Upsampling queries

1. If your original source file has a native rez of 16/44, does upsampling add a "fake" layer of resolution to make it 24/96, 24/192, DSD... or whatever?

Bluntly put - YES !

2. Will a file "upsampled" to say 24/192 sound the same as a native 24/192 file?

Since the upsampled file is a 'Fake' it will not sound as good as the original.

That said ... I think I remember reading a couple of years ago, that dCS first experimented with Upsampling, and were not quite sure if they should introduce it as a commercial product, since they were not quite sure how to provide a convincing argument as to how it works !

However, the demand for the products from audiophile in the East was so strong, that they decided to sell it anyway, even if they could not provide a theoretical explanation of how (why? ) it worked...

think of it as up-sampling plain DVD's to 1080P resolution. It is similar in concept.

Amp_Nut schrieb:

1. If your original source file has a native rez of 16/44, does upsampling add a "fake" layer of resolution to make it 24/96, 24/192, DSD... or whatever? Bluntly put - YES !

But then, what's the point?

Shahrukh schrieb:

With all the hoopla going on about computer audio and DACs, I too have developed a keen interest and am seriously considered going "PC" with my music. On my journey I have come across upsamplers and upsampling DACs. So here's my question: 1. If your original source file has a native rez of 16/44, does upsampling add a "fake" layer of resolution to make it 24/96, 24/192, DSD... or whatever? 2. Will a file "upsampled" to say 24/192 sound the same as a native 24/192 file? 3. What's oversampling? Gyaan awaited.

(1) Upsampling does not add a fake layer of resolution. One thing should be clear in your mind - upsampling (& oversampling) create NO NEW INFORMATION. Whatever information exists in the 16/44.1 format exists in the upsampled (overampled) data EXCEPT that the BIT RATE of the new data is 24/96, 24/192, DSD, etc, etc. No new information is created in the process. OK?
When you use the word "resolution", to me, you seem to be indicating that the upsampled data is an improvement over the original 16/44.1 data. From a purely digital data stream perspective the upsampled data is neither an improvement nor a deterioration over the original 16/44.1 data - it's merely different.
In audio, the listener can decide if he/she prefers the original 16/44.1 data or the upsampled version - it's highly subjective.
Your last post asked

But then, what's the point?

originally, the point of upsampling (oversampling) was to ease the design of the analog filter that proceeded the D/A converter in the CDP. You know that humans can (officially) hear upto 20KHz & that the redbook CD sampling rate is 44.1KHz. This makes the half sampling frequency 22.05KHz; only 2.05KHz higher than the upper human hearing limit. So, the analog filter proceeding the DAC (this is called the reconstruction filter) has to have a skirt/roll-off that is very steep - from 20KHz-22.05KHz it's got to provide some 90-100dB or more of attenuation to get the full 16-bit rez (Signal-Noise ratio) out of the CD music. That is a challenging design & with the design being analog (output of D/A converter is analog, of course!), the variations in the filter were likely to be too much. These variations would manifest themselves as change in filter bandwidth, change in steepness of the slope, etc - parameters that would affect CD playback performance. So, the engineers decided to upsample/oversample. They believed that if the bit rate was changed to 24/96 (just pulled one example from your list) then the half sampling frequency would be 48KHz. Now, the analog filter proceeding the D/A converter would have to provide the same high attenuation between 20KHz & 48KHz i.e. the filter gets more frequency range to achieve its attenuation. IOW, the filter skirt is shallower & the design is much eased. This also shows up in reduced cost.

(2) I believe Amp_Nut already answered this question. I agree w/ his take on this.

(3) Upsampling is a technique in digital signal processing to increase the bit rate of an input data stream while preserving the original information & without actually creating any new information in the process. It was created so that the input digital data stream clock rate could be changed as needed to be compatible with the clock rate of the next digital signal processing stage. Generally, in a complicated digital signal processing machine, there are several different clock rates which are dictated by the amount of data thru-put required, power dissipation allowed, spurious clock coupling into other nearby circuits, area occupied by the circuit in silicon, etc, etc.
It is hard to explain this concept anymore to someone w/o an engineering background 'cuz it requires an understanding of Nyquist's & the Shannon theorems ('cuz there are certain criteria for clock rate selection), understanding of signals & systems ('cuz upsampling/oversampling creates what is called side-bands & how one deals with them), etc.

