[Duncan]

Thanks for that NEC

Was a good read... meaning that everything is as good as the same as everything else (if i read it correctly

)

 

[DanG]

Quote:

Originally posted by markjia From what a professor, specializing in signals and systems, tells me, upsampling is effective at upto 2x. Anything beyond this is pointless and unnoticable.

This same person would likely argue that DSD information is overly excessive, as is DVD-A's 24/96 and 24/192 data streams. Would audiophiles agree? No. As we often argue about here, it's hard to support or refute normative claims of benefit when it relates to audio equipment, and objective claims cannot be supported since we don't always know what numbers to look for. But according to the listening tests of many who have heard upsampling at over 2x, the difference is noticeable and often beneficial.

 

[Dusty Chalk]

Quote:

Originally posted by nec In case anybody is really interested in facts and not in marketing BS: http://www.madrigal.com/upconversion.htm

First of all, most of my post, other than the MSB link, was not marketing BS, I am in no way affiliated with any company in audio. It just struck me that a picture is worth a thousand words, and those pictures explained a lot to me.

Secondly, ...this is from Madrigal? OMG, I so completely disagree with this. For example, this quote: Quote:

Obviously, a "96/24" (96,000 samples per second at 24 bits per sample) recording using a Fisher- Price microphone will not sound significantly different than the 44/16 version of the same thing.

"Obviously"? Please, don't insult me. Guitarists go to great lengths to get their analog tone/distortions/etc. Yes, they even use esoteric and unorthodox items such as a Fisher-Price microphone. Sure, it specs badly, but it's a completely different distortion than that introduced by digital. I would still much rather hear Lou Read's Metal Machine Music on 24/96 rather than non-upsampled 16/44.

Noise is not noise. There are many many varieties.

Now, that said, I do have a non-upsampling DAC from the now defunct company of Audio Alchemy that I listen to on a daily basis. It is wonderful. I would have a hard time replacing it. So again upsampling is not the be-all and end-all, but...you gotta hear it.

 

[Nezer]

Quote:

Originally posted by Dusty Chalk First of all, most of my post, other than the MSB link, was not marketing BS, I am in no way affiliated with any company in audio. It just struck me that a picture is worth a thousand words, and those pictures explained a lot to me. Secondly, ...this is from Madrigal? OMG, I so completely disagree with this. For example, this quote:"Obviously"? Please, don't insult me. Guitarists go to great lengths to get their analog tone/distortions/etc. Yes, they even use esoteric and unorthodox items such as a Fisher-Price microphone. Sure, it specs badly, but it's a completely different distortion than that introduced by digital. I would still much rather hear Lou Read's Metal Machine Music on 24/96 rather than non-upsampled 16/44. Noise is not noise. There are many many varieties. Now, that said, I do have a non-upsampling DAC from the now defunct company of Audio Alchemy that I listen to on a daily basis. It is wonderful. I would have a hard time replacing it. So again upsampling is not the be-all and end-all, but...you gotta hear it.

I was going to say... Doesn't Less Claypool use these mics in his studio?

As for me, I get my tone with digital modeling of vintage gear.

 

[bearwise]

I have a bit of a DAC fetish and own several attached to various rigs. I have the P-1A/P-3A in my main system and it is phenomenal, but I can't say that is due to the fact that it upsamples. It does beat the pants of the (non-upsampling) CAL Alpha that i formerly used and the CAL Sigma (24/96) that i use in one of my headphone rigs. I have only heard a BelCanto 1 briefly at a dealers and thought it was good but I wasn't blown away (although a quick audition in a strage system in a different room is hardly fair). LOts and lots of reviewers love the 1.1 and the 2 is getting raves already. One relatively (~1k$) inexpensive DAC that I have that isnt talked about much is a Birdland Odeon Lite. A bit cheesy looking (small blue plastic case), mine was heavily modded (superior electronics and power cord), but is upsampling and sounds superb, I use it in my main headphone rig.

