SACD ripping using your PS3 (part 2)

[Synfreak]

Hi Ted!

 

Yeah, thats how I see it too.

With Saracon there is absolute no "real" benefit using higher samplingrates as i.e. 88,2khz.

The steep anti-imaging filter cuts of anything way below that theretical maximum, so ...

(and even Bruce seem to had recognized it ... ;-) )

 

@lich000king:

If you want to save some space, and therefor using "only 48khz SR, I would at least suggest to save in 24Bit. I would think it's worth it!

 

For the most cases 88,2/24 should be enough, but that's up to the individual ...

 

Cheers

Harald

 

P.S.:

"Avalon" is GREAT!

But one of the "hotter" SACDs, so I had to reduce the level by 3dB to avoid clipping!

 

[ted_b]

rip DSD, then convert it to simply 16/48k? It is worthy (theoretically, assuming good source of course) of at least 24/176k, and 24/352k is called DXD for a reason.

 

[lich000king]

I'm not interested in DSD, I'm interested in surround music.

Also, I am not aware of any audible benefit of sampling rate/bitdepth above 48/16.

 

[ted_b]

n/t

 

[Kevin_O]

See this then...

 

 

http://www.temporalcoherence.nl/docs/HighReso.pdf

 

[Miska]

In DSD there is only noise well above 20kHz anyway, or did I missunderstand something?

 

Here are some spectrograms of frequency sweep from two of my converters. That faint harmonic is coming from the encoder.

 

 

 

 

[Kevin_O]

Yeah...keep believing that there is no such thing as 2nd and 3rd harmonics, overtones and tape bias noise in DSD.

 

[Miska]

http://www.temporalcoherence.nl/docs/HighReso.pdf

 

Regarding the time domain resolution discussed in this doc, I'd just comment that with careful design it is possible to make PCM converge with DSD at 352.8 kHz sampling rate.

 

Yeah...keep believing that there is no such thing as 2nd and 3rd harmonics, overtones and tape bias noise in DSD.

 

Would you mind clarifying what this is referring to and what you mean? I didn't quite understand to which posting this was related to...

 

 

[audiozorro]

I just skimmed the article and saw nothing obvious that I disagree with. But I am puzzled with your reference to tape bias noise in DSD.

 

I use a Korg MR-2000S to record to DSD128. Where does the tape bias noise enter the process?

 

[Kevin_O]

Seems everyone wants to get rid of DSD noise, then why isn't everyone concerned with tape bias noise as well?

 

If you get rid of DSD noise, then that removes all the harmonics/overtones as well.

Might as well be listening to CD's since DSD noise starts at 22k.

 

We use a Korg MR2000s as well as a Sonoma and Pyramix workstations. When I convert DSD to PCM, I don't use any filters. That's how HDtracks got into trouble with the Rolling Stones downloads. They used Saracon which puts at steep filter in and the 88.2 and 176.4 files look exactly the same.

Leave the filters out!

 

[Miska]

Seems everyone wants to get rid of DSD noise, then why isn't everyone concerned with tape bias noise as well? If you get rid of DSD noise, then that removes all the harmonics/overtones as well.

 

I'm not concerned about tape bias noise if it's correctly dealt with. Something to complain about the figures I posted?

 

And both contain filtering that I consider correct. Latter one is for those especially concerned about the noise.

 

When I convert DSD to PCM, I don't use any filters.

 

So you are using 2.8 MHz PCM?

 

Your own algorithm or by some 3rd party software?

 

 

[wappinghigh]

Harald wrote

 

"P.S.:

"Avalon" is GREAT!

But one of the "hotter" SACDs, so I had to reduce the level by 3dB to avoid clipping!"

 

No shit Harald. The surround SACD of Avalon is THE sacd......

 

Prices of that thing on ebay prove it

I'm lucky to own a copy...I call it "My Precious" LOL

It's pretty much tho only physical disc I play these days...

 

It's THE benchmark

The pinnacle of a great artform

The top of Mt Everest

 

The Music industry has been sliding down the slippery slope ever since...

 

Wap

 

 

 

[wappinghigh]

Whilst not in anyway attempting to rain on the parade, or pass judgement on what is right or wrong, may I simply point out to all posters on this thread to be very careful....

 

...."Intelligence agencies keep things secret because they often violate the rule of law or of good behavior." Julian Assange

 

I don't wish to cast dispersions on the music industry, but may I remind everyone in the past, that with the help of the FBI, they have not hesitated to try and make life miserable for even the most insignificant and well meaning of college students: http://news.dmusic.com/article/16172

 

And I'd be very surprised if this site wasn't on their "watch list".

 

 

 

[audiozorro]

is not insignificant and well meaning.

 

The Computer Audiophile website is against music piracy and illegal file sharing.

