IPB

Welcome Guest ( Log In | Register )

2 Pages V   1 2 >  
Reply to this topicStart new topic
DTS 1536kbps, how transparent it manages to be?, Is there any ABX test that's been conducted?
gorman
post Sep 16 2008, 11:12
Post #1





Group: Members
Posts: 66
Joined: 19-March 05
Member No.: 20746



Well, title and description say it all.

I hear a great deal of noise on the new lossless multichannel codecs and I wonder... considering perceptual transparency is achieved in all major lossy codecs at around 192kbps (or even lower if you don't have golden ears), what could be said about full bitrate DTS?
Go to the top of the page
+Quote Post
memomai
post Sep 16 2008, 12:57
Post #2





Group: Members
Posts: 264
Joined: 13-February 05
From: Germany, Kempten
Member No.: 19808



nothing. Making ABX tests at that bitrate range is useless. It'll be transparent for sure.


--------------------
FB2K,APE&LAME
Go to the top of the page
+Quote Post
2Bdecided
post Sep 16 2008, 14:07
Post #3


ReplayGain developer


Group: Developer
Posts: 5108
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



I wouldn't be so sure. Basic DTS is a very simple algorithm. It must have its weak points. Whether they are audible at 1536kbps, I don't know - but it's not "high" compared with 192kbps if the 192kbps in question is for two channels with a state-of-the-art codec - while the 1536kbps is for 6 channels with old technology.

Cheers,
David.
Go to the top of the page
+Quote Post
SebastianG
post Sep 16 2008, 14:35
Post #4





Group: Developer
Posts: 1318
Joined: 20-March 04
From: Göttingen (DE)
Member No.: 12875



QUOTE (2Bdecided @ Sep 16 2008, 15:07) *
I wouldn't be so sure. Basic DTS is a very simple algorithm. It must have its weak points.

I actually downloaded the DTS specification from ETSI a couple of years ago (it was free of charge!). From what I remember it's like MP2 + optional linear prediction within subbands.

The Wikipedia article also contains some more hints like
  • optional entropy coding of quantized samples (which is lacking in MP2)
  • optional use of "joint frequency coding" (whatever that may be exactly)

Divide 1536 kbps by 6 full-band channels and you'll get 256 kbps per channel. That seems quite okay for something that's probably a bit better than MP2.

Cheers,
SG
Go to the top of the page
+Quote Post
GeSomeone
post Sep 16 2008, 14:58
Post #5





Group: Members
Posts: 921
Joined: 22-October 01
From: the Netherlands
Member No.: 335



Some time ago I tried a compare between DTS-CD (44.1kHz/24 bit input) and lossless (48kHz/24b), both 5.1 .
It was not a (double) blind AB/X test, so it's just my impression. With (certain) Classical music I found the two very hard to distinguish, Rock music with a whole lot going on in all channels seemed to make some more of a difference.

Such behaviour may be expected from a Constant Bit Rate codec, good as long as it his enough bits to spend.

P.S. it is also possible to make DTS 2.0 or 4.0 so it does not necessarily have to be 6 channels, although that's the common format. (At full bit rate 2.0 would make no sense anyway.)
Go to the top of the page
+Quote Post
[JAZ]
post Sep 16 2008, 18:50
Post #6





Group: Members
Posts: 1778
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



QUOTE (SebastianG @ Sep 16 2008, 15:35) *
optional use of "joint frequency coding" (whatever that may be exactly)

http://en.wikipedia.org/wiki/Joint_Stereo#...quency_encoding

In other words, MP2's intensity Stereo. tongue.gif

QUOTE (GeSomeone @ Sep 16 2008, 15:58) *
Some time ago I tried a compare between DTS-CD (44.1kHz/24 bit input) and lossless (48kHz/24b), both 5.1


Could it be that you were hearing resampling differences? (44khz vs 48khz) (on playback, that is)
Go to the top of the page
+Quote Post
knutinh
post Sep 17 2008, 13:34
Post #7