 

[nec]

Quote:

Originally posted by Dusty Chalk First of all, most of my post, other than the MSB link, was not marketing BS, I am in no way affiliated with any company in audio. It just struck me that a picture is worth a thousand words, and those pictures explained a lot to me.

I think you missed the point of the article. In summary it explains that upsampling is by default used in all DACs and CD players since 1984. Upsampling and oversampling are just different names for the same process.

Sticking word "upsampling" on the DAC is a marketing ploy by manufacturer to fool people into thinking that their product is inherently better than the competition.

Even worse: some ads (for example MF's on the last page of 06/02 Stereophile) imply that by using upsampling DAC you magically transform your CDs to SACD level.

Here's another short article:
http://www.stereophile.com/showarchives.cgi?344
Amusingly it was written by Stereophile's editor; and now Stereophile hosts those funny MF ads that I just mentioned.

 

[Duncan]

Quote:

Originally posted by Mel4 I will try and hunt down a good quality lead of this type and let you know how I get on...

Mel4...

Maybe of help, maybe not...

MY approach to this was to buy a standard BNC-BNC cable and also this adapter (for plugging into the phono plug, digital out of my CD Player) and I can't say I notice anything degrading to the sound, and it definetly does to my ears at least sound superior to optical... (bear in mind that in the UK Tandy is the name for Radioshack)

OR... if you can get your hands on a BNC-Phono connector, could you please let me know the manufacturer, and the cost?

Thanks

 

[Dusty Chalk]

Quote:

Originally posted by nec I think you missed the point of the article. In summary it explains that upsampling is by default used in all DACs and CD players since 1984. Upsampling and oversampling are just different names for the same process.

No, I got that, I was just so flabbergasted that I did not repeatedly "dis" it. It is my understanding that these are not the same thing. Oversampling does not introduce new "interpolated" values, as the article states. It introduces intermediate zero values, as your second link states. It is then the analog filter's job to do this so-called "interpolation". This is more work for the analog filter. Doing this interpolation in the digital domain makes less work for the analog filter.

For example (meaning, as I meant last time, that this is only one datum of my argument, not the entirety of my argument) how could they provide intermediate values of a 16-bit signal on a 16-bit DAC, which is the most those older players had, back when oversampling was first introduced? Quote:

Sticking word "upsampling" on the DAC is a marketing ploy by manufacturer to fool people into thinking that their product is inherently better than the competition.

Nope. For example, in the MSB Tech Link DAC III, it's an upgrade. There's a whole separate chip/daughter board. Installed it myself.

Also, there are standalone upsampling devices, such as the GW Labs DSP and (going from memory on the name of this one) the Sonic Frontiers D2D-1. Their input: a 16/44.1 digital signal; their output: a 24/96 digital signal. I have physical proof and evidence that these do more than nothing.

I understand where you're coming from. When upsampling was first introduced onto the market, there was some confusion, and quite a bit of skepticism about what they were doing, if anything. This trickled up into the magazines. (The same thing is happening with DSD and high-res PCM.) But I'd be quite curious about all the follow-up that occured since then, and what the "audiophile" conclusions about these are now. Not the company line, which, as you've implied, can be affected by contributing advertisers. Quote:

Even worse: some ads (for example MF's on the last page of 06/02 Stereophile) imply that by using upsampling DAC you magically transform your CDs to SACD level.

Alright, have a hard time disagreeing with that. That's poppycock. One of the reasons that I got the GW Labs DSP is that it has a nice little feature wherein it also downconverts a 24/96 signal to CD level. I was doing this so that I could take some of my DAD's, downconvert them to CD's 16/44.1, then upsample them back to 24/96, and see if I could tell the difference. I suspect I will. They (upsampled CD and fully high-res material, such as DAD's) are not the same. Quote:

Here's another short article: http://www.stereophile.com/showarchives.cgi?344 Amusingly it was written by Stereophile's editor; and now Stereophile hosts those funny MF ads that I just mentioned. [/B]

Yeah, well, it would be nice if Stereophile was consistent. It's not like they've never been wrong before. I rememeber reading in a paragraph somewhere where the criticism of "transients" was invoked against DSD. The answer that they came back with? Something to the effect of, we asked Sony, and Sony said that SACD was really a hybrid of DSD and PCM. I would still like to see some follow-up on this, because every article that I've read since, both in Stereophile and otherwhere, contradicts this.

