192 KHz/24bit DAC - Is it true?

+A -A
Autor
Beitrag
vpriyan
Ist häufiger hier
#1 erstellt: 22. Nov 2008, 11:53
Hi

I am able to see many CD Players, DACs with this tag "192 KHz/24 bit" upsampling DAC. How is that possible if the CD Recording is just 44.1KHz/16bit?

Does this upsampling concept work?

Regards
Visak
toob_noob
Schaut ab und zu mal vorbei
#2 erstellt: 26. Nov 2008, 06:01
I am keen to get a proper understanding of this as well.
square_wave
Inventar
#3 erstellt: 26. Nov 2008, 10:14
Here is a simple explanation.

http://www.soundstag...gtechnical200311.htm
bombaywalla
Stammgast
#4 erstellt: 26. Nov 2008, 17:27

vpriyan schrieb:
Hi

I am able to see many CD Players, DACs with this tag "192 KHz/24 bit" upsampling DAC. How is that possible if the CD Recording is just 44.1KHz/16bit?

Does this upsampling concept work?

Regards
Visak



This *IS* a loaded question & would require a pretty long explanation as one would have to get into DSP fundamentals, which might not be interesting to many others.
Anyway some short answers:
* the 192KHz/24-bit" tag is largely a marketing gimmick to distinguish one CDP from another. The idea to communicated to the user is that more bits at a higher rate directly equates to better sonic performance.
* this is *not* necessarily the case.
* yes, the concept of upsampling or oversampling is being employed here.
* yes, the concept works & it works like a charm ELECTRONICALLY. whether that xlates to better sound is dependent on the listener's ear.
* the central reason for even emploiying up/oversampling was to give the analog filter after the DAC (called a reconstruction filter) a break in meetings specs. As you know, after the DAC the signal is analog so the filter after the DAC has to be analog. In comparison to digital, analog circuits vary "all over the map" in terms of voltage & current variations. If no up/oversampling would be used, then the transition of the filter from pass-band to stop-band would have been very steep - human sound limit is 20KHz & the half sampling rate is 44.1KHz/2 = 22.05KHz. That would mean that the filter would have to transistion from pass-band to stop-band in just 22.05-20KHz = 2.05KHz! Holy cow! that would have meant a very steep filter skirt. That would in turn mean a very high order filter, which would have xlated to an expensive analog filter.
(you'll have to accept the usage of half sampling rate to calc the filter transition region unless you have studied DSP in school).
So, the Philips engineers conjured up up/oversampling - now if the up/oversampling rate is, say, 192KHz then the half sampling rate is 192KHz/2 = 96KHz. *NOW* the filter transition region is 96KHz - 20KHz = 76KHz. Whew! NOW we have lots of freq range between pass-band & stop-band. Thus, the order of the filter can be dramatically reduced, which xlates to designing a sloppier analog filter. This keeps the amount of analog design to the minimum, keeps cost down AND now the manuf also has marketing tool to pull wool over the consumer's eyes (higher sampling rate is better sonically!)
* several people think that up/oversampling is the pitts & that it creates artifacts in the re-produced sound. These people are favour non-oversampling DACs.
* several other people love up/oversampling & they think that it's the best thing since sliced bread!
* several others are on the fence on this topic - they like the non-os DACs when they hear a good one & they can be swayed to like up/oversampling DACs when they hear a good one.
* both up/oversampling DACs & non-os DACs have very good commercial implementations. Names will be withheld as they are personal opinions & also to protect the "innocent" (they are hardly innocent but.....)


Using the up/oversampling concept it is easy to change the bit rate of the music data. Also, by up/oversampling at 1 integer rate & downsampling at a second unrelated rate one can even achieve a net fractional bit rate change. That's exactly how they convert from 44.1KHz to, say, 48KHz. If you calc the 48/44.1 ratio you get 1.088. How the heck are you going to change the rate by 1.088???
you can achieve that by up/oversampling by 160 times & then downsampling by 147 times.

Hopefully this explanation will shed some light & help some.
Amp_Nut
Inventar
#5 erstellt: 03. Dez 2008, 11:05


I am able to see many CD Players, DACs with this tag "192 KHz/24 bit" upsampling DAC. How is that possible if the CD Recording is just 44.1KHz/16bit?


I have been meaning to write on this thread, but have SOOO Much to say, that I never got started !

Bombaywalla has been kind enough to put time aside and post... most of the points, in a nut shell.

Many of these deserve to be further elaborated.

As time permits, I will try to post, on this topic, but would also like to learn from others, who better know the theory, or have audiotioned upsampling CD players. Most will agree that it DOES improve sound..

Heck, maybe even talk about Oversampling on this thread ?


However, to start with your question... "How is that possible if the CD Recording is just 44.1KHz/16bit?"

I wonder if you are familiar with PhotoShop ?

Try this experiment :

a. Take a small pized pic, and increase its size by 200%, in a progarm like word, by stretching the pic from one of its corners.

b. Take that same pic, and "Resize" it by 200% inh Photoshop.

