Hi

I have done a fun little audiotest for my friends here at AVS.

Were you can download a zipfile with 3 - 5.1 PCM files (48/16)

And just pick the one you think sound best of the 3.

The files are 1 minute each. Total file size (ZIP) is 68mb.


http://www.4shared.com/file/63120397/5a018119/test.html


Hopefully I will not have any bandwith problem this time. Poll is open for 10 days.

I hope they are not listening to this with their PC speakers :D

Sounds like you heard the track, Xylon. ;)

My PC speakers are 2.1 so am I disqualified? :)

My PC speakers are 2.1 so am I disqualified? :)

Nope, im interested in every opinion, I can tell you I had worse when I mixed this. ;)

Impressive sounds. I would say they all sound the same, but I would go with Track 3.

The low frequency pluses causes overload distortions on all tracks. It is the least objectionable on track two. The opening thunder sounds on track two is not up to the same level as tracks one and three. When time permits I will check this out on the big system.

Unfortunately I'm at work using headphones and they all sounded the same. I'll have to try it again in my theater when I get home add see if I hear a difference.

BTW is the recording from King Kong?

ok, I think they sound all the same, because I do not trust my ears that much. My impression was that track 1 and two had better bass but that 1 lacked dynamics at one part. Track three had a more detailed dinosaur roar, but worse bass.

Listening in stereo on fairly cheap towers and diy subwoofer. spdif out from my pc.

Only going by 2.1 pc speakers here, but I'd say it's almost a toss up between 1 and 3. I think 1 has more compression so I chose 3. 2 sounds weak to me.

Hi, back from a little trip to Germany.

I will release all details about the track, but I think its better to leave as much info out as possible so everyone has time to give their own impression.

But please write as much as possible from what you experience from the track, as it gives us some interesting stuff to discuss.

EDIT

We also need that people vote, 4 voters dont give me much data. ;)

Hi, back from a little trip to Germany.

I will release all details about the track, but I think its better to leave as much info out as possible so everyone has time to give their own impression.

But please write as much as possible from what you experience from the track, as it gives us some interesting stuff to discuss.

EDIT

We also need that people vote, 4 voters dont give me much data. ;)

I'm surprised that there aren't more audiophiles jumping in on this. I'm going with track 2. The volume wasn't very loud.

Ok, I bit. I thought track 2 was the best with track 3 being second best although if you told me track 2 and 3 are the same, well, I think I detect just slighlty better clarity on track 2.

I'm listening to it on a 5.1 sound system using Martin Logans all around. Scared the you know what out of my cats! The reason I like track 2 more is it seems to have more "air". For example, the echoes from the lightening strike seemed to fade of more smoothly. This was not true for track 1.

Now I'll really go out on a limb here. Track one is what I would expect if the sound was not dithered but merely truncated. The second track would be what I would expect with dithering applied. Third track, well, just not enough difference for me to make any judgement. It just sounds a tad bit off from track 2 but it is really close.

Created M2TS files so people can play it from their PS3 or other device that can play M2TS files.

http://www.4shared.com/file/63844435/6c822d05/AudioTest.html

C'mon guys...we need more votes / people to try this out. Where are all the audiophiles?

Created M2TS files so people can play it from their PS3 or other device that can play M2TS files.

http://www.4shared.com/file/63844435/6c822d05/AudioTest.html

Thanks MovieSwede. Have to copy these clips and take it to the store for testing.

It all sounds the same with my crappy speakers and crappy ears. :)

So, the poll is closed.

A quick little hint before bed.


One track has been compressed as Dolby Digital 448 kbps

One track has not been compressed.

One track has been compressed as Dolby Digital 448 kbps

One track has not been compressed.

Might I say that the tracks did not appear demanding? There is not a huge amount of detail present and usually only a few of the channels are being used at once. So it would be fairly easy to compress...or am I missing something?

Might I say that the tracks did not appear demanding? There is not a huge amount of detail present and usually only a few of the channels are being used at once. So it would be fairly easy to compress...or am I missing something?

I have no idea how demanding its for this codec, since different codec compress a bit different.

But many channels should have been used at the same time (gonna recheck that) But in a surround enviroment some channels become more dominant, and I would guess that a lossy codec takes advantage of that.

But when I mixed this, I didnt aim to just add alot of different sound in different speakers, I wanted effects that sounded like they belonged to each other.

And I didnt mix this to be easy on the Dolby encoder.