Group: Members
Posts: 569
Joined: 1-November 06
Member No.: 37047



QUOTE
' date='Sep 16 2008, 19:50' post='588668']
QUOTE (SebastianG @ Sep 16 2008, 15:35) *

optional use of "joint frequency coding" (whatever that may be exactly)

http://en.wikipedia.org/wiki/Joint_Stereo#...quency_encoding

In other words, MP2's intensity Stereo. tongue.gif

QUOTE (GeSomeone @ Sep 16 2008, 15:58) *
Some time ago I tried a compare between DTS-CD (44.1kHz/24 bit input) and lossless (48kHz/24b), both 5.1


Could it be that you were hearing resampling differences? (44khz vs 48khz) (on playback, that is)

I would have guessed that even an old 5.1ch lossy format was able to exploit inter-channel redundancy so that the needed bitrate needed for a given level of perseptual degradation for "5.1" material would be less than 5 times that needed for the same degradation of mono material?

-k
Go to the top of the page
+Quote Post
me7
post Oct 15 2008, 14:48
Post #8





Group: Members
Posts: 177
Joined: 23-August 06
Member No.: 34375



What about AC3? I've read several claims that AC3 is more efficient than DTS and produces similar sound quality at lower bitrates, but only on not-reliable forums. ABXing is useless since AC3 and DTS tracks are usually mastered differently.
Does anyone here know some facts about the specification of AC3 and how it performs compared to DTS?
Go to the top of the page
+Quote Post
smok3
post Oct 15 2008, 17:28
Post #9


A/V Moderator


Group: Moderator
Posts: 1728
Joined: 30-April 02
From: Slovenia
Member No.: 1922



if you need to author a classic SD DVD, then you have to use dolby as well anyhow.


--------------------
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung
Go to the top of the page
+Quote Post
me7
post Oct 15 2008, 19:02
Post #10





Group: Members
Posts: 177
Joined: 23-August 06
Member No.: 34375



No, I mean to backup my favourite movies on an external harddrive and I'M unsure which audio format to keep.
Go to the top of the page
+Quote Post
user
post Jan 14 2009, 12:49
Post #11





Group: Members
Posts: 873
Joined: 12-October 01
From: the great wide open
Member No.: 277



5.1 ac3/DD can be max. bitrate 448 kbit/s, "standard" bitrate on some movies with 5.1 ac3 is even only 384 kbit/s.

I got recently a DVD-V with music+ some video, it had optional a 640 kbit/s or so ac3 stream, addtionally to the 448.
That was advertised as great bonus feature for HiFi etc, btter sound quality, but they did not warrant, that this stream plays on all ac3/DD systems, as it is already out of ac3 standard.

I recall, it played on mine.

But 5.1 DTS is another thing.

Bitrate is nearly unlimited, consider the commercial CD-Audio , 16 bit stereo, 44,1 kHz PCM,
it offers 1411 kbit/s, and 5.1 DTS can be packed into this wav container.

How many of the 1411 kbit/s could be used for DTS data ?

I guess at least 2/3 , 70% ?

That would be a bitrate of 1000 kbit/s for dts on CD-Audio, as the .1 bass channel needs less bitrate as the 5 channels, about 180 - 200 kbit/s per single channel could be possible !
So it should be enough for good quality.
That is doubled bitrate compared to 448 kbit/s ac3.

IMo, dts sounds better, nearly perfect, than ac3 at 448. of course, that could be due to mastering of movie sound also.


--------------------
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials
Go to the top of the page
+Quote Post
GeSomeone
post Jan 14 2009, 18:24
Post #12





Group: Members
Posts: 921
Joined: 22-October 01
From: the Netherlands
Member No.: 335



QUOTE (user @ Jan 14 2009, 13:49) *
But 5.1 DTS is another thing.

Bitrate is nearly unlimited, consider the commercial CD-Audio , 16 bit stereo, 44,1 kHz PCM,
it offers 1411 kbit/s, and 5.1 DTS can be packed into this wav container.

How many of the 1411 kbit/s could be used for DTS data ?

I guess at least 2/3 , 70% ?

