[Dusty Chalk]

Quote:

Originally posted by AndreYew And what is so terrible about a ringing pulse response? It may look terrible, but we don't listen with our eyes.

No, there have been many reports that this is probably the main problem with PCM and CD. Mark Levinson, for one, thinks this. Quote:

What the original poster was talking about was leaving out the analog filter entirely, and that's what I was responding to.

The original poster in this thread said nothing of the sort. I'm not sure to which "original poster" you are referring, but if you're talking about my references to pictures provided by MSB, that is not at all what I was implying. Yes, those pictures are taken before the analog filter stage, but the point was to show what it does, just as many of Joe's pictures that he quoted showed intermediate stages to show how the ultimate result is achieved.

 

[MirandaX]

Quote:

Originally posted by Dusty Chalk I go back to the sine waves, and that's not a fair representation of "sound", because (IMHO) sound is not made up of sine waves, but it's better than nothing, which is what this whole discussion is about (to me).

You're both right and wrong. Physically, sound is a function of air pressure versus time. In other words, what our ears "sense" are changes in air pressure on the cillia in our ears.

That's all sound is.

Math, however, tells us that *any* function can be thought of as a set of sine waves, called the Fourier basis. This breakdown is perfect -- it's not an imprecise engineering concept/approximation of the real world -- it's purely a mathematical tool for thinking about signals in a different way. We tend to like thinking about sound as a set of sine waves, because it's natural; each sine wave represents a frequency component in the original signal. But we don't have to. There are other ways of representing/thinking of sound (and people have done this, e.g. the wavelet and Bernstein bases), but they don't tend to have any correspondence to things we think of in the real world, so they're not widely used.

 

[Dusty Chalk]

Quote:

Originally posted by MirandaX Math, however, tells us that *any* function can be thought of as a set of sine waves, called the Fourier basis. This breakdown is perfect...

I guess my point was that I don't know how useful this mapping is to audio. Noise is basically all frequencies, yet I think it is more useful to analyze as noise, rather than to have a component of the signal that is "all frequencies". Like a snare drum attack. I'm not saying that it can't be accurately represented using Fourier analysis, I'm just saying I'm not sure how useful it would be to do so, or to think about it that way.

That was the point I was trying to make. It's an opinion, not a fact. Sorry if I didn't make that clear.

 

[Joe Bloggs]

Say, dusty,

1. What have you got to say on the topic of what's the difference between over- and upsampling? Seeing as that article shows that oversampling already includes interpolating intermediate values...

2. What do you mean by dither? Judging by the article I just read and linked to, there is one sense of 'dither' that is quite useful and not contradictory to over/upsampling (although not as good as noise shaping, but that's another thing again)

Also,

Quote:

"ok, um, do upsampling DACs use oversampling filter chips?" Yes, but the OS does very little. However there is another issue not considered here: the quality of the 44.1 kHz brickwall filter in oversampling chips vs. those in sample-rate converters (which are the chips typically used in most so-called upsampling DACs). People have done listening tests wherein the quality of the brickwall filter has been very important. That's one thing to consider why upsampling may sound better --- it may just do a better job of 44.1 kHz brickwall filtering than your run-of-the-mill OS filter. The HDCD filter was well-known as being much better than most other OS filters. Nothing magical, just better engineering.

Uh, bear with me--so are the OS chips in upsampling DACs set to oversample as well? E.g. In the specs for the ART DI/O I saw 128x oversampling as well as upsampling to 96kHz...

(the funny thing is that the DI/O does NOT add bit depth but still calls it upsampling. Oh the irony...)

 

[Musiclover4]

Joe,
the DIO does not have any upsampling. If I remeber correctly, it, like many pro audio units, has 128x oversampling. The clock select is purely to determine if the master clock is derived from the DIO itself or the transport in external sync mode.

From what I understand, I have to side with Andre Yew, although I do think his statement about ringing is a little off-base. As Dusty pointed out, ringing is one of the problems many feel is the key to unthe dissatisfaction with redbook audio. About a year ago I read a bunch of technical papers and was following some very good technical debate over at Audio Asylum about the process of upsampling. After all of that I came away feeling that it was just a slightly tweaked method of digital filtering. Definitely not the end all be all, more like a slightly more complicated oversampling, though it's difficult to tell because the company's try to be ambiguous and not let anyone know if the process is real or just marketing jargon. I just decided to not worry about it and not look at it like a feature, instead just go with how it sounds. I do think the digital filtering is vastly underrated in importance, but proper design and implementation is much more important than a certain type.

Oh, and bit-depth is just like Andre Yew said in upsampling. It is a side effect. I had not heard his idea that it is due to the calculations and leaving it at 24 bit for simplicity's sake, but that could very well be. Otherwise they just dither for 24 bit depth and actual 17 bit resolution. The P1-A, though, actually does use a different algorithm that tries to predict what the signal would have been back when it was 24 bit, but it is not liked by all since I've heard many stories about people preferring it off.