EDIT
Examined the track again, and I would say that for 80% of the track we have 4 or more channels active. Just before and after the monster, the we go down to one channel.

Examined the track again, and I would say that for 80% of the track we have 4 or more channels active. Just before and after the monster, the we go down to one channel.

Ok, I stand corrected...from what I could tell on my stereo speakers, there seemed to be only stereo cues, not surround...usually I can tell when something is SUPPOSED to be on the surround speakers, even when listening with stereo, but I guess I did not hear correctly this time.

Ok, I stand corrected...from what I could tell on my stereo speakers, there seemed to be only stereo cues, not surround...usually I can tell when something is SUPPOSED to be on the surround speakers, even when listening with stereo, but I guess I did not hear correctly this time.

Did you listen to the PCM files or the M2TS files?

did you listen to the pcm files or the m2ts files?

pcm

So when will we learn which is which?

So when will we learn which is which?

Well I can tell it right now, was wondering if anybody had different idea when they knew what to look for.

But here it is

Track 1 - DD448 with volume +2db
Track 2 - PCM (16/48)
Track 3 - DD448 with volume +2,5db

I the audio loops that I created the 5.1 mix, came from Sony Pictures Sound Effects Series.

The loops were 44,1 khz but the master is 48khz.

By extension, don't you agree that a comparison of uncompressed PCM vs DD at 640kbps or DTS at 1.5Mbps would show that almost nobody can tell the difference, despite many claims that uncompressed tracks on Blu-ray are a "night and day" improvement over the legacy codecs?

Also, one problem with tests like this is that the listener could be influenced by the file size. I didn't have time to do the listening test before the poll closed, but when I downloaded the ZIP file I noticed that track 2 compressed to a significantly smaller size than the other two. I guessed (wrongly) that meant it was the inferior track, and I'm sure that would have subconciously influenced my perception. Maybe an ISO file would be a better format to use for future tests.

By extension, don't you agree that a comparison of uncompressed PCM vs DD at 640kbps or DTS at 1.5Mbps would show that almost nobody can tell the difference, despite many claims that uncompressed tracks on Blu-ray are a "night and day" improvement over the legacy codecs?


I was planning including a DD640 track aswell, but I run into some complication with the retransfer to PCM, so I skipped it this time.

(The editor only accepted AC3 in VOB, and the VOB program only accepted DD448 etc etc)


Also, one problem with tests like this is that the listener could be influenced by the file size. I didn't have time to do the listening test before the poll closed, but when I downloaded the ZIP file I noticed that track 2 compressed to a significantly smaller size than the other two. I guessed (wrongly) that meant it was the inferior track, and I'm sure that would have subconciously influenced my perception. Maybe an ISO file would be a better format to use for future tests.

There are so many factors in this, that you can never really draw any conclusion, but you get some data that you can build some theorys on.

And that would be that any difference is harder to spot if you only have your ears then if you can use your eyes aswell.


The biggest problem here is (besides my limited mixing abilities) that people have different equipment to listen to this. Equipment shouldnt be a variable in testings like this, but as we are on the internet, We dont have so much other options.

A test should be done in a more pro enviroment.

So I got the right track. Just for a data point I did not look at the file sizes so I was using "just my ears". I guess my 53 year old ears aren't completely gone. :) Amazing how much a .5db change in volume can make when doing A/B comparisons. Didn't catch that.

So now I'm wondering, paraprashing a line from My Cousin Vinny, does my theory hold any water? Specifically, I would think that compression isn't the only factor, it's how the compression was done. I would venture to guess that the PCM track, which was the pro mix, would have been dithered to get to 16 bits. The other tracks, during compression would *probably* be using integer math to make the codec faster (floating point math is much slower). This would mean that the LSB's would be truncated. So in addition to compression, we have dithered vs truncation. If dithering was used as part of the codec then I would suspect the tracks would sound even closer.

What do you all think? Am I all wet?

The compressed file is corrupted, I hate messages like that oh, well I guess its nothing special upload it on youtube a simple clink an listen and then I’ll listen to it though the big sound system.

Well I download the other file does the file need to be renamed because it wont play, come on something simple otherwise I’m tossing this in the dustbin.

Well I download the other file does the file need to be renamed because it wont play, come on something simple otherwise I’m tossing this in the dustbin.
FWIW, I downloaded the files and they all played just fine. Did not change the file extensions. I played them back from my computer to my main sound system using a Turle Beach USB sound card and a TOSLNK connection.

