[GloP]

Ok, here is what I am looking for maybe someone can tell me if that even exists.


1) I put an audio CD in the CD player and press play in software X.


2) The HTPC reads the Audio CD using direct DAE (Digital Audio Extraction). The HTPC applies fancy error correction and interpolation algorithm (using or not the one provided by the hardware CD reader).


3) HTPC does a very nice upsample of the signal to 96Khz/24 bits (processing power to spare woohooo)


4) HTPC outputs digital signal using a quality SPDIF output


5) separate quality DACs (from my AV32R) turn the signal into analog and sends it on it's way


6) enjoy the beautiful analog signal


So is there **any** software / hardware combination that does all that by just doing (1) ?

 

[Sherbona]

"3) HTPC does a very nice upsample of the signal to 96Khz/24 bits "


Are you sure you want to upsample to that frequency and change the bits? CDs are 44.1Khz/16 bits, so that is not going to be a very 'clean' conversion... if you want to upsample, I think maybe you'd be better off going to an integer multiple of the original frequency, like 88.2khz; and also not dealing with dithering errors by keeping the original 16 bits.


Even then, keep in mind that simply doubling the samples from 44.1 klhz to 88.2 khz doesn't magically create information and won't sound as good as if the music were originally sampled at 88.2khz in the first place, but I guess the idea is that maybe the D to A conversion is better at the higher frequency...

 

[GloP]

You raise an interesting point, of course it doesnt create information but it "smoothe" the signal, the same way you can double the resolution of a picture, and it will definitly look better if you interpolate the new pixel instead of just doubling each pixel in each direction. Results for sounds have no reason to be less impressive than what you can get with a picture given enough processing power and resolution. Line doubler/tripler/... for video works exactly that way.


Given enough power I am sure the convertion can be "clean" enough, algotihm to do this kind of things exists and dont introduce dithering artifacts. Mathmatically dithering comes when you go the other way (trying to project a given information space on a smaller base). Everything is in the interpolation algorithm that should not introduce bias or artifacts. Granted it might be a calculation intensive task but todays CPU can deal easily with that amount of work. This technique also allow you to get rid of any jitter problem at read time since you read data asynchronously in a buffer, even 5 times in a row if you dont get consistant numerical reading from the CD drive.


It looks like softwares like EAC allow to play music by extracting the data from the CD audio drive directly but I am looking for an amateur, more hi-fi minded, solution. Some pretty well known very high end CD player (Meridian for example) work exactly that way, without even an incredible custom made transport. Why not use the same method to play high quality music from our HTPC? Dont anyone else think that would be possible with a HTPC ? And definitly way cool ? (and cheaper than one of these Meridian 800 ?)



(excuse my english if it's not perfect, it's not my native language)

 

[GloP]

Read this for error correcting capabilities of different drives:

http://www.exactaudiocopy.de/eac13.html


It's a very interesting read.

 

[thomaspf]

For exactly this application I use


M-audio DIO24/96


Winamp 2.9 + CDDA plugin + ASIO plugin


Configure the sample rate converter to whatever you like and enjoy. I was under the impression that the TAG MacLaren can do upsampling in the processor?


Cheers


Thomas

 

[GloP]

I found this URL for the CDDA plugin but I could not find anything for the ASIO plugin do you have a URL around?

 

[0db]

Interesting, but I was under the impression that decent DACs in the digital-analog conversion process already perform some sort of interpolation. I suppose it's possible that DACs only have the brute power for a linear interpolation and the HTPC may have the ability to better fit the data. Seems like an interesting application though - we've got excess power, let's make it work for us!

 

[Goi]

Hey GloP, I have the AV32R too, but interestingly when I use the Winamp ASIO upsampling plugin to upsample to 88.2KHz, the AV32R detects it at 96KHz instead.

 

[PGw]

I feed my AV32REX from M-Audio Audiophile 24/96 (digital) using WinAmp 2.8 with the CDReader plugin

and the ASIO plugin. It sounds great! (but I'm not using the upsampling mechanism.)

Winamp 2.x ASIO plugin

 

[KikeG]

Quote:

Originally posted by 0db Interesting, but I was under the impression that decent DACs in the digital-analog conversion process already perform some sort of interpolation.