IMHO, w/o getting overly technical the above explanations might suffice. Let me know if they do not & what other info you might want on this subject. Thanks.

The real reason for all this is because, modern DACs (which perform better than ancient 16 bit DAC), simply cannot take the original PCM (16 bit) format **directly** for processing.

Since the introduction of PCM in the 80's, the IC manufacturers have constantly improved thier technology, not willing to wait for music-publishers to upgrade the format in which Music is published (which in turn is because of money involved in license fee for all these formats !!)

It is a complex economic & power struggle.

The poor CD player manufactuers have to now resample the original format, so that the modern DACs will accept the signal and do the necessary function.

While this was a technical necessity...the companies added some extra mumbo-jumbo to make all this look like great inventions !!!. Unfortunately many do fall prey for all this.

I'm not saying the modern DACs are not better...but only that - we can do without the extra mumbo-jumbo.

Just crude way to look at reality...

sivat schrieb:

The real reason for all this is because, modern DACs (which perform better than ancient 16 bit DAC), simply cannot take the original PCM (16 bit) format **directly** for processing. Since the introduction of PCM in the 80's, the IC manufacturers have constantly improved thier technology, not willing to wait for music-publishers to upgrade the format in which Music is published (which in turn is because of money involved in license fee for all these formats !!) It is a complex economic & power struggle. The poor CD player manufactuers have to now resample the original format, so that the modern DACs will accept the signal and do the necessary function. While this was a technical necessity...the companies added some extra mumbo-jumbo to make all this look like great inventions !!!. Unfortunately many do fall prey for all this. I'm not saying the modern DACs are not better...but only that - we can do without the extra mumbo-jumbo. Just crude way to look at reality... :KR

Without meaning to get into argument with you, I'd like to say that I'm not in total agreement with you on this.

I'm not sure that increasing the clock rate of the DAC is an "improvement" in DAC technology. I will agree with you that modern day DACs do perform better than those in the 1980s & I feel that is because circuit techniques to design DACs & silicon processes to manuf them have both made great strides in the past 20-30 years.
Running a DAC at higher clock rates actually costs more - the DAC burns more current, you need wider bandwidth circuits because the clock rate is 96KHz or 192KHz, etc & the bandwidth of the circuits need be much higher than the clock rate to keep the DAC distortions very low. Thus, in turn, the DACs occupy more silicon area. Plus, the higher speed clock leaks more into the silicon substrate so if you have other circuits residing on the same silicon substrate, there is far more spurious coupling of the DAC clock into these other circuits. This adds distortion. So, all of this has a cascading effect on cost. Today, this has been largely offset by the steady decrease in cost of CMOS silicon processes such that one can put larger circuits for almost zero additional cost vs. the 1980s.
I'm still convinced that the priciple reason to increase the clock rate of the DAC was to ease the reconstruction filter design.

We are sooooo far down this upsampling/oversampling road now that it's hard to remember what the reason was for upsampling/oversampling. It now appears that the upsampling/oversampling are some major new inventions cooked up by these CD manuf companies & there is so much marketing hype (mumbo-jumbo in Sivat's words) around this concept so as to ensure a sale to the customer & to ensure that the customer buys into this concept & accepts it as a necessity.
Today, non-oversampling DACs have proven that this upsampling is not a requirement at all. So, now the listener can judge for him/herself which technique he/she prefers in his/her digital playback. We have a choice!

[quote="bombaywalla"]
I'm not sure that increasing the clock rate of the DAC is an "improvement" in DAC technology.
[/quote]

We are not in argument bcos this is not what i mean

I mentioned DAC's have improved.....how ?....that's another big story with many chapters. All that is not in Syllabus for this class. . For example, some of the recent improvements in dynamic ranges are shockingly good....

But what does all these improvement means to CDP designers....u need to patch up b/w old & new. As a bare minimum - reclocking/resampling is essential to put thru vanilla rebook via a modern DAC. As i mentioned already, most(all) modern DACs do not accept the Redbook I2S in native form...so there is really no other choice.