BTW: Thanks MovieSwede. This has been interesting.

Error -2048: couldn’t open the file track1 m2ts because it is not a file that QuickTime understands

That’s it in the dustbin it goes not interested.

Open the RAW PCM file instead.

Open the RAW PCM file instead.

Now where is that may I ask because all I see is file zip, I open it and then I see another file I click open and nothing happens, so may I ask where is the Raw PCM file is it hidden on the file or other?

I used to argue that 99.99 percent of the people can't tell the difference between high bit-rate compressed versus uncompressed, and I still feel that's correct. I do believe, though, that blu-ray should have no less than a 1.5mbps track, and I prefer lossless because (at the very worst case) it will sound no worse than DD 448,640,1.5 or DTS 768/1.5 (assuming they are using the same mix and bit-depth).

There are some easy to hear differences between 384-448 kbps DVD soundtracks versus lossless or even 1.5mbps DD/DTS. I was watching the Sopranos seasons and rented them all on DVD and then got Season 6 on blu-ray. The minute the intro song started playing , it was clear that there was a difference... it sounded less muffled. A lot of people claim the differences is in LFE, but this was just basic intro-song and 5-10 seconds... and i wasn't looking for it. The other clear difference was Entrapment. I was doing some PQ comparisons and when I heard the difference in sound, I was shocked. Granted, this was an old DVD from 2001 compared to a blu-ray from 2006, but even only playing at CORE DTS 1.5mbps, it was clearly better. In both of these cases, I wasn't "looking" (listening) for it.

In the case of this test, it wasn't nearly as easy to tell the difference, so there's obviously more to equation (Just like PQ and the whole bit-rate debate). I do feel a more demanding soundtrack may have been useful. Glad we did the test...I just wish more people participated in this.

I used to argue that 99.99 percent of the people can't tell the difference between high bit-rate compressed versus uncompressed, and I still feel that's correct. I do believe, though, that blu-ray should have no less than a 1.5mbps track, and I prefer lossless because (at the very worst case) it will sound no worse than DD 448,640,1.5 or DTS 768/1.5 (assuming they are using the same mix and bit-depth).

Yeah, BD might as well have lossless, there's plenty of space.

There are some easy to hear differences between 384-448 kbps DVD soundtracks versus lossless or even 1.5mbps DD/DTS. I was watching the Sopranos seasons and rented them all on DVD and then got Season 6 on blu-ray. The minute the intro song started playing , it was clear that there was a difference... it sounded less muffled. A lot of people claim the differences is in LFE, but this was just basic intro-song and 5-10 seconds... and i wasn't looking for it. The other clear difference was Entrapment. I was doing some PQ comparisons and when I heard the difference in sound, I was shocked. Granted, this was an old DVD from 2001 compared to a blu-ray from 2006, but even only playing at CORE DTS 1.5mbps, it was clearly better. In both of these cases, I wasn't "looking" (listening) for it.

You've got to be really, really careful with comparisons like that. Starting with entrapment, there were there's 5-6 years between those releases, it's quite possible, probable perhaps that there was a lot more different between the two releases than just the compression. You'd be better off comparing the DTS core to it's full lossless version.

Somewhat dito with Soprano's, you're comparing completely different releases, different years/seasons, there's quite possibly more different there than just the compression.

You've got to be really, really careful with comparisons like that. Starting with entrapment, there were there's 5-6 years between those releases, it's quite possible, probable perhaps that there was a lot more different between the two releases than just the compression. You'd be better off comparing the DTS core to it's full lossless version.


I probably couldnt tell the difference between the core and the lossless, but it's too hard to actually test since I can't switch from lossless to core without a huge pause in between. I did note one was 5-6 years older, so it's not a true fair comparison, but I'm just saying it was extremely easy to tell the difference between old dvd and DTS core...


Somewhat dito with Soprano's, you're comparing completely different releases, different years/seasons, there's quite possibly more different there than just the compression.

This was season 5 versus season 6, so I think the comparison was more "fair" than the entrapment comparison as far as time between.

I also believe that when Dexter comes out on blu-ray, there will be a significant difference in sound quality... and Dexter is a very new DVD... it just sounds really compressed. Can't wait for the blu-ray!

Great test MovieSwede. I wish more people took part in this. maybe post on the speaker thread.