There is no need to guess. DTS-CD is a format that has been around for years now (even before DVD).
The bitrate for the datastream is 1235 kb/s as the 2 most significant bits (of the 16 bit CD data) are not used. The reason is so that the static, you hear on non-DTS capable equipment, will not be full scale (but -12dB).

Of course the bit rate is not unlimited, in fact it is CBR only, either 768, 1536 (48kHz on DVD) or 1411 kb/s (44.1 kHz on CD) (not considering the newer DTS-HD on BluRay). The original topic is about DTS at 1536 bit rate anyway.
Go to the top of the page
+Quote Post
Ron Jones
post Jan 14 2009, 19:07
Post #13





Group: Members
Posts: 412
Joined: 9-August 07
From: Los Angeles
Member No.: 46048



There have been two that I'm aware of. One was performed by AES, which requires you to be a member to view the results, while the other was performed by the EBU, and that's freely downloadable as a PDF. I'll find a link a bit later, but they show that DTS at 1.5mbps fares very well on the MUSHRA scale, between 4.4 and 4.9 on a number of samples. It also shows that Dolby Digital at 448kbps fares equally as well.
Go to the top of the page
+Quote Post
Nil Einne
post Feb 27 2009, 08:04
Post #14





Group: Members
Posts: 10
Joined: 27-February 09
Member No.: 67440



Any luck finding the studies you refer to?
Go to the top of the page
+Quote Post
sauvage78
post Feb 27 2009, 10:16
Post #15





Group: Members
Posts: 677
Joined: 4-May 08
Member No.: 53282



some weeks ago I have been playing shortly with AC3 & DTS to see how transparent both can be compared to mp3/aac/vorbis

for ac3 I tested with the killer sample Castanet, which I know by heart (I tested every modern lossy codecs on it)
Sonic Foundry Soft Encode (Dolby V6.6.2) Soft Encode 1.0, Castanet 2.0 Wav>AC3, 100% easyly ABXable up to 224Kbps. (maybe ABXable at 256Kbps but it requires attention, I gave up as quality was OK)
ac3enc (Used by BeSweet) (from ffmpeg) V1.20 (18-02-04), Castanet 2.0 Wav>AC3, 100% easyly ABXable even at 256Kbps. (Very Bad, I mean awfull)
I wanted to test Aften but the download link for the binary was broken.

my personnal conclusion is that ac3 is a bad transform codec, clearly worse than non-lame mp3. (something like fhg mp3 or even worse IMHO...)
On music, with transform codecs killer samples, it is not transparent at the bitrate used both on DVD (5.1 448Kbps /5x2= 2.0 180Kbps) & HD-DVD (5.1 504Kbps /5x2= 2.0 201Kbps)
But it is most likely transparent at the bitrate used by BD (5.1 640Kbps /5x2= 2.0 256Kbps) specially if professionnal hardware encoder are better than those I tested.
Quality decrease VERY quickly with AC3.

I tried to test DTS but it was a pain, I couldn't find any encoder doing 2.0 at equivalent bitrate to 5.1 (DTS 5.1 768Kbps /5x2= 2.0 307Kbps, DTS 5.1 1536Kbps /5x2= 2.0 614Kbps)
so I tried to create an artificial 5.1 file from castanet 2.0 by splitting channels with audacity then duplicating them & then I encoded with SurCode DVDPro DTS Encoder V1.0.21
the result wasn't scientific AT ALL as I compared a lossless 2.0 file against a lossy 5.1 file (with fake channels) downmixed to 2.0 ... anyway both at 768 & 1536 the result was MUCH louder (without speaking about audio artefacts).
my personnal conclusion is that I cannot draw any conclusion except:
DTS encoders are very unfriendly (no 2.0, no scale of bitrate, need splitted channels, change the loudness), so anyone claiming DTS superiority most likely didn't ABX it ... anything you will find is spontaneous newbie opinions like "the louder the better" or "the bigger the better".
Due the loudness, it was almost impossible to ABX artefacts without applying replaygain to DTS, as my DTS file was in an MKV I couldn't replaygain it so I gave up ... I should have decoded to 5.1 wav recoded to flac then replaygain it & ABX ... but at this point I was bored & my file was a weird fake 5.1 anyway.
the rumors that AC3 & DTS streams would be encoded from differents masters is IMHO most likely a myth due to this huge loudness boost ... (this may happen but you can never know as it will sound different (louder) even with the same master)