For the record, I'm no engineer and this is all based on my interpretations of what I read about a half a year ago so I may be off base. I am confident, though, that upsampling is definitely not a must have feature and more like a slightly different implementation of something we've had for a long time.

 

[Joe Bloggs]

Musiclover4,

There is an external lock mode, which is when the DI/O just locks onto the clock of the outside source, and then passes that directly to the filter and DAC chips. But there are also other settings, like the 88.2kHz setting--which is definitely 'upsamling' of some sort.

To others:

about the ringing--is this in the digital domain--an artifact of the FFT?

Hmm, seems like we are stuck between a rock and a hard place here--use oversampling + digital filtering and get ringing (and then an analog rolloff filter of course) or use an analog brickwall filter and get phase distortion and ****...

Of course, it's long since been accepted that the analog brickwall is not an option...

It's so ironic that audiophiles are all so down with EQ when in fact all your CDs pass through a digital EQ of a sort (the digital filter) before reaching your ears

From the discussion here and from the 'Audio DAC' project page, it seems to me that the pinnacle of current CD DAC technology is oversampling (or upsampling or whatever

) with dither and noise shaping. Noise shaping, in particular, in all probability implies interpolation (the much-vaunted difference between over- and upsampling?)

Now then the strange thing is there doesn't seem to exist a filter / DAC combination in the bottom of that page that DOESN'T include dither and noise shaping.

 

[nec]

Quote:

nec Here's a scheme

Quote:

Dusty Chalk Where do you get this?

Quote:

nec From the books on digital signal processing that I found yesterday at the library at work. I don't have titles written down; I'll look them up again next time I get to the office.

Quote:

Dusty Chalk one of the reasons you're having a hard time with it

Thank you for trying to explain this to such an idiot like myself...

Anyway... I like to keep my promises so today I got over my laziness and drove to the office. Here's a list of books I was talking about before:

Madisetti, Vijay K. (ed), Williams, Douglas B. (ed)
The digital signal processing handbook
IEEE Press, 1998
ISBN 0849385725
pp 43-2 - 43-4

Proakis, John G.
Digital signal processing, 3rd edition
Prentice-Hall, 1996
ISBN 0133737624
p 774

Webster, John G. (ed)
Wiley Encyclopedia of electrical and electronics engineering, vol 5
John Wiley & Sons, 1999
ISBN 0471139467
pp 533-534

The bottom line:
1. There's interpolation needed (and is happening) in oversampling DACs. You can not "just put zeroes".
2. Don't confuse digital and analog filters. Digital filter interpolates oversampled data. Analog filter on the output of DAC filters out high-frequency noise generated during D/A conversion.

 

[AndreYew]

DC,

I apologize for any snideness on my part.

"First you're saying that it doesn't do anything to the bit depth, now you're saying that what it does to the bit depth is just a side-effect of the arithmetic. Which is it?"

It doesn't have anything to do with bit depth because the bit depth is a side effect of the computation.

"Alright, let me try this another way. Do you agree that, if a 16/44.1 signal could be accurately* upconverted to a 24/96 or 24/192 signal -- say, for example, converting it to the frequency domain and then back again -- that it would be an improvement to the signal?"

I don't agree. Accurate upconversion should do nothing to the baseband signal. You don't need to go to the frequency domain to do upsampling, BTW. Almost no one does it this way.

I also disagree on ringing being a problem with PCM and CD. First, I would not trust anything Mark Levinson (the Madrigal company) says about digital. Look inside their flagship DACs and transports, and you'll see an overpriced, underengineered piece of equipment. Those guys could not design their way out of a wet paper bag. Wadia hardware engineering is really the good stuff, but their software filters are terrible.

What has more credibility, and has listening tests to back it up, is that the majority of 44.1 kHz brickwall filters are bad, ie. audible. Good ones can be transparent, but those are computationally expensive, and uncommon.

Joe,

Ringing is caused by the initial sampling, and the necessary reconstruction filter used on playback.

--Andre

 

[ddriveman]

Originally posted by Nezer


here's the deal, this thread started-off with some poor soul asking about reviews or opinions of different products and it's turned into a pissing match over what coler the bike shed should be (and by a bunch of folks that don't even build bike sheds!)! ....

Dear Nezer,

Thank you for getting us back to my orginal post. I had no idea that my post would start such controversy over upsampling vs oversampling. I do agree with Kelly though that I do find the responses interesting and informative myself.