Different sampling rates, make the same DAC behave differently...and it can sound drastically different. This is really tricky issue with respect to algorithm...and mathematics. I've been really scratching head on these issues for a while now...

sivat schrieb:

bombaywalla schrieb: I'm not sure that increasing the clock rate of the DAC is an "improvement" in DAC technology. We are not in argument bcos this is not what i mean

OK, cool! Thank you for clarifying.

sivat schrieb:

I mentioned DAC's have improved.....how ?....that's another big story with many chapters. All that is not in Syllabus for this class. ;).

exactly correct!

sivat schrieb:

For example, some of the recent improvements in dynamic ranges are shockingly good....

completely agree with you.

sivat schrieb:

But what does all these improvement means to CDP designers...u need to patch up b/w old & new. As a bare minimum - reclocking/resampling is essential to put thru vanilla rebook via a modern DAC..

OK, how do I write this in a friendly manner so we can have an amiable discussion????

The improvements in today's modern DACs come from improved circuit design techniques & modern CMOS fabrication processes. These very same circuit design techniques can be utilized to produce a "shockingly good dynamic range" (quoting from your post) DAC running at 44.1KHz. There is nothing (& I mean really nothing) preventing this from happening. And, I believe that Burr-Brown & Philips (among many others have such products in the marketplace). In that case, you would not need to upsample/oversample/resample to get the DAC to play with the output data from the CD transport. Correct?

OK, so we've now done away w/ upsample/oversample/resample & we have a hi-perf modern-day DAC running at 44.1KHz.

But...........what problem do we create for ourselves with this modern-day hi-perf DAC running at 44.1KHz? IMHO, we have resurrected the problem of designing a very challenging reconstruction filter. The very issue we wanted to circumvent with the introduction of upsampling/oversampling/resampling.....

I hope that I'm making sense......

bombaywalla schrieb:

The improvements in today's modern DACs come from improved circuit design techniques & modern CMOS fabrication processes. These very same circuit design techniques can be utilized to produce a "shockingly good dynamic range" (quoting from your post) DAC running at 44.1KHz. There is nothing (& I mean really nothing) preventing this from happening. And, I believe that Burr-Brown & Philips (among many others have such products in the marketplace). In that case, you would not need to upsample/oversample/resample to get the DAC to play with the output data from the CD transport. Correct? OK, so we've now done away w/ upsample/oversample/resample & we have a hi-perf modern-day DAC running at 44.1KHz. But...........what problem do we create for ourselves with this modern-day hi-perf DAC running at 44.1KHz? IMHO, we have resurrected the problem of designing a very challenging reconstruction filter. The very issue we wanted to circumvent with the introduction of upsampling/oversampling/resampling..... I hope that I'm making sense......

In theory, it is true. But unfortunately, our friends at TI, Wolfonson, Analog Devices, Cirrus, etc., ..have different ideas ...

No modern burr-brown DAC can accept Vanilla I2S signal now. Either u need to use a filter like DF1704 or alternatively use on of the ASRC devices for resampling...

Vanilla = redbook = (16 bit CD Format)

sivat schrieb:

I mentioned DAC's have improved..... But what does all these improvement means to CDP designers....

The increase in DAC clock rate & the performance improvements in modern DACs are mostly/completely unrelated to each other. I think that we need to separate these 2 items.
If one can design a hi-perf DAC running at 96KHz or 192KHz then one can run that same DAC design (or design a similar hi-perf modern DAC running) at 44.1KHz. There is no limitation in doing this at all i.e. running a DAC a lower speed & extracting the same high performance out of it as another modern DAC running at 96KHz or 192KHz.

sivat schrieb:

Different sampling rates, make the same DAC behave differently...and it can sound drastically different. This is really tricky issue with respect to algorithm...and mathematics.

Absolutely correct! The algorithm for upsampling/oversampling is the entire crux of the matter. That's what makes or breaks the entire CD playback chain! And, this why you have the likes of Meridian, Wadia, EMM Labs, Theta, etc, etc, etc. All of them have different algorithms hence different sound. There's atleast 20 years worth of work in this area by various CD manuf!! Each one of these algorithms is a compromise & the listener gets to choose the least offensive one to his/her ears.