I think the problem is that it is not sufficiently understood, and it is hard to understand, because the two are so close. It really is getting down to the nitty-gritty, but they are different.

 

[nec]

Quote:

Originally posted by Dusty Chalk It introduces intermediate zero values

That is plainly wrong. If you put three zero values between each two points that you extracted from CD you would hear constant noise - popping sounds every 1/(44100*4) seconds.

They add zero bits (not zero values!) to the values they extracted from CD.

Quote:

Originally posted by Dusty Chalk how could they provide intermediate values of a 16-bit signal on a 16-bit DAC, which is the most those older players had, back when oversampling was first introduced?

What do you mean "how"????
They could take a sum of two values and divide it by 2. What's the problem?

16 bit words allow us to distinguish between 2^16=65536 values. Let's assume that we code them as integers between 0 and 65535.

For example let's assume we do 2x oversampling and we extracted two subsequent values from CD: 10 and 15. We take average of 10 and 15 which is 12.5. Because we can only code integers we take 13 as interpolated value:
so on input we get 10, 15, ... at 44.1kHz
on output we have 10, 13, 15, ... at 88.2 (44.1*2) kHz

Now let's assume we have 20 bit DAC. We can use 4 additional bits to code fractional part and represent interpolated values with greater precission. In our example we would have 10.0, 12.5, 15.0, ... on the output - we didn't have to round 12.5 to 13 because we had extra space (4 bits) to store fractional part.

So having 20 bit DAC versus 16 bit allows us to represent interpolated values with greater precission - but that's all. You can do oversampling even if you have 16 bit input and 16 bit DAC.

Quote:

Originally posted by Dusty Chalk Also, there are standalone upsampling devices, such as the GW Labs DSP and (going from memory on the name of this one) the Sonic Frontiers D2D-1. Their input: a 16/44.1 digital signal; their output: a 24/96 digital signal.

They add 8 (24-16) less significant zero bits to the values they extracted from CD and they calculate interpolated values with greater precission than 16 bit DAC could.

Quote:

Originally posted by Dusty Chalk it would be nice if Stereophile was consistent.

It's not a question of consistency but a question of moral: "Can you print lies (misleading ads) if you've been paid high enough?"

Quote:

Originally posted by Dusty Chalk I think the problem is that it is not sufficiently understood, and it is hard to understand, because the two are so close. It really is getting down to the nitty-gritty, but they are different.

When I read that the physics of cable processes is not sufficiently understood I can believe that. But something that was programmed and implemented in silicon 20 years ago can not be "not sufficiently understood".

And this is not hard to understand. In fact I believe that Madrigal's article does very good job in explaining the process in layman terms - you just need to read it thoroughly.

 

[MirandaX]

Quote:

Originally posted by nec That is plainly wrong. If you put three zero values between each two points that you extracted from CD you would hear constant noise - popping sounds every 1/(44100*4) seconds.

Actually, Dusty Chalk is right. DACs don't work in the time domain; they work in the frequency domain. A major topic in most basic courses in signal processing is that it really doesn't matter what values you insert in between each two points, so most DACs that oversample do indeed insert zeroes (or noise). The reconstruction filter generates the same output (theoretically) regardless of what you insert in the time domain. The point of oversampling is to make it easier to construct a good reconstruction filter, not to interpolate the source data.

 

[DarkAngel]

Miranda
You have come to the essence of the matter:
>>>>>>>>
A major topic in most basic courses in signal processing is that it really doesn't matter what values you insert in between each two points, so most DACs that oversample do indeed insert zeroes (or noise). The reconstruction filter generates the same output (theoretically) regardless of what you insert in the time domain. The point of oversampling is to make it easier to construct a good reconstruction filter, not to interpolate the source data.
<<<<<<<

As I stated in my post earlier:
>>>>>>>
Very simply put, the advantage of upsampling (24/96, 24/192 etc) and the reason it sounds better is that it allows removal of noise and distortion from the DAC process in ways that don't negatively effect orignal music signal. CDs sound much more natural
<<<<<<<<

The ability to more effectively filter noise/distortion created by the DAC without harming the original signal is the main advantage of upsampling. No need to get bogged down in all these technical discussions.