When I wote "upsampling" DAC in my original post, I was indeed referring to DACs that:

1) Interpolated 16 bits to 24 bits. My current Musical Fidelity X-Ray CD player does this already. I truly enjoy this CD player and would not think of going back to a 16, 18 or 20 BIT DAC

2) Upsample original 44.1 kHz sampling freq. to either 96 kHz or 196 kHz. I was intrigued by upsampling ( as describe by the likes of Bel Canto, Perpetual Technologies and DCS) through the reviews in various Hi-Fi magazines. I recommend checking out their technology paper on this subject at http://www.dcsltd.co.uk/papers/aes97ny.pdf.

I have had the opporunity to listen to the Musical Fidelity A3-24 DAC at both 96 kHz and 196 kHz upsampling frequencies (this unit allows you to switch between the two). I was astonish by the easily discernable difference between the 2 upsampling frequencies, with my preference going to the 192 kHz upsampling freq. So why would a difference of upsampling freq. in the sam DAC with the same filter topology etc make a difference. Go figure.

Anyway, thank you for the responses and happy listening all.

 

[Nezer]

Quote:

Originally posted by ddriveman [BI have had the opporunity to listen to the Musical Fidelity A3-24 DAC at both 96 kHz and 196 kHz upsampling frequencies (this unit allows you to switch between the two). I was astonish by the easily discernable difference between the 2 upsampling frequencies, with my preference going to the 192 kHz upsampling freq. So why would a difference of upsampling freq. in the sam DAC with the same filter topology etc make a difference. Go figure. Anyway, thank you for the responses and happy listening all. [/B]

And if I may ask, what was the difference to your ears? I'm looking at some of these DACs and such and have many of the same questions you have.

Is it worth it, to your ears, to buy one?

It's awfully hard to sample these things without buying them and who wants to lay-out that kind of cash and take the risk? Before I lay out that kind of cash on ANYTHING I have to be DAMNED sure or else the girlfriend will just go nuts (and rightfully so) if I order it then want to send it back! This would likely put the brakes on buying a second product if the first one didn't work-out.

 

[ddriveman]

Originally posted by Nezer


And if I may ask, what was the difference to your ears? I'm looking at some of these DACs and such and have many of the same questions you have....

Dear Nazer.

When comparing the 192 kHz upsampling freq to the 96 kHz upsampling freq on the MSF A3-24 DAC, I found that the higher upsampling freq gave greater presence and warmth to the music. There seems to be more airiness around each instrument leading to a somewhat wider soundstage. On vocals, the sibilants (i.e. Shhh..) tends to linger a little longer. In either case, the A3-24 DAC even with 96 kHz upsampling gave similar improvements over the MSF A3 CD player by itself. The difference with the higher upsmapling freq is that the improvement was more noticeable.

Which is more accurate ? The A3-24 DAC or the A3 CD player by itself ? One could argue that the "improvements" I heard are actually colourations to the actual recorded music. However, to me, I found the A3-24 DAC more natural sounding and hence more enjoyable. Hence my thirst for similar sounding DACs and a quest for the best DAC upgrade for around US$1500.

 

[kelly]

Quote:

Originally posted by Nezer It's awfully hard to sample these things without buying them and who wants to lay-out that kind of cash and take the risk? Before I lay out that kind of cash on ANYTHING I have to be DAMNED sure or else the girlfriend will just go nuts (and rightfully so) if I order it then want to send it back! This would likely put the brakes on buying a second product if the first one didn't work-out.

*cracking whip sound* Damn Nezer, I'd be in trouble if I couldn't audition by buying and selling. Tried explaining the "trial" thing to her?

 

[Nezer]

Quote:

Originally posted by kelly *cracking whip sound* Damn Nezer, I'd be in trouble if I couldn't audition by buying and selling. Tried explaining the "trial" thing to her?

To be entirly honest, it's not just her... I think long and hard ebfore spedning $300 let alone $3000. When I drop that kind of cash I want to be damned sure it's the right product for me and it will be sticking around in-use for a while.

 

[kelly]

Quote:

Originally posted by Nezer To be entirly honest, it's not just her... I think long and hard ebfore spedning $300 let alone $3000. When I drop that kind of cash I want to be damned sure it's the right product for me and it will be sticking around in-use for a while.

My philosophy is to audition or borrow whenever possible. When not possible, look for used so that I can resell and minimize losses. Only as a last resort will I even buy new--if I do buy new, I'm looking for a return policy AND a price cut. Buying something I've never heard with no return policy and at full retail? Not me.

 

[bifcake]

Since we're on the subject of DACs, I have a question:

A friend of mine owns the Mark Levinson Model 31 DAC without the upsampling upgrade ($3500). He's considering getting the Musical Fidelity A3 24 DAC and sell the Mark Levinson. Does anyone think that's a good way to go? The MF DAC is less than a third of the Mark Levinson upgrade and I wonder, whereas the upgraded Levinson DAC may sound better than MF, it may still be worth his while to buy the MF DAC, which will sound better than the non-upgraded Levinson, which he'll sell to more than offset the cost of the MF DAC.

Any opinions?

Thanks