The point is that even with no video, fast switching (PS3), and intent to listen, I failed the test. I thought that they all sounded the same. It often makes me wonder why bother with lossless at all if compression can be that good, and if only by careful testing can you (maybe) pick the flaws out. Maybe different sound effects can better show off the differences, maybe not. But I'm not in this hobby anymore to sit in a dark room to detect negligible differences in audio or video. My life is not that long. Sadly, I think too many AVSers get obsessed with specs, rather than the actual performance. I laugh at the "Speed Racer BD" thread when everyone was ready to burn Warner for its BD-25 treatment, only to discover it is a damn good video transfer. Yet they still yearn for BD-50.

Another idea for a future test:
Mix 10 tracks. Only one is PCM. The other 9 all the same. See if anyone can pick it out. This should eliminate guessing. or even more cruel...

Mix 10 all identical lossy tracks. Tell people that one is lossless and ask them to pick it out. This should be funny.

So now I'm wondering, paraprashing a line from My Cousin Vinny, does my theory hold any water? Specifically, I would think that compression isn't the only factor, it's how the compression was done. I would venture to guess that the PCM track, which was the pro mix, would have been dithered to get to 16 bits. The other tracks, during compression would *probably* be using integer math to make the codec faster (floating point math is much slower). This would mean that the LSB's would be truncated. So in addition to compression, we have dithered vs truncation. If dithering was used as part of the codec then I would suspect the tracks would sound even closer.

What do you all think? Am I all wet?
Since you quoted one of my favorite movies which I have watched a 100 times, I have no choice but to answer :D.

Your case holds 50% water :). Codec manufacturers have different specs with regards to certifying the implementation. And some have none. For example, you can implement AAC and MP3 as long as something plays, you can say it is that type of codec! The signal to noise ratio of such an implementation could be way down in 65 to 70 db (same as cassette tape) but what folks don't know, well, doesn't bother them I guess.

Companies with certification requirements aim higher but if they go too high, as you implied, it may get so hard to implement in the limited precision of the DSP processors in low-end gear as not cause you to lose business. I don't know what Dolby or DTS require as far as accuracy here. But it is certainly possible that it is well below that if CD/96 db. I seem to recall for WMA we had certification requirement but due to MP3/AAC having none, we had to settle for lower number there too.

Note that finding the level of accuracy is very difficult. A codec may reproduce 99% of the sources transparently (with regards to accuracy) but then fall down to 65 db for a few segments.

All in all, not every implementation will sound the same.

As an aside, voice codecs for cell phones and such usually require bit-exact implementation meaning that no room is left for the decoder to deviate from an ideal implementation. They worry that their signal quality is low enough as it is, that allowing some accuracy loss could make things really bad.

Hi amirm! Long time no see. Looks like you have a new job. Thought you retired. :D

So I'm not all wet, just partially soaked. :) Not sure I understood everything you wrote. I can unsderstand that if you use a hardware DSP encoder, you're going to get better results because you can put the entire algorithm in optimized hardware. Hence the cell phone example you used. But just to be clear, I would guess that MovieSwede used a software compression program and not hardware. I'm sure there are differences in software codecs but I would still guess that some/most use integer math for speed which results in truncation.

I guess what I am really trying to get to is, a properly dithered lossy compression is going to have actual data in the LSBs that can improve the sound. Perhaps by 1 to 2 bits of resolution. On 24 bit depth, no big deal. On 16, could be a problem. A truncated signal is just gonna drop this information. Could that be why I claim to hear more "air" in track 2 on the thunderclap?

BTW: I LOVE My Cousin Vinnie. Cracks me up every time! And I really appreciate your posts. I've learned alot from you!

Hi amirm! Long time no see. Looks like you have a new job. Thought you retired. :D

So I'm not all wet, just partially soaked. :) Not sure I understood everything you wrote. I can unsderstand that if you use a hardware DSP encoder, you're going to get better results because you can put the entire algorithm in optimized hardware. Hence the cell phone example you used. But just to be clear, I would guess that MovieSwede used a software compression program and not hardware. I'm sure there are differences in software codecs but I would still guess that some/most use integer math for speed which results in truncation.

Actually often times it's the other way around. Usually hardware is designed for realtime operation, and for techincal and financial reasons, such and encoder will often sacrifice quality/accuracy to meet the design goals (both technical and financial) of realtime encoding.