Very informative discussion !!

bombaywalla schrieb:

Each one of these algorithms is a compromise & the listener gets to choose the least offensive one to his/her ears.

I did not get this part, I can understand each one being different because the Engineer thinks that is the best but what compromises can be associated in a sampling algorithm .


Another thing Bombaywalla, it might sound very amateurish at this level but I always thought, theoretically, when the sampling rate is higher you are able to plot a more exact wave form of the signal and thats the reason it should ideally sound more accurate. But your explanation suggests there is nothing like that!

abhi.pani schrieb:

Very informative discussion !! bombaywalla schrieb: Each one of these algorithms is a compromise & the listener gets to choose the least offensive one to his/her ears. I did not get this part, I can understand each one being different because the Engineer thinks that is the best but what compromises can be associated in a sampling algorithm .

Abhi, you wrote "because the Engineer thinks that is the best".
I ask you, what is "that" referring to??
you answer this question & you'll find your answer!!

abhi.pani schrieb:

Another thing Bombaywalla, it might sound very amateurish at this level but I always thought, theoretically, when the sampling rate is higher you are able to plot a more exact wave form of the signal and thats the reason it should ideally sound more accurate. But your explanation suggests there is nothing like that!

correct! Now we are getting into the technical details of upsampling/oversampling & it can get a bit hairy.

Take a look at this pix:
http://upload.wikime...-Signal_Sampling.png
you can see the original signal in green & the blue dots which represent the digitization of this analog signal.

In upsampling, they do what is called "zero-stuffing". when they upsample the original data @ the new clock rate, they insert a logic 0 wherever they do not find the original data. So, you end up having a data stream that looks like this: orignal data, logic 0, logic 0....,orignal data, logic 0, .... (I'm inserting 2 logic 0s merely for argument's sake. You need to insert atleast 1 logic 0 but there is no theoretical limit to how many more you can insert).
So, imagine that on the time (t) axis (where the numbers 1, 2, 3.... are written) you have some blue dots. These blue dots will be a zero voltage level (the zero stuffing process).

You feed this to the DAC & you'll get garbage out!
In the above example, the 2 logic 0s between the 2 original datas did not exist in the original data stream off the CD. So, how does one convert these 2 logic 0s to some meaningful values such that there is a smooth transition between the 2 original datas??
Ahh-ha! that's where the various algorithms come into play. One engineer could say: let's use the least-squares error method, another could say: let's simply join the space between the 2 original datas by a ruler straight-line. A 3rd engineer might say: let's use a polynomial curve fit. And, so it goes on & on & on & on & on & on & .......
How many techniques to join to join 2 dots in space??? I would say that it's limited by your imagination, no??
Not all your ideas will sound good but many will.

Question: what is the theoretically correct way to join the space between the 2 original data points???
Answer: there is NONE!!! Why? the space between the 2 original data points never existed in the original data. YOU created it by upsampling!! So, now YOU come up with an algorithm to connect the space between the 2 original data points.
So, if YOU come up with an algorithm to the space between the 2 original data points, is your algorithm THE correct one??
If yes, why is my (or sivat's or anybody else's) algorithm to connect the space between the 2 original data points incorrect? Give a technical reason, please!

The thing is, you can't give a bonafide technical reason! There is NONE!
The algorithm is an interpretation of what it should be & it's totally subjective!!
Hence, it's a compromise.....

bombaywalla schrieb:

The increase in DAC clock rate & the performance improvements in modern DACs are mostly/completely unrelated to each other. I think that we need to separate these 2 items. If one can design a hi-perf DAC running at 96KHz or 192KHz then one can run that same DAC design (or design a similar hi-perf modern DAC running) at 44.1KHz. There is no limitation in doing this at all i.e. running a DAC a lower speed & extracting the same high performance out of it as another modern DAC running at 96KHz or 192KHz.

For the first part, key is to read my earlier post more carefully. The answer is there "old & new"...and note the words "as a bare minimum". These are important to understand what I'm talking about (Modern DACs are not designed with redbook in mind).