Despite the loudness I am confident that DTS 1536Kbps is more than acceptable quality as the bitrate is overkill even for 2.0 ... it's 614Kbps.

the biggest problem IMHO is to know weither DTS 5.1 768Kbps is better or worse than AC3 5.1 640Kbps
I have no clue. I would need several native 5.1 killer samples that affect both subband & transform codecs ... which means IMHO that it's so hard to test DTS that no one knows the real quality of DTS.

Quote SebastianG:
"Divide 1536 kbps by 6 full-band channels and you'll get 256 kbps per channel."
==> you must divide by 5, as when you encode with nero aac or vorbis 1.2 at 128Kbps for 2.0 you end up with slighly above 320Kbps for 5.1 (64Kbps per channel) I think LFE is not a full channel.

gorman:
DTS 1536kbps is not lossless, DTS-HD ~3200Kbps /5x2= 2.0 1280Kbps for 24Bits is lossless. maybe at 16Bits you may end with a lossless bitrate near 1536kbps (I have seen 1400Kbps with TrueHD 5.1 16bit) but it's not the same. With AC3 & DTS you will always have a lowpass near 18Khz, not with TrueHD or DTS-HD.

GoSomeone:
What software did you use to get 2.0 ? I tried 3 ... none of them was doing 2.0 ... I read somewhere that it was possible but I don't know with which software ... I am interested even if 768/1536Kbps cannot be compared to between 2.0 & 5.1 as it's not the same bitrate by channel. Anyway DTS is so unfriendly that I suspect that even with a software that outputs 2.0, it cannot output 307/614Kbps which would be needed to draw a conclusion on 2.0 castanet 44.1Khz & extrapolate on 5.1, 768/1536Kbps, 48Khz.

Edit:
I just tested aften ac3 V0.0.8, I was able to instantly ABX castanet at up to 320Kbps 2.0, 384Kbps may be ABXable but I stopped as quality was becoming OK (By OK I mean that I couldn't instantly ABX it, not that it was good). This is a 5.1 bitrate for a 2.0 sample !
The quality of aften at 384Kbps is very roughtly equivalent to the quality of Sonic Foundry AC3 at 256Kbps. (would need carefull ABX to be more precise but at such a low efficiency I don't care for this codec) So, actually professional AC3 encoder (Sonic Foundry) seems much better than free ones on this short test. People using aften & ac3enc are deaf wink.gif
The quality of aften at 320Kbps is very roughtly equivalent (maybe worse, but not better) to the quality of vorbis/aac at 96Kbps on castanet ... none is transparent. (Edit: I was generous with Aften here)

This post has been edited by sauvage78: Mar 1 2009, 12:07


--------------------
CDImage+CUE
Secure [Low/C2/AR(2)]
Flac -4
Go to the top of the page
+Quote Post
DARcode
post Feb 27 2009, 10:59
Post #16





Group: Members (Donating)
Posts: 681
Joined: 10-January 05
From: Italy
Member No.: 18968



QUOTE (sauvage78 @ Feb 27 2009, 10:16) *
[...]clearly worse than non-lame mp3. (something like fhg mp3 or even worse IMHO...)[...]
Pretty groundless comparison in light of the recent audio listening tests.


--------------------
WavPack 4.70.0 -b384hx6cmv/qaac 2.41 -V 100
Go to the top of the page
+Quote Post
probedb
post Feb 27 2009, 11:17
Post #17





Group: Members
Posts: 1214
Joined: 6-September 04
Member No.: 16817



640Kbps DD is standard, it's the bitrate of the core DD tracks within Dolby TrueHD as far as I'm aware? It's mentioned in Wikipedia's DD article http://en.wikipedia.org/wiki/Dolby_Digital
Go to the top of the page
+Quote Post
sauvage78
post Feb 27 2009, 11:28
Post #18





Group: Members
Posts: 677
Joined: 4-May 08
Member No.: 53282



Darcode:
I meant that AC3 as the same kind of artefacts at high bitrate (192-224Kbps) as mp3 at low bitrate (64-96Kbps) ... not mp3 at near transparent bitrate indeed ... all mp3 encoders, even lame, are very bad on castanet at low bitrate, so it wasn't a criticism of fhg ... it sounds like a cavern with a river in it if you see what I mean ...