You're right, virtually all DACs use oversampling, even cheapest ones.

Quote:

I suppose it's possible that DACs only have the brute power for a linear interpolation and the HTPC may have the ability to better fit the data.

Both good software resampling and DAC oversampling use the proper method, which is sinc convolution (brickwall filtering).


However, brute-force software resampling allows better filtering (allows a filtering closer to the ideal), but in practice, comparing with a decent DAC it will make a difference just on frequencies over 21 KHz or so.

 

[thomaspf]

That is interesting. If you start with a 16bit 44.1Khz signal and add one interpolated data point between 2 samples lets say with simple linear interpolation or with splines (Wadia).


Does the difference in the analog waveform only manifest itself in frequencies higher than 21Khz? How about real DACs vs. theoretical ones.


Cheers


Thomas

 

[KikeG]

thomas, I don't understand what you mean. Good real DACs don't use linear or spline interpolation, they use sinc interpolation, or what is the same, FIR brickwall filters. Same as SSRC, with the difference that SSRC uses a much "sharper" filter.


Despite the use of spline interpolation from Wadia converters and similar, spline interpolation is not the right way to oversample a signal. With spline interpolation the waveform just looks better in time domain, but results into improper filtering/reconstruction, if you analize the signal in frequency domain.

 

[GloP]

This is really interesting so we're saying that a computer can do better upsampling than the upsampling done inside a reasonable quality DAC? This kind of make sense to me but I wasn't sure, hence this thread. I am also very interested in the error correction/jitter removal process gained by reading the audio CD in the digital domain with a FIFO buffer and re-reading capabilities. I wonder why more hi-fi CD player dont use this technique. The Meridian player for example doesnt strike me as having anything beside high quality component that justify it's price and I am looking for reasons why the same quality would not be achievable with a properly programmed/configured HTPC. I am a software engineer and even though I lack experience in the DSP field I'd be willing to start a new project if I dont find anything matching my needs.

 

[thomaspf]

The point being that if you use any type of simple geometric interpolation the result changes the waveform all across the frequency band.


Intuitively this seems to be the a good thing to do to make a graphical representation of an analog wave in the time domain look more pleasing to the eye. That is what the marketing material from many upsamplers show...


Wadia even claims that their geometric smoothing sounds particularly good as a kind of effect processing.


Cheers



Thomas

 

[KikeG]

Quote:

Originally posted by GloP This is really interesting so we're saying that a computer can do better upsampling than the upsampling done inside a reasonable quality DAC? This kind of make sense to me but I wasn't sure, hence this thread.

I think so. Still, I don't see the point of doing software upsampling (or using "upsampling" DACs) when any decent DAC does a very good work of oversampling

Quote:

I am also very interested in the error correction/jitter removal process gained by reading the audio CD in the digital domain with a FIFO buffer and re-reading capabilities. I wonder why more hi-fi CD player dont use this technique.

Because even the cheapiest CD player already uses FIFOs when playing the data. The data is read from the disc, unpacked, decoded, error-corrected, and then, sent to the DAC at a fixed clock rate. The disc reading is speeded or slowed so that the internal FIFOs don't under or overrun. The clock is fixed, it is not recovered from anywhere, unlike the case of external DACs that have to recover the clock from the spdif stream.


About re-reading, cd players error correction is able to correct most read errors so that no bit played is wrong. This is because there is redundant information recorded at the disc. When the error is too big so that it can't be totally corrected, most players use simple intepolation to minimize audibility of the lost data. Still, this happens only when the disc is very scratched.

 

[thomaspf]

There is one core issue that high-end marketeers never seem to talk about that has not been mentioned in this thread. Consecutive bits on a redbook CD are not located in consecutive sectors on the disc but spread out to reduce the risk of loosing to much in a scratch.


Even for normal playback the player has to buffer everything to deinterleave the sectors and then apply the processing outlined by KikeG. The clock is therefore never directly coupled to the stream of bits read of the CD.


While high mass transports made a lot of sense for analog turntables they are more of a decorative feature similar to gold plating on a CD player.


Cheers



Thomas

 

[sly]

 MAD Plug-in for Winamp


I think this is kinda what you are looking for..


this is for mp3s but it does what you want..

it is a start.