 

[terminus]

as the vast majority of my own video and audio content is paid for and/or downloaded under licence, but I wouldn't believe for a moment the industry's figure of $50m. They are notorious for cooking up figures like this. If we assume retail value of 99c per song and $15 per movie, that's either 50 million songs or 3.3 million movies, or some combination thereof... which frankly, even if Dhaliwal was an inveterate collector of pirated content, beggers belief. I don't doubt that they are talking about the amount of money that they think they could recover at trial inclusive of the maximum statutory damages award, which would be $25,000 per item. In that light, it scales back to 2,000 songs or movies... probably less than an iPod's worth. Sure, it's still copyright infringement, but I'm just saying... don't take what the industry says at face value.

 

[audiozorro]

"They are notorious for cooking up figures like this."

 

"don't take what the industry says at face value."

 

This is not a winning strategy in court unless your strategy is to intimidate defendants and avoid court altogether.

 

[Manelus]

Is there a way to decompress DST to DSD?

Weiss saracon (the one I use to convert to PCM), great quality BTW, does not read DST, just DSD.

 

[lich000king]

The foobar plugin ca do this, though it takes a lot of time.

But why don't you let your PS3 do the job? You can rip as DSDIFF DSD directly, this will be much faster.

 

[Manelus]

Unfortunately i don't have a ps3, so I must deal with ISO files.

maybe I can get one in a near future, but not now.

foobar DSD to PCM conversions are poor quality. Weiss Saracon algorithm does way better job with high frequencies.

 

[Sik_Lescinovid]

I've heard allusions in this thread of DSD being equivalent (without filtering) to 2.8 MHz PCM or theoretically equivalent to 24bit/176.4kHz PCM or a variation of the above. Neither is true, DSD is in fact equivalent to a resolution slightly below 24bit/96kHz PCM (see http://www.extremetech.com/computing/48844-surround-sound for a nice overview or simply do the math yourself comparing good literature on PCM and the technical specification of DSD). Hence I don't understand why anyone would want to convert a DSD signal to a resolution higher than 24bit/96kHz...then suddenly Saracon's behavior isn't so odd, is it?

 

EDIT: Just in case someone doesn't know how to go about doing the "math": 1*2822.4 (the bit rate of DSD) is slightly bigger than 24 * 96 (the bit rate of PCM), hence one would first assume that DSD is slightly better than 24bit/96kHz PCM in terms of resolution. The way that DSD works though (by storing quantization noise in the upper frequency register) compared to how PCM works implies with a bit more thought that DSD is actually comparable to something like 20bit/96kHz PCM and hence of slightly lesser resolution (in terms of actual music) than 24bit/96kHz PCM.

 

[audiozorro]

http://www.theaudioarchive.com/TAA_Resources_File_Size.htm

 

Bitrates for 2 channel audio:

 

16/44.1 = 1411.2 kbps

24/88.2 = 4233.6 kbps

24/96 = 4608.0 kbps

DSD64 = 5644.8 kbps

24/176.4 = 8467.0 kbps

24/192 = 9216.0 kbps

DSD128 = 11289.6 kbps

DXD = 16934.4 kbps

 

When given a choice, I want 24/176.4 or higher digital. And when I do my own recordings it is to 24/192 or DSD128.

 

 

[Miska]

...for a nice overview or simply do the math yourself

 

That article is incomplete and gives wrong picture. I've posted bunch of figures already on other threads.

 

For reminder, here's frequency sweep converted to 176.4/24 PCM with standard filtering applied:

http://www.computeraudiophile.com/files/dsd64-sweep_0.jpg

 

Hence I don't understand why anyone would want to convert a DSD signal to a resolution higher than 24bit/96kHz...

 

Because you don't want to spoil the frequency and time domain properties of DSD.

 

The way that DSD works though (by storing quantization noise in the upper frequency register) compared to how PCM works implies with a bit more thought that DSD is actually comparable to something like 20bit/96kHz PCM and hence of slightly lesser resolution (in terms of actual music) than 24bit/96kHz PCM.

 

It is not quite that simple. Define resolution? Frequency domain? Dynamic range? Transient response? Low level linearity?

 

Depending on modulator, DSD can give ~160 dB dynamic range at low/mid- frequencies. While also having roughly 100 kHz total audio bandwidth without time-domain side effects of brickwall filtering...

 

As an after note, please also remember, that all modern DACs convert PCM to SDM for the conversion stage...

 

 

[Sik_Lescinovid]

That article is incomplete and gives wrong picture. I've posted bunch of figures already on other threads.

 

Could you be a bit more specific as to why this article is incomplete and why it gives the wrong picture? Also if you don't particularly like that article there's a plethora of literature available on the net coming to the same conclusion...