You will be surprised. Photoshop's resize will provide MUCH better visual results.

How ?

Photoshop "Interpolates" or add extra dots in the exopanded empty spaces. These Extra dots place are of an average value of adjacent dots.

This provides a smoother continious picture, than just creating a larger picture by expanding the dots and the inter-dot spaces, ( as done by Word )

Similarly, you can take a CD Recording at 44.1KHz/16bits and interpolate extra info between the original bits...

This is ofcourse just a loose example....


[Beitrag von Amp_Nut am 03. Dez 2008, 11:07 bearbeitet]
abhi.pani
Inventar
#6 erstellt: 04. Dez 2008, 07:58
Hi Amp_Nut,
While your Photoshop example was simple and practical but wouldnt this also say that you are adding your own bits to the original picture and ultimately fiddling with the originality. While it magnifies objects which were too small in the original picture to be viewed at a glance and projects them out, isnt it in the process presenting the whole picture in a manner different than the original. If the same picture would have actually been taken in a hi-rez format (originally big) would it look the same as this blown-up/cooked-up picture ??


[Beitrag von abhi.pani am 04. Dez 2008, 08:00 bearbeitet]
Amp_Nut
Inventar
#7 erstellt: 04. Dez 2008, 08:21
Hi Abhi,

Ofcourse what you say is true.

The Data on the CD and the CD format are fixed, at a level tghat was mass producaable 10 or 20 years ago.

The aim now is to make that data sound as good as possible.

As you have rightly pointed out, itr would never sound as good as a True Hi Res recording eg 24 Bit / 192 KHz.

Incidentally, 24/192 has 250 TIMES as much data (bits) compared to a standard CD.
Arj
Inventar
#8 erstellt: 04. Dez 2008, 09:12
Hmm nice example. is it possible that some of the higher end CDPs do the upsampling and then a downsampling to make the sound from the cd better in terms of "details? in anyway ?
Amp_Nut
Inventar
#9 erstellt: 04. Dez 2008, 09:35
The Digital data is like points forming a smooth curve - the final analog signal to output from the CD player.

The more dots, the closer the dots will represent the final waveform.

The Final Digital-To-Analog (D-to-A ) process, simply joins all these dots ( no matter how many or how few the dots ) into a continious curve, called the analog signal.

So we dont need to reduce the number of dots, before joining them. ....

In fact, the more the merrier as long as the dots are not out of line

( these Out Of Line Dots can loosely be said to be similar to Aliasing... I hope to post on that this evening .... )
Amp_Nut
Inventar
#10 erstellt: 04. Dez 2008, 13:33
To better understand Oversampling, we need to review some basics on digital audio :

1. Nyquist in 1928 Harry Nyquist of AT&T reckoned that to properly digitise any frequency, the sampling frequency must be at least twice as high as the highest frequency to be digitised.

Since the CD was to deliver 'perfect quality' over the audio spectrum of 20 Hz to 20 KHz, a sampling frequency of at least 20 KHz x 2 = 40 KHz was required. For practical purposes (to accommodate a filter slope) a higher frequency was selected... finalised at 44.1 KHz.

The reason why 44.1 KHz instead of 44 KHz, is explained here :

http://www.cs.columbia.edu/~hgs/audio/44.1.html

ALIASING :

If a digital system sampling at 44.1 KHz is fed a signal higher than 22 KHz, some really weird and horrible results are obtained !

As an example if the sampling frequency is 44 KHz and an Analog signal of 30 KHz enters the system, the digital system will create 'Alias' or 'Ghost' frequencies, within the audio band !

The analog frequency (30KHz) exceeds the Half sampling frequency (22 KHz) by 8 KHz.

It creates Alias (Ghost) frequencies at 22 KHz - 8 KHz = 14 KHz.

(The frequency will be folded back across half the sampling frequency by the amount it exceed half the sampling frequency)

Note that no 14 KHz frequency was fed to the digital system ! Only a 30 KHz signal was fed. However a 14 KHz alias or Ghost appeared !



BRICK WALL FILTER

The only way to avoid creating Alias or Ghost frequencies is to ensure that the digital system is never fed any frequency above 22 KHz.

Common, analog Low Pass filters gradually reduce the signal fed to them at 6 dB or 12 dB or even 24 dB per octave. These are not just good enough for digital use because if they are fed a 30 KHz signal it would be reduced by about 24 dB and fed to the digital system. The digital system would then create Alias or Ghost frequency, approx. at -24 dB and at a frequency of 14 kHz. That would represent about 10% distortion ! Since its not a multiple ( Harmonic ) of the original signal, it will sound Horrible !

Digital filters have been designed that will create a VERY VERY sharp cut off at 70 dB to or 90 dB per Octave. Their slope is almost vertical and their response appears like a brick wall through which high frequencies cannot pass.

These brick wall filters provide excellent attenuation but badly mess up the phase and timing of signals that they attenuate or of signals close to their cut-off frequency. These filters have often been condemned by some audiophiles.