The second part is still theory. Even the ancient AD1865, does not accept the 16 bit/44.1 khz into it (it needs 18 bit at the minimum). The only DACs that can still operate without resampling are the likes of TDA1541A, TDA1545, etc.,

This is as much about "practical" limitation, as it is for theoritical possibilities. There is only so much you can learn from reading....hence the need to DIY

Maybe my english is bad...and folks might fint it difficult to understand.

Let me try to put this in another way...

The DAC designers from TI (Burr Brown), Analogue Devices, Wolfonson,etc., have long ago stopped treating Redbook as the center of thier universe.

Depending on the DAC, today there is a minimum sampling rate that is expected. The 16 bit/44.1Khz forumla falls short the minimum sample rate of allmost all DACs that were launched in the last decade or more...

Hence the only solution for CDP designers is to resample the redbook stream into higher resolution, so that the DAC will accept the signla.

This was the "technical necessity" that i referred to earlier. A bit of masala to this...and it becomes a big discovery by CDP designers

sivat schrieb:

For the first part, key is to read my earlier post more carefully. The answer is there "old & new"...and note the words "as a bare minimum". These are important to understand what I'm talking about (Modern DACs are not designed with redbook in mind). The second part is still theory. Even the ancient AD1865, does not accept the 16 bit/44.1 khz into it (it needs 18 bit at the minimum). The only DACs that can still operate without resampling are the likes of TDA1541A, TDA1545, etc.,

there seem to be a few more, for example, Philips TDA 1305, maybe even the UDA1334BTS & there is one from the old Burr-Brown PCM63K.

sivat schrieb:

This is as much about "practical" limitation, as it is for theoritical possibilities. There is only so much you can learn from reading....hence the need to DIY :prost

Go ahead & DIY all you want. Have fun, don't get lost. I stil do not think that DIY'ing will tell you why only 96KHz or 192KHz or any other freq DACs are available in the market. DIY'ing will lay in front of you the available DACs in the market. You still do not know why you have only these choices; only that you do. You can keep guessing & hypothesizing but that's all you are doing.
It's entirely possible (obviously I'm guessing why the TI/B-B, ADI, Wolfson, etc no longer make redbook their universe) that the redbook CD market was not a large enough market for the DAC vendors & that they needed to make DACs for other applications to make this biz worth their while. And, since you need to upsample to make the reconstruction filter design much easier (I've read this time & time again in several DAC data sheets), it's entirely possible that this aspect played right into the DAC vendors' hands of wanting to upsample the clock to make DACs for a overall larger market space. Very possible.

bombaywalla schrieb:

Go ahead & DIY all you want. Have fun, don't get lost. I stil do not think that DIY'ing will tell you why only 96KHz or 192KHz or any other freq DACs are available in the market. DIY'ing will lay in front of you the available DACs in the market. You still do not know why you have only these choices; only that you do.

To understand "why" modern DAC's do what they do...you just have to ready the white-papers they publish or various application notes that occampany thier products. It does not mean the manufacturers solve all problem, they do leave a bunch of math unanswered, making DIY all the more intresting :prost.

DIY is not just about hooking up wires and screwing up the cabinet box for some of us.

bombaywalla schrieb:

The improvements in today's modern DACs come from improved circuit design techniques & modern CMOS fabrication processes. These very same circuit design techniques can be utilized to produce a "shockingly good dynamic range" (quoting from your post) DAC running at 44.1KHz. There is nothing (& I mean really nothing) preventing this from happening. And, I believe that Burr-Brown & Philips (among many others have such products in the marketplace). In that case, you would not need to upsample/oversample/resample to get the DAC to play with the output data from the CD transport. Correct?

This was your earlier post .... which cleary showed lack of understanding ..as to why there is a "fundamental" need to resample today !!

However, your last message indicates that you knew this was wrong (you have read the datasheets & understood them !!)

All this while i was trying to clarify

..a bit confused

bombaywalla schrieb:

It's entirely possible (obviously I'm guessing why the TI/B-B, ADI, Wolfson, etc no longer make redbook their universe) that the redbook CD market was not a large enough market for the DAC vendors & that they needed to make DACs for other applications to make this biz worth their while. And, since you need to upsample to make the reconstruction filter design much easier (I've read this time & time again in several DAC data sheets), it's entirely possible that this aspect played right into the DAC vendors' hands of wanting to upsample the clock to make DACs for a overall larger market space. Very possible.