 

For reminder, here's frequency sweep converted to 176.4/24 PCM with standard filtering applied:

http://www.computeraudiophile.com/files/dsd64-sweep_0.jpg

 

So if I understand correctly the argument here is that apparently without applying a hard lowpass filter there is genuinely some of the original signal visible in the upper frequency range where theoretically only quantization noise should be stored and hence one should not apply that hard of a lowpass filter and use higher sample rates when converting to PCM? That's an interesting suggestion, although it kind of implies that DSD isn't capable of encoding data higher than the one found on 24/96 PCM in the sense that it will mix with quantization noise. I.e. it's up to the listener to either see DSD as a container capable of handling such high frequencies with added noise or simply not being able. I will have to do some listening tests to see what I prefer here...

 

Because you don't want to spoil the frequency and time domain properties of DSD.

 

My question was quite obviously asking for why one would spoil the "frequency and time domain properties" (whatever it is you are actually referring to) if one limits oneself to 24/96 PCM...

 

It is not quite that simple. Define resolution? Frequency domain? Dynamic range? Transient response? Low level linearity?

 

And this is not helpful at all. Additionally I was using only one of the terms you are asking me to define and hence I can only interpret this as a nonsensical straw-man. If you want me to define "resolution" though I'd liken it to bit rate restricted to actual music signal (i.e. not bit rate used for encoding the quantization noise). The basic idea here is that superficially DSD (not DSD128 nor DXD) has a slightly higher bit rate than 24/96 PCM, but if one compares the strength of lowpass filters applied by SACD players and the frequency range usually taken up by quantization noise one can quite easily actually see that in DSD the bit rate for pure music data (i.e. without added noise) is slightly lower than the bit rate of 24/96 PCM, it just happens to be so even under ideal conditions. Unless of course you're counting music data which has been mixed with noise in which case it seems to be a matter of taste again...though I'm pretty sure the engineers who made DSD have never intended to use the noise domain as storage for musical data, but that's just my opinion of course.

 

Depending on modulator, DSD can give ~160 dB dynamic range at low/mid- frequencies. While also having roughly 100 kHz total audio bandwidth without time-domain side effects of brickwall filtering...

 

Here I just want to mention that all of the literature I've read on DSD usually claims that even under ideal conditions a maximum of about 120db dynamic range can be achieved. Do you have any sources for your claim? I'd be genuinely interested.

 

As an after note, please also remember, that all modern DACs convert PCM to SDM for the conversion stage...

 

Could you be a bit more specific as to what this has to do with what I've said? I can't seem to find a viable connection and any help would be appreciated.

 

[Miska]

Could you be a bit more specific as to why this article is incomplete and why it gives the wrong picture?

 

Because it doesn't contain enough detail to actually draw a complete picture. It's a sketch.

 

So if I understand correctly the argument here...

 

Yes and no, you seem to misunderstand how DSD is supposed to work and be used (no unusual though). Just as with PCM, quantization noise is everywhere. It's just not constant across frequencies. Just like in case of PCM I use, I have much more dynamic range in < 10 kHz frequencies than >20 kHz frequencies. On purpose.

 

but if one compares the strength of lowpass filters applied by SACD players and the frequency range usually taken up by quantization noise one can quite easily actually see that in DSD the bit rate for pure music data (i.e. without added noise) is slightly lower than the bit rate of 24/96 PCM

 

Wrong, in the figure I posted filtering was done according to Scarlet Book SACD standard.

 

And don't look at those multi-format players that use Mediatek's 88.2 kHz DSD to PCM conversion. There are players out there that do the filtering right.

 

My question was quite obviously asking for why one would spoil the "frequency and time domain properties" (whatever it is you are actually referring to) if one limits oneself to 24/96 PCM...

 

It would limit the available bandwidth in frequency domain while reducing time domain resolution by slowing down transients (due to frequency domain limitation) and introducing ringing (due to brickwall filtering).

 

Here I just want to mention that all of the literature I've read on DSD usually claims that even under ideal conditions a maximum of about 120db dynamic range can be achieved. Do you have any sources for your claim?

 

Just look at performance of some of the better modern DSD modulators.

 

Could you be a bit more specific as to what this has to do with what I've said? I can't seem to find a viable connection and any help would be appreciated.

 

DSD is just marketing name for one bit SDM. All modern ADCs and DACs are SDMs or hybrids. DSD is native language of modern ADCs and DACs, unlike PCM. That's one of the reasons DSD was invented, to avoid SDM->PCM conversion in ADC and PCM->SDM conversion again in DAC. That's why the first D comes from "Direct".

 

Output of modern ADC and DAC have similar noise slope. On PCM ADC it is just brickwall'ed away on PCM conversion and on DAC it is filtered out by analog filters inside and outside the chip.

 

 

[Synfreak]

@ Miska:

 

It might be a bit OT, but ...

Have you considered getting a Mytek 192DSD DAC?

That _should_ work with ASIO to stream DSD data to the DAC-chip (ESS9016?), and with your player that would be the (currently at least) chepest solution to pöay native DSD on a Windoze PC.

 

Cheers

Harald

 

P.S.:

I have the mytek on order, should materialize in about 3 to 4 weeks.