Software on the other hand often doesn't have realtime requirements and can thus be designed for quality. You'll see this a lot in video codecs. You'll find hardware MPEG-2 encoders on PC tuner cards, but these are far less efficient and to a poorer job compared to a decent software encoder without the requirement to run in real time.

Hi amirm! Long time no see. Looks like you have a new job. Thought you retired. :D
Still am retired. But it seems that when you are retired, folks want you to write articles for free :). I enjoy doing it though so it keeps me busy a bit.

So I'm not all wet, just partially soaked. :) Not sure I understood everything you wrote. I can unsderstand that if you use a hardware DSP encoder...
OK, let's step back for some introduction.

The process here is simple. You have PCM input to the encoder. The encoder compresses it and produces what we call a "bit stream." The bit stream is put on disc. At playback, this bit stream needs to be converted back to PCM in order to be played. None of this has anything to do with it being in hardware or software. The process is one in the same.

Now, I think it is well understood that encoders are NOT standardized. That is, they can continue to be changed and improved as long as they produce a bit stream which the decoder (which can NOT be changed) can understand and convert back to PCM. This is the reason quality of these systems improve over time whether we are talking about audio or video.

The finer point being raised here is that not all decoders work the same, or have the same fidelity. If the implementation does not require bit-exact decoders, then there are variations between decoders, given the identical encoder/bit stream.

How can there be differences? Reason is that there are many different decoder systems. In case of audio, we usually use a special kind of processor we call a DSP. The DSP is software programmed just like a PC would run an application. DSP systems come in various speed grades and different architectures with respect to ability to handle floating point math (i.e. anything with fractions in it). And of course, there is a person involved, interpreting the specifications and attempting to write a compliant decoder.

A codec like Dolby Digital uses transforms which convert a time varying signal into its frequency components. Such math requires fair amount of computation which can be too slow to implement on some processors. In addition, the reference implementation provider by the original vendor may use floating point, whereas the target may not. Or even if it is an integer implementation, the target may lack the same level of resolution. For example, the DSP may have 20 bits of resolution yet the PCM samples may be 24-bits. There are tricks to handle this still but depending on the implementation skill of the programmer and processing horsepower available, shortcuts may need to be taken.

What do the shortcuts do? Let’s look at a simple example. Assuming that we have an ideal system with zero loss from encoder to decoder. That is, we compress the signal, create a bit stream and then decode that into PCM. In that situation we expect what we put in the encoder to come out the decoder. So let’s assume the PCM samples are 35, 38, 42, 43, and 45. An ideal system would reproduce exactly the same values on the output. The system taking some shortcuts in order to meet system constraints, would potentially produce 35, 37.99, 42, 43, 45.01. As you can see, if you round these numbers, you get to the same PCM values. But there is still some error.

What does the error sound like? Well, we can treat it for now as noise. If you take a clean signal like the source was, and varying it slightly as I did above, the end result is not that different than simply adding noise to it. In other words, I can take the PCM value 45 and add 0.01 amount of noise to it and arrive at 45.01. This is the reason I used signal to noise ratio to describe the effect here.

What gets complicated however is that while noise is random, the accuracy error here may not be. It is too complicated to describe here but essentially the noise would be “correlated” with the source which is not good. Imagine anything there is a high frequency you get noise, but not other times. Your ear will be able to hear this distortion more than if the noise was always there.

I guess what I am really trying to get to is, a properly dithered lossy compression is going to have actual data in the LSBs that can improve the sound. Perhaps by 1 to 2 bits of resolution. On 24 bit depth, no big deal. On 16, could be a problem. A truncated signal is just gonna drop this information. Could that be why I claim to hear more "air" in track 2 on the thunderclap?
I have not listened to the test tracks here. But yes, in general, 24-bit signals are self-dithering so you don’t hear the truncation noise and as a result, are able to better discern low level detail which provides “air” and ambiance.

BTW: I LOVE My Cousin Vinnie. Cracks me up every time! And I really appreciate your posts. I've learned alot from you!
Good to know I don’t have to walk around embarrassed that I am the only one who loves this movie so much. The writing and acting is so much better than it needed to be: “Two Yoots... Two what? Two Yoots? What is a Yoot? Sorry your honor, I meant to say youths…”

Just wanted to add, that I didnt have a 24bit source on this one.

16bit 44,1 khz stereo and mono loops, mixed in a 5.1 workspace (48/16).