This is again how mumbo-jumbo's are created.

The DAC manufactuers do not want to support only redbook, but many other (advanced) formats as well. Redbook is still the biggest player in audio format...and for a long time to come and i'm sure the manufacturers understand that.

However for the real reason, we will take a analogy - It is like Windows 7, having to support application developed for "windows 95". I have already mentioned this fact in my first post...i'm trying to make it simple for a layman

sivat schrieb:

:? bombaywalla schrieb: The improvements in today's modern DACs come from improved circuit design techniques & modern CMOS fabrication processes. These very same circuit design techniques can be utilized to produce a "shockingly good dynamic range" (quoting from your post) DAC running at 44.1KHz. There is nothing (& I mean really nothing) preventing this from happening. And, I believe that Burr-Brown & Philips (among many others have such products in the marketplace). In that case, you would not need to upsample/oversample/resample to get the DAC to play with the output data from the CD transport. Correct? This was your earlier post .... which cleary showed lack of understanding ..as to why there is a "fundamental" need to resample today !! However, your last message indicates that you knew this was wrong (you have read the datasheets & understood them !!) All this while i was trying to clarify ..a bit confused

Not sure what your confusion is.....

sivat schrieb:

bombaywalla schrieb: It's entirely possible (obviously I'm guessing why the TI/B-B, ADI, Wolfson, etc no longer make redbook their universe) that the redbook CD market was not a large enough market for the DAC vendors & that they needed to make DACs for other applications to make this biz worth their while. And, since you need to upsample to make the reconstruction filter design much easier (I've read this time & time again in several DAC data sheets), it's entirely possible that this aspect played right into the DAC vendors' hands of wanting to upsample the clock to make DACs for a overall larger market space. Very possible. This is again how mumbo-jumbo's are created. The DAC manufactuers do not want to support only redbook, but many other (advanced) formats as well. Redbook is still the biggest player in audio format...and for a long time to come and i'm sure the manufacturers understand that. However for the real reason, we will take a analogy - It is like Windows 7, having to support application developed for "windows 95". I have already mentioned this fact in my first post...i'm trying to make it simple for a layman :)

you seem to be talking as tho you are an authority on the subject!
Between you & me I do not know who the layman is - I design ADCs, DACs & other mixed-signal ICs for a living. So, I know what it takes to get such circuits from concept to production & I make exactly the design choices I'm speaking about. You, OTOH, seem to be good at using ICs available in the market & implementing for your various DIY projects. So, you tell me who's the layman here??

This leaves me every more confused

Upsampling does not add a fake layer of resolution. One thing should be clear in your mind - upsampling (& oversampling) create NO NEW INFORMATION. Whatever information exists in the 16/44.1 format exists in the upsampled (overampled) data EXCEPT that the BIT RATE of the new data is 24/96, 24/192, DSD, etc, etc. No new information is created in the process. OK?

Ok!

When you use the word "resolution", to me, you seem to be indicating that the upsampled data is an improvement over the original 16/44.1 data. From a purely digital data stream perspective the upsampled data is neither an improvement nor a deterioration over the original 16/44.1 data - it's merely different. In audio, the listener can decide if he/she prefers the original 16/44.1 data or the upsampled version - it's highly subjective.

So, are you saying that 24/192 file is NOT superior to a 16/44 - just different? Then why this insanity to go "hi-rez"? Why upsamplers?

Sivat and Bombaywalla



a lot of

but finally, it's

Shahrukh schrieb:

So, are you saying that 24/192 file is NOT superior to a 16/44 - just different? Then why this insanity to go "hi-rez"? Why upsamplers?

I believe that Amp_Nut already addressed this. What I'm saying that 16/44.1 upsampled to 24/192 is different sonically to 16/44.1. Many people like the upsampled 16/44.1 sound. There are almost an equal number of people that do not like the upsampled 16/44.1 sound.