This post has been edited by sauvage78: Feb 27 2009, 11:31


--------------------
CDImage+CUE
Secure [Low/C2/AR(2)]
Flac -4
Go to the top of the page
+Quote Post
Kujibo
post Feb 28 2009, 00:59
Post #19





Group: Members
Posts: 38
Joined: 4-January 08
Member No.: 50127



QUOTE (sauvage78 @ Feb 27 2009, 01:16) *
I tried to test DTS but it was a pain, I couldn't find any encoder doing 2.0 at equivalent bitrate to 5.1 (DTS 5.1 768Kbps /5x2= 2.0 307Kbps, DTS 5.1 1536Kbps /5x2= 2.0 614Kbps)
so I tried to create an artificial 5.1 file from castanet 2.0 by splitting channels with audacity then duplicating them & then I encoded with SurCode DVDPro DTS Encoder V1.0.21
the result wasn't scientific AT ALL as I compared a lossless 2.0 file against a lossy 5.1 file (with fake channels) downmixed to 2.0 ... anyway both at 768 & 1536 the result was MUCH louder (without speaking about audio artefacts).
my personnal conclusion is that I cannot draw any conclusion except:
DTS encoders are very unfriendly (no 2.0, no scale of bitrate, need splitted channels, change the loudness), so anyone claiming DTS superiority most likely didn't ABX it ... anything you will find is spontaneous newbie opinions like "the louder the better" or "the bigger the better".
Due the loudness, it was almost impossible to ABX artefacts without applying replaygain to DTS, as my DTS file was in an MKV I couldn't replaygain it so I gave up ... I should have decoded to 5.1 wav recoded to flac then replaygain it & ABX ... but at this point I was bored & my file was a weird fake 5.1 anyway.
the rumors that AC3 & DTS streams would be encoded from differents masters is IMHO most likely a myth due to this huge loudness boost ... (this may happen but you can never know as it will sound different (louder) even with the same master)


I just wanted to defend DTS a bit here. It seems like you might be confusing a couple of non-DTS issues with the DTS format itself.

1) You copied the stereo image into the remaining channels then downmixed it and complained that DTS causes a loudness issue? Downmixing correlated content is always going to result in a louder 2.0 file. Did you downmix PCM and listen to it before you went to DTS? A better test would be to simply take the front left and right channels out of the decoded 5.1 DTS mix instead of downmixing.

2) It comes across like you are saying DTS isn't a flexible format, but I think the issue you are having is with your DTS encoding software, not the format itself. DTS is capable of more than just a 5.1 channel config. I might hope there is some software out there that can accept more flexible inputs.

Cheers.
Go to the top of the page
+Quote Post
sauvage78
post Feb 28 2009, 02:16
Post #20





Group: Members
Posts: 677
Joined: 4-May 08
Member No.: 53282



Kujibo:
Thks for the information, I didn't realize I was introducing the loudness boost by downmixing & I didn't had the idea of removing fake channels after encoding. It sounds clever, maybe I'll try later.
As I said the test wasn't valid to my own mind anyway, that's why I gave up.
I didn't really attack the DTS format but clearly the encoding software I tested. I don't know much about the format itself & I don't care as I will transcode it to vorbis anyway.
I was just trying to understand what was the original quality of BD streams to determine which is the best to keep & what setting to use for transcoding ... I still favor DTS 1536Kbps over AC3 640Kbps ... but I don't know what to think of DTS 768Kbps. In the end the original quality/flaw of both DTS & AC3 didn't matter when I determined my vorbis setting ... I used the lowpass of the source encoder & the transparency of the target encoder as my criteria ... DTS 1536Kbps/AC3 640Kbps/Vorbis 320KBps 5.1 all use something near 18Khz & are all near transparent. (EDIT: AC3 640Kbps is not transparent see next post)

My tests are not finished for DTS, I also gave up as, at time, I expected to grab a more friendly DTS encoder. I know for sure that AC3 is a bad codec now, but I still don't know the real quality of DTS ... I was just pointing that testing DTS is not easy compared to other codecs, so one should be very carefull when he listens to the opinion of people concerning DTS. There is lot of FUD concerning DTS.