OVER SAMPLING

If instead of a 44.1 KHz sampling frequency, the frequency is increased to 192 KHz. Hence it is only necessary to filter out frequencies above 192/2 = 96 KHz. Hence audio signals of even 96 KHz and below (compared to 22 KHz for the CD standard) will not create Ghost or Alias frequency.

That horrible 'Brick Wall' digital filter can be done away with or made to operate at 96 KHz.. a frequency at which there is very little audio information, and the digital filter can do no harm.

P.S: Wonder it this explanation has got tooo involved ?

bombaywalla
Stammgast
#11 erstellt: 04. Dez 2008, 15:16

Amp_Nut schrieb:

P.S: Wonder it this explanation has got tooo involved ?

:prost


I like it Amp_Nut! well done explaining the concepts.
looks like you have studied DSP in school or picked it up real well along the way.
For the non-technical this might be too much to swallow.
It's hard to explain DSP concepts in any other language/method.



A good visual for the concept of folding frequency is as follows:-
* draw a horizontal line on a napkin/paper.
* divide this horizontal line into 2 EQUAL halves - make a vertical mark at the mid-point of the horizontal line.
* this is the half sampling freq - in Amp_Nut's example, this mid-point represents 22KHz. The right-most point represents 44KHz.
* physically fold the paper/napkin at the mid-point mark.
* now you can visually see that any frequency content in the righ half (to the right of the mid-point) FOLDS into the left half area (to the left of the mid-point).
* this concept of any signal higher than 1/2 the sampling frequency is called ALIASING.




why 1/2 the sampling frequency?
Because Nyquist (of AT & T Bell Labs) showed that if an analog signal is digitized & needs to be re-constructed back into analog, one needs a minimum of 2 points of that analog waveform.
That's why one needs to sample at 2X the highest frequency that one is likely to see coming in.
It follows that 1/2 the sampling freq is the highest freq signal that can be allowed into the system WITHOUT aliasing.
Any input freq higher than 1/2 the sampling rate will become an imposter/alias & fold into the audio band & rain on our parade (i.e. spoil the sonics).



Amp_Nut,
since you seem to have a good understanding of DSP, the only peeve I seem to have w/ sampling higher than 96KHz is settling time for the analog circuits.
If the input to the analog ckts is being updated every 192KHz & the analog circuit needs to behave in a deterministic manner then the bandwidth of the analog circuits needs to be ~10X higher than the update/clock rate i.e. one needs 1.92MHz of closed-loop bandwidth!
I really do not think that any audio related analog circuit has a bandwidth that high.
What this tells me that the analog circuits cannot settle in-time before the next update hits its inputs. This is bound to create signal-dependent distortion, which is B-A-D!
Amp_Nut
Inventar
#12 erstellt: 04. Dez 2008, 16:06


I like it Amp_Nut! well done explaining the concepts.
looks like you have studied DSP in school or picked it up real well along the way.


Thank you Sir !

I am rather long in the tooth... when I finished my Engeneering, the CD player was not invented.
Suche:
Das könnte Dich auch interessieren:
IS IT TRUE THAT 1 WATT = 2.83 volt
SUB_BOSS am 25.01.2005  –  Letzte Antwort am 25.01.2005  –  3 Beiträge
DAC
neono am 02.09.2005  –  Letzte Antwort am 05.09.2005  –  2 Beiträge
Rethm the second + Unison research 6 + Trivista 192 tube DAC = Magic
square_wave am 11.06.2007  –  Letzte Antwort am 11.06.2007  –  3 Beiträge
GOing balanced.is it worth it?
Savyasaachi am 03.01.2009  –  Letzte Antwort am 10.01.2009  –  26 Beiträge
NAD @ soundnvision - is it recomended?
vpriyan am 22.11.2005  –  Letzte Antwort am 29.11.2005  –  5 Beiträge
Connector.Is it important?
vpriyan am 16.01.2007  –  Letzte Antwort am 17.01.2007  –  12 Beiträge
Is Jamo E-series worth it?
Neutral am 21.04.2005  –  Letzte Antwort am 26.04.2005  –  21 Beiträge
Best Enclosure Structure.... is it sphere?
Sonic_Master am 28.11.2005  –  Letzte Antwort am 29.11.2005  –  6 Beiträge
Yamaha AX497 amp - how is it?
Bibs am 03.02.2006  –  Letzte Antwort am 08.02.2006  –  4 Beiträge
Is the power ratings on AVR - 100 into 6 channels true.??
SUB_BOSS am 12.01.2005  –  Letzte Antwort am 12.01.2005  –  4 Beiträge
Foren Archiv
2008

Anzeige

Aktuelle Aktion

Partner Widget schließen

  • beyerdynamic Logo
  • DALI Logo
  • SAMSUNG Logo
  • TCL Logo

Forumsstatistik Widget schließen

  • Registrierte Mitglieder925.513 ( Heute: 3 )
  • Neuestes Mitgliedemoryrascon
  • Gesamtzahl an Themen1.550.306
  • Gesamtzahl an Beiträgen21.521.551

Hersteller in diesem Thread Widget schließen