engineering on the manufacturers' part rather than a tweak for the golden-ears brigade, but it's a known fact that golden ears rarely listen to reason.

 

[Saurav]

Still doesn't address my first question though - what does the filter after the DAC do? What frequency does that filter operate at? And, will upsampling make any difference to the implementation of that filter?

 

[KeithH]

Mark, your attitude about writing to Stereophile is like saying, "Why should I vote? It's only one vote." If somethings bothers you ('you' meaning anyone) enough, write a letter and tell them about it. Maybe they will publish the letter. Perhaps you aren't the only one with the complaint. If enough people issue the same complaint, perhaps the editor will take notice.

 

[Kevin C Brown]

Mark- I learned something new today. Thanks!

I did not know that even simple 16/44.1 DACs *can* do interpolation and noise shaping on their own.

 

[John Mannix]

Gentlemen,

I am not an electrical engineer. Just thought I'd get that out of the way up front. However, I do love music, and I think I've got a pretty discerning ear. Personally, I don't care what a gizmo does or does not do, as long as what's reproduced sounds live, and conveys the emotional content of the music.

Having said that, I too am interested in the Musical Fidelity A324 and the Bel Canto DAC-II. Not specifically because they do "upsampling", but because they're considered to be outstanding DAC's for the dollar.

If fact, the best CD player I've ever heard does not have upsampling... at least that I've ever read or been told... and it sounds FANTASTIC. I'm speaking of the Linn Sondek CD12.

The DAC I'd dearly love to own, and I'm quite confident would beat either the Musical Fidelity A324 or the Bel Canto DAC-II hands down... and is the DAC that the Perpetual Technologies P-1A/P-3A combination is said to "approach"... also does not have "upsampling". I'm speaking, of course, of the Mark Levinson ML-360s. Now that's one smooth, sweet, honey of a DAC!

At any rate, what I'm trying to indicate is that I think "upsampling" has less to do with the overall sound quality than the total package: design of the digital and analog stages, design of the power supply, component quality, construction, attention to detail, etc. However, I think there may be a possibility that upsampling can improve the perceived sound quality of more cost effective DACs.

Like anything else, upsampling can undoubtedly be done "right" or done "wrong" - and with this particular technology being the current buzz word and "hot ticket" it wouldn't be surprising for everyone to jump on the bandwagon... without doing the research, investigation, and design requisite to do it right.

That's why I'll put my money on Musical Fidelity and Bel Canto - two highly regarded companies who have a proven track record, and with whom the upsampling DAC I'm interested in is not their initial foray, but a refinement of a proven product that they've been working on for years.

 

[Doug_B]

Musical Fidelity does provide some verbiage wrt upsampling as part of their Nu-Vista 3D literature that appears on their Home Page.
Doug

 

[Mark Tranchant]

Kevin - the DAC as a package interpolates voltage by definition. You give it two samples a certain time distance apart, and it gives you a time-continuous analogue voltage changing from the first sample to the next. At any point between the samples, you have an interpolated output voltage.

In reality, this is probably done with a simple low-pass filter in most cases: the DAC just holds the sample voltage until the next one arrives, and the filter smooths the steps. In a properly designed system, this will approximate to linear interpolation.

There is no theoretical reason why advanced methods of interpolation cannot be used, but these are hard to implement in the analogue domain. If you choose to do it digitally, you're effectively upsampling...

 

[KeithH]

John, certainly the design of the DAC will determine the ultimate performance. That said, I still would be interesting in comparing a quality DAC like the Musical Fidelity A324 in 16/44.1, 24/96, and 24/192 modes. Given the difference in price, I would expect the Mark Levinson DAC to outperform the Musical Fidelity unit running in any mode. However, if I preferred the Musical Fidelity unit in upsampling mode to using it in 16/44.1 mode, I would naturally be curious to hear what the Mark Levinson model could do if it upsampled.
In the end, I am hoping that the $299 upsampling board for the Ah! Njoe Tjoeb 4000 incorporates a quality DAC. A person who posts on Audio Asylum under the moniker "janis" claims to live in Holland, where the Ah! player and upsampling board are manufactured. This individual further claims to live near the dealer manufacturing these products and to have heard a prototype upsampling board in the Ah! player. The person has been saying for weeks that the upsampling board is outstanding. I hope this is true.

 

[Sihan Goi]

Interesting...I tried upsampling on my HTPC using a Audiophile 2496 soundcard's S/PDIF output to my TagMcLaren AV32R. I tried it on many source CDs and mp3s on 44.1KHz 48KHz, 88.1KHz and 96KHz and could never really tell the difference, but then I came across a 128kbps mp3 that my friend recorded of himself with a guitar, and the 88.1KHz and 96KHz version just sounded better, with the guitar being more natural and clear...and this is with a 128kbps mp3. I'm going to ask him to send me the original PCM wav file...