This post has been edited by sauvage78: Mar 1 2009, 07:46


--------------------
CDImage+CUE
Secure [Low/C2/AR(2)]
Flac -4
Go to the top of the page
+Quote Post
C.R.Helmrich
post Feb 28 2009, 15:01
Post #21





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (Nil Einne @ Feb 27 2009, 08:04) *
Any luck finding the studies you refer to?


http://tech.ebu.ch/webdav/site/tech/shared/tech/tech3324.pdf

See page 21 ff. if you want to skip the technicalities. Apparently, at least 448kbps are needed to code any 5.1-channel signal transparently. According to this test, DTS at 448kbps is not transparent for all items, but DTS at 1500, Dolby Digital Plus at 448, and WMA 9 at 448 are. AAC at 320 is quite close.

Edit: By "transparent", I mean a score greater than 80 on the MUSHRA scale, but since that's not everyone's opionion, I should have used "excellent" I guess.

This post has been edited by C.R.Helmrich: Feb 28 2009, 15:08


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
Nil Einne
post Feb 28 2009, 18:29
Post #22





Group: Members
Posts: 10
Joined: 27-February 09
Member No.: 67440



QUOTE (C.R.Helmrich @ Mar 1 2009, 03:01) *
http://tech.ebu.ch/webdav/site/tech/shared/tech/tech3324.pdf

See page 21 ff. if you want to skip the technicalities. Apparently, at least 448kbps are needed to code any 5.1-channel signal transparently. According to this test, DTS at 448kbps is not transparent for all items, but DTS at 1500, Dolby Digital Plus at 448, and WMA 9 at 448 are. AAC at 320 is quite close.

Edit: By "transparent", I mean a score greater than 80 on the MUSHRA scale, but since that's not everyone's opionion, I should have used "excellent" I guess.

Thanks greatly for the study, rather interesting stuff.

In particular the DTS bit proves to me that DTS Connect (i.e. transcoding content from a PC to DTS 1536 for output over SPDIF) can be considered good enough as there's no evidence for any benefit of LPCM or other lossless for 5.1 channel 48/16 output. Given that there's little evidence for any benefit of 192/24 it really seems to me there's little benefit at all to the new HDMI standards from an audio quality viewpoint. While that doesn't surprise me, it's good to actually have some hard evidence.

However I'd really like to see some DD (i.e. AC-3) 640 stuff. Unfortunately the study only tested DD up to 448 where it's not transparent. As far as I'm aware Dolby Digital Live, the competitor of sorts to DTS Connect outputs DD (not Plus) at the maximum bitrate 640 kbps. Since some mobos have one but not the other, it's interesting to see if it really matters which one we choose (beyond hardware support). I'm guessing it will be transparent, or at least awfully close (since 448 is already quite good) but it would be nice to actually see some confirmation as with DTS. Wikipedia says "Then the DTS audio track is encoded at its highest legal bitrate (1,536 kbit/s), technical experts rank DTS as perceptually transparent for most audio program material (i.e., indistinguishable to the uncoded source in a double blind test.) Dolby claims its competing AC-3 codec achieves similar transparency at its highest coded bitrate (640 kbit/s)" but it's unsourced and in any case, who cares about what Dolby claims? :-P

We can't of course rule out the possibility of crappy codec implementations in the drivers/chipset meaning the output of either DTS Connect or DD Live is not what we would expect. But since these are licensed techs and the encoding is hardly difficult for modern hardware I'm guessing not.