What the madness is about "hi-rez"? IMHO there are 2 aspects to this, (1) is that people are buying/downloading NATIVE 24/96 & 24/192 music i.e. music that is initially recorded in 24/96, 24/192 right at the get-go. NO upsampling & (2) there are some websites (one such is HDTT) that are offering music files of old(er) music that have been upsampled to 24/96, 24/192.
It's a personal opinion whether or not hi-rez music sounds very good or not - I've seen an equal number of believers & non-believers. Music done correctly in 16/44.1 can sound really excellent & it'll surprise you just how good it can sound! But, try hi-rez music for yourself & see if you subscribe to this "insanity" (as you call it).

For those people who do not buy (or do not want to buy) hi-rez music, they can pretend that they are listening to hi-rez music by upsampling their 16/44.1 music. As I explained before in a post addressed to Abhi, there are upsamplers & then there are upsamplers. You'll like some/many & there are many others you will not like. So, it's very likely that you'll have to try a lot of upsampling units before you settle on one.
Hope that this makes sense....

"One good thing about music, when it hits you feel no pain" - Damian Marley

Shahrukh, I like you quote.
What I've found is that if your system can deliver music with its emotions then, when music hits you some of that feeling can be pain depending upon the music content & your overall mood.
If Damian Marley was not feeling any pain when the music hit him, his system was probably not delivering the emotion attached to the music during playback!

FWIW, here's a very short historical perspective of the CD player & its associated DAC:
* as many of you already know, the CD player was introduced in the EU in 1982 & in the USA in 1983. Back then, the state-of-the-art (SOTA) was 12-14 bit DACs. So, yes, the 1st generation of CDPs were sold with mostly 14-b DACs & these DACs were clocked at the 44.1KHz rate. One could not even get full CD performance out of these CD players BUT even a 14-b DAC had 86dB of signal-to-noise compared to 60dB signal-to-noise for analog (vinyl & reel-reel). So, their 1st objective to unseat analog was supposedly successful. Also, note, that since the DACs were clocked at 44.1KHz, the reconstruction filter following the DAC was a steep cut-off filter. These filters, tho' manufactured, were very expensive. Needless to say, since the CD player was a brand new technology, the CDP carried a very high price tag. One of the internal electronics contributing to the high price tag was these steep cut-off filters.
* the 2nd generation of CDPs were introduced in the later 1980s (1988, 1989 time-frame). According to me there were 2 noteworthy features to pay attention to - (1)by this time the various DAC manuf (Analog Devices, Burr-Brown, etc) had come a long way & were introducing 16-b & 18-b DACs. So, now one could get full CD performance & (2) the DACs used were running at 2X, 4X & even 8X sampling frequencies! One such DAC was the Analog Devices AD1856 introduced in 1988. This DAC could oversample upto 8X, which is 352.8KHz.
Back in the late 1980s, neither DVD-A (which is 24/96) nor DSD audio (which runs at 2+ MHz) nor hi-rez FLAC at 24/192 existed. Yet, the DACs used in the 2nd generation CDPs were already oversampling DACs. Why??? The reason was that the CDP manuf wanted to bring down the price tag & one of the areas to address was the reduction in cost of manuf the analog reconstruction filter following the DAC. Having a steep cut-off reconstruction filter was not the solution to a cheaper CDP. This is where upsampling came to the CDP manuf's rescue (as I already explained in my earlier post #5) - higher the DAC sampling freq, higher the Fs/2 frequency, shallower the cut-off slope of the reconstruction filter & cheaper the cost to manuf in mass.
it was not due to

The real reason for all this is because, modern DACs (which perform better than ancient 16 bit DAC), simply cannot take the original PCM (16 bit) format **directly** for processing.

nor was it due to

It is a complex economic & power struggle.

& nor was it due to

But what does all these improvement means to CDP designers....u need to patch up b/w old & new. As a bare minimum - reclocking/resampling is essential to put thru vanilla rebook via a modern DAC.

.
In the words of that author, the above quotes are "mumbo-jumbo" reasons.

Today, when 24/96, 24/192 & DSD audio exist, the CD quality DACs have already been upsampling to 96, 192 for several generations of products hence it's no skin off the DAC manufs to support hi-rez audio formats.