 

[nec]

Quote:

Originally posted by MirandaX it really doesn't matter what values you insert in between each two points, so most DACs that oversample do indeed insert zeroes (or noise).

Right. Let's take just two sample points: one at the beginning of 70-minutes CD and another - at the end, the rest "really doesn't matter"; our DAC will fill in the rest by itself. You compressed 640M to 2 bytes - Nobel Prize committee is waitinig for you to accept the check

I'm afraid you didn't pay enough attention to your "most basic course in signal processing". Here's some more quoting Quote:

(http://www.benchmarkmedia.com/appnotes-d/hicks1994.htm): The first step in this process is to insert three new samples in between each of the original ones. The value is unimportant but zero is often used as it enables an efficient hardware implementation. < skip> Next, this faster sample stream is passed through a digital filter whose action is to make the new samples a smooth interpolation of the original data.

They put zeroes, that's true. But that's only an intermediate step!!

Anyway, I give up!..

Quote:

Originally posted by DarkAngel No need to get bogged down in all these technical discussions.

Yeah! Why bother?! If they tell you some egg-heads found a way to magically transform your CDs into SACDs - that must be true! Order toll-free! It works even better with that green marker they sold you last year

 

[Nezer]

Quote:

Originally posted by nec Right. Let's take just two sample points: one at the beginning of 70-minutes CD and another - at the end, the rest "really doesn't matter"; our DAC will fill in the rest by itself. You compressed 640M to 2 bytes - Nobel Prize committee is waitinig for you to accept the check

BYTES??!!?? Why waste that even, use a 1-bit DAC and now your down to two *bits*!! Of course, you can't even get a shave and a haircut for that these days let alone a nobel prize...

 

[MirandaX]

Quote:

Originally posted by nec Right. Let's take just two sample points: one at the beginning of 70-minutes CD and another - at the end, the rest "really doesn't matter"; our DAC will fill in the rest by itself.

You're misrepresenting the upsampling process. When we double the number of samples, we double the sample rate. This effectively doubles the Nyquist frequency (the highest frequency representable in the original data). The added zeroes do not affect the information content of the signal in the range [0,original Nyquist frequency].

Quote:

I'm afraid you didn't pay enough attention to your "most basic course in signal processing". Here's some more quoting They put zeroes, that's true. But that's only an intermediate step!!

Tell that to the prof who gave me a 4.0 in a graduate signal processing class last fall

The quote you cite confirms precisely what I'm talking about. The "interpolation" is a natural byproduct of the low-pass reconstruction filter applied after upsampling. The whole reason we upsample is to make the construction of such a filter easier. Believe me, creating good-performing FIR filters that can be implemented cheaply in hardware is nontrivial. Upsampling makes this possible.

 

[kelly]

Well, since I'm basically an idiot and don't understand the bulk of this stuff despite reading it, I'll offer up subjectivisim instead because any ******* can have an opinion.

Here's mine...

I think a good redbook CD on a good DAC can sound better than an SACD on an average player. So.. it's not necessarily a lie when someone says "SACD-like quality from your CD collection." The fact is that the SACD has more *potential* than the redbook CD format but if the potential is unrealized in the average player than the DAC could hypothetically equal or surpass the SACD.

Anyway, I do enjoy the engineering debate. I wish I could say I was understanding more of it. What I don't understand is that it seems like if you have a straight line between 10 and 15 that adding an extra datapoint of 12.5 wouldn't much matter (a straight line is a straight line no matter how many dots represent it). Therefore, it seems to me that all this talk of upsampling and such would only effect frequency extremes where data was previously undersampled. I'm not trying to correct you guys I'm just trying to show how I'm confused so one of you guys might take pity and dumb it down for me.