P.S. I would largely agree with your definition of transparency. Perhaps better, as I expressed above, there's no evidence for a statistically significant difference between DTS et al and the source audio from the available studies.
P.P.S. Of course this isn't really the post to discuss DD 640 but since it has been discussed :-P

This post has been edited by Nil Einne: Feb 28 2009, 18:51
Go to the top of the page
+Quote Post
C.R.Helmrich
post Mar 1 2009, 01:14
Post #23





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



You're welcome. Actually, looking at page 49 and 75, I noticed that even DTS at 1500kbps is not transparent to very critical listeners on the bach_organ2 item (the first 20 seconds of Bach's Toccata, I believe). Someone (or some location on average?) gave it an 87.9 out of 100. But of course, these tests were done with very high-quality equipment in a room with very little reverberation, so chances are pretty high that even such an item would be transparent on a home theater system.


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
sauvage78
post Mar 1 2009, 05:09
Post #24





Group: Members
Posts: 677
Joined: 4-May 08
Member No.: 53282



thks to Kujibo's trick, I tested DTS & AC3 furthermore:

Here is what I did:
I took Castanet 2.0 44.1Khz I splitted it in two channels using Audacity then I resampled each channels to 48Khz, I rejoined the channels to have a lossless Castanet 2.0 48Khz as a reference.
Then, as SurCode wants 6 channels as input, I copied the right & left channels 3 times & I encoded to DTS 1536 & 768Kbps (default settings) creating two fake Castanet 5.1 48Khz files, with duplicated channels.
Then, I remuxed the two DTS file in MKA with MKVMerge & decoded it to wav using EAC3to, I deleted he fake duplicated channels with Audacity (keeping only front left & front right) & I encoded to flac & replaygained the result with F2K.

see how complicated it is to test DTS ... (with the final files, I didn't have the loudness problem I had previously)

I ended with 3 files:
Castanets Lossless WAV 2.0 48Khz.flac
Castanets Lossy DTS (SurCode V1.0.21) 307Kbps 2.0 48Khz.flac which AFAIK is equivalent to DTS 5.1 0768Kbps
Castanets Lossy DTS (SurCode V1.0.21) 614Kbps 2.0 48Khz.flac which AFAIK is equivalent to DTS 5.1 1538Kbps

Then I tried to ABX it,
I don't know if my methodology is really valid but I don't know how to do it better ...

DTS (SurCode V1.0.21) 614Kbps 2.0 48Khz Vs. WAV 2.0 48Khz:
ABX failed.

DTS (SurCode V1.0.21) 307Kbps 2.0 48Khz Vs. WAV 2.0 48Khz:
ABX 100% sucess

CODE
foo_abx 1.3.3 report
foobar2000 v0.9.6.3
2009/03/01 02:57:07

File A: C:\Documents and Settings\Sauvage\Bureau\Castanets Lossless WAV 2.0 48Khz.flac
File B: C:\Documents and Settings\Sauvage\Bureau\Castanets Lossy DTS 307Kbps 2.0 48Khz.flac

02:57:07 : Test started.
02:57:48 : 01/01  50.0%
02:58:31 : 02/02  25.0%
02:59:08 : 03/03  12.5%
03:01:43 : 04/04  6.3%
03:02:41 : 05/05  3.1%
03:04:01 : 06/06  1.6%
03:05:18 : 07/07  0.8%
03:07:32 : 08/08  0.4%
03:07:39 : Test finished.

----------
Total: 8/8 (0.4%)


Conclusion:
On castanets killer sample & with SurCode V1.0.21
DTS 5.1 1538Kbps is most likely transparent
DTS 5.1 0768Kbps is not transparent

Now I wanted to test DTS 5.1 768Kbps Vs. AC3 640Kbps to know which one was better, I took the best AC3 encoder I previously quickly tested & encoded my 2.0 48Khz wav to 2.0 256Kbps 48Khz AC3.
& first I tried to ABX the AC3 vs the original:

AC3 (Soft Encode V1.0 Build 19) 256Kbps 2.0 48Khz Vs. WAV 2.0 48Khz:
ABX 100% sucess