Bombaywalla, I think I now understand your point.
Just a question, Why is the constraint of F/2 for the cut off ? I mean if the sampling freq is 44.1khz, why dont we have the liberty to cut off just before 44.1khz, why is it 44.1/2 ?

So the DAC today takes in 44.1/16 bit input ?

It is true the DAC manufacturers innovated. That is the point.

The CDP designer, who did not have anything do with it..have been touting around oversampling ...and making mumbo-jumbo...that's the point.

bombaywalla schrieb:

it was not due to The real reason for all this is because, modern DACs (which perform better than ancient 16 bit DAC), simply cannot take the original PCM (16 bit) format **directly** for processing. nor was it due to It is a complex economic & power struggle. & nor was it due to But what does all these improvement means to CDP designers....u need to patch up b/w old & new. As a bare minimum - reclocking/resampling is essential to put thru vanilla rebook via a modern DAC. . In the words of that author, the above quotes are "mumbo-jumbo" reasons. formats.

For 1) I will await response from the author who will indicate a modern DAC that will take PCB directly as input

For 2) The author has clearly indicated Economy as the reason why DAC's evolved...and in time it also became a power issue. Else why will Sony buy so many Music business and prevent us from having a better format...the DAC manufactuers just did not wait for Sony to loose its grip.

For 3) Same as (1). I will await info on "One" DAC that will accept incoming PCM in Vanilla format..

abhi.pani schrieb:

Bombaywalla, I think I now understand your point.

Cool, Abhi.

abhi.pani schrieb:

Just a question, Why is the constraint of F/2 for the cut off ? I mean if the sampling freq is 44.1khz, why dont we have the liberty to cut off just before 44.1khz, why is it 44.1/2 ?

To understand this, you need to understand the Nyquist theorem which states that to accurately represent an analog signal in the digital domain one needs to sample it at a freq that is 2X the highest expected freq in the analog signal. So, for audio, the highest freq is 20KHz hence we will need to sample the analog signal at no lower than 40KHz. For redbook CD we actually sample at 44.1KHz (this freq came from video raster scanning from the Philips labs during the creation of the CD standard) hence the maximum frequency that we can accurately capture in the analog signal is 22.05KHz (44.1KHz/2). If we try to capture a signal greater than 22.05KHz, we get something called "aliasing" because we violate the Nyquist theorem. These alias (imposter) frequencies manifest themselves in the 20Hz-20KHz audio band & create distortion (a bad thing). In sampling theory, the Fs/2 (sampling freq divided by 2) is called the fold-over frequency because all the unwanted frequencies in the 22.05KHz --> 44.1KHz fold over into the wanted audio band of 20Hz --> 20KHz creating distortion.
Hope that this answers your question.

I got it...thanks.

bombaywalla schrieb:

I believe that Amp_Nut already addressed this. What I'm saying that 16/44.1 upsampled to 24/192 is different sonically to 16/44.1. Many people like the upsampled 16/44.1 sound. There are almost an equal number of people that do not like the upsampled 16/44.1 sound. What the madness is about "hi-rez"? IMHO there are 2 aspects to this, (1) is that people are buying/downloading NATIVE 24/96 & 24/192 music i.e. music that is initially recorded in 24/96, 24/192 right at the get-go. NO upsampling & (2) there are some websites (one such is HDTT) that are offering music files of old(er) music that have been upsampled to 24/96, 24/192. It's a personal opinion whether or not hi-rez music sounds very good or not - I've seen an equal number of believers & non-believers. Music done correctly in 16/44.1 can sound really excellent & it'll surprise you just how good it can sound! But, try hi-rez music for yourself & see if you subscribe to this "insanity" (as you call it). For those people who do not buy (or do not want to buy) hi-rez music, they can pretend that they are listening to hi-rez music by upsampling their 16/44.1 music. As I explained before in a post addressed to Abhi, there are upsamplers & then there are upsamplers. You'll like some/many & there are many others you will not like. So, it's very likely that you'll have to try a lot of upsampling units before you settle on one. Hope that this makes sense....

It does make sense. Thanks.

I think , you people can start making CD players.. Seriously.