CODE
foo_abx 1.3.3 report
foobar2000 v0.9.6.3
2009/03/01 03:46:27

File A: C:\Documents and Settings\Sauvage\Bureau\Castanets Lossless WAV 2.0 48Khz.flac
File B: C:\Documents and Settings\Sauvage\Bureau\Castanets Lossy AC3 (Soft Encode V1.0 Build 19) 256Kbps 2.0 48Khz.flac

03:46:27 : Test started.
03:47:37 : 01/01  50.0%
03:48:26 : 02/02  25.0%
03:49:32 : 03/03  12.5%
03:51:31 : 04/04  6.3%
03:52:34 : 05/05  3.1%
03:53:51 : 06/06  1.6%
03:54:58 : 07/07  0.8%
03:56:16 : 08/08  0.4%
03:56:21 : Test finished.

----------
Total: 8/8 (0.4%)


Unlike what I previously said, with a regular ABX test, AC3 from Soft Encode V1.0 Build 19 is ABXable even at 256Kbps & quality is not OK. (It is just better than worst settings which are awfull)

Then I compared AC3 (Soft Encode V1.0 Build 19) 256Kbps 2.0 48Khz Vs. DTS (SurCode V1.0.21) 307Kbps 2.0 48Khz:
I can ABX a difference between both but I cannot really declare which is best, both are bad (really bad) but not terribly awfull (not unlistenable & not instantly ABXable)

for me
DTS preserve the fluidity better but lose the sharpness of the attacks.
AC3 sounds flatter but the attacks is better than DTS (maybe because being flat it sounds slightly slower).
both sound like "a metallic spoon rubbed against a grid", with DTS the spoon is rubbed faster but softer & with AC3 the spoon is rubbed slower but harder. (I don't know how to tell it better)
On this sample that is said to be a killer sample for transform codecs, AC3 being a transform codec & DTS being a subband codec, doesn't help DTS much.

so far personnaly I think I still favor AC3 256Kbps over DTS 307Kbps as I think I favor the "slow" effect artefact vs. the "soft" effect artefact ... if you get what I mean.
for me the sound seems more stable with AC3 256Kbps, DTS 307Kbps seems to sound slightly blurry. But it's really hard to chose between artefacts. To be fully honest I must say that I am undecided yet.
It's a matter of taste here & I may change my mind here as I am curently still ABXing them. Maybe in a week when I get used to the artefacts, I could tell which one I really favor.
I am currently in full paradox here as I think DTS seems to sound more naturals, but it softens the attack which modify the agressivity of the sample. AC3 is un-natural (slightly metallic) but doesn't soften the attacks.

It is the same kind of artefacts than between musepack & MP3 at mid/low bitrate (96-128Kbps), none is transparent at all, but it sounds different. It's like as if transform codecs were metallising the sound while suband codecs were ereasing sounds. At first try, what you don't hear sounds more natural than what hear as robotic, but I don't know why but I favor hearing artefact than not hearing data. That's why I never liked musepack personnaly, as I listen to loud music (metal) I don't like its tendancy to soften the rhythmic.

finally I don't consider AC3 5.1 640Kbps & DTS 5.1 768Kbps transparent on music anymore.
the quality of both is bad & roughtly equivalent, but it is acceptable for speech & background music.

DTS 5.1 1538Kbps is a good source for transcoding & I consider it always far superior to AC3 640Kbps ... there is no competition here.
For me, the claim that AC3 5.1 640Kbps is as transparent as DTS 5.1 1538Kbps is a commercial lie. NO doubt here.

people transcoding DTS to AC3 are nuts (specially DTS 768Kbps to AC3 & specially with free AC3 encoders which are terrible)... they should keep the original or use Nero AAC-Vorbis at 320Kps IMHO

Hope it helped. See ya.

This post has been edited by sauvage78: Mar 1 2009, 17:20


--------------------
CDImage+CUE
Secure [Low/C2/AR(2)]
Flac -4
Go to the top of the page
+Quote Post
me7
post Mar 1 2009, 11:40
Post #25





Group: Members
Posts: 177
Joined: 23-August 06
Member No.: 34375



If you'd like to include aften, you can get it here.
Go to the top of the page
+Quote Post

2 Pages V   1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 30th August 2014 - 14:05