FFDShow vs. AC3Filter Audio Quality

[Ditto666]

Please answer only if you actually tested this with high quality surround speakers, using some sort of surround audio streams...
This is not a question of which you like better as a program or which you're used to.

The results would only make sense if one outweighed the other by a lot, like it would be if the poll was between 128kbps vs. 256kbps MP3s, as an example. Mixed results would only mean one of two things; people are either voting with a bias or, both create exactly the same quality...

[Dark Shikari]

This poll doesn't even make sense. The only variation allowed between AC-3 decoders is the random distribution used for frequency coefficients with a bap value of zero and the floating point rounding during decoding.

None of these would affect how "rich" or "clear" the audio sounded.

[Ditto666]

This poll doesn't even make sense. The only variation allowed between AC-3 decoders is the random distribution used for frequency coefficients with a bap value of zero and the floating point rounding during decoding.

None of these would affect how "rich" or "clear" the audio sounded.

Lol, maybe. I was thinking that too. But the sound was so much more powerful coming from AC3Filter. Maybe I was imagining things. I'll have to compare when my mind is clear. Too much obsessing in one day...

[Ditto666]

No.. I'm not crazy. Maybe you just don't know what difference actually exists. With all the settings exactly the same, AC3Filter definitely sounds better to me. I think this is a legitimate poll. Sometimes, you just gotta put all the technical stuff aside.

[Inspector.Gadget]

Sometimes, you just gotta put all the technical stuff aside.

No. The "technical stuff" - ABX tests, etc. - is the only reliable measure of quality.

[Midzuki]

ffdshow has two options to decode AC3: via liba52, and via libavcodec. Do they both sound "equally bad" when compared to the output from AC3Filter? :confused:

[Ditto666]

ffdshow has two options to decode AC3: via liba52, and via libavcodec. Do they both sound "equally bad" when compared to the output from AC3Filter? :confused:

Well, I'm actually comparing DTS - REAL DTS... And FFDShow doesn;t sound bad. Just worse...

[Midzuki]

Ooops --- I misread, and miswrote. Apologies. :stupid:

Instead of "liba52", read "libdts". :o

[Ditto666]

Ooops --- I misread, and miswrote. Apologies. :stupid:

Instead of "liba52", read "libdts". :o

Well, anyway, funny thing is, I find it even worse when passing the audio through to the receiver... Some people found that puzzling. It sounds different but not in a good way I think.

[_xxl]

Well, I'm actually comparing DTS - REAL DTS... And FFDShow doesn;t sound bad. Just worse...
A sample please.

[Ditto666]

A sample please.

http://www.filedropper.com/dtsdvd-a-lazingonasundayafternoon

EDIT: Just want to mention.. I've gone though more songs, listening with FFDShow. Then, I just decided to play the same thing with AC3Filter and my eyes finally opened. Everything is clearer and sharper. Whatever difference exists, it makes AC3Filter sound better for sure. I am convinced now.

[leeperry]

yeah, libdts sounds terrible(hissy/distorted dialogs)..libavcodec is cool, too bad libavcodec decodes to 16int when libdts is 32float.

but because of that bug, AC3Filter cannot be used in ffdshow if you wanna make autoload profiles: http://sourceforge.net/tracker/?func=detail&atid=867360&aid=2725991&group_id=173941

I personally like Sonic 4.2(at least they have an official DTS licence, it's no dodgy reverse engineering....but you can't completely disable DRC :()

[RunningSkittle]

http://www.filedropper.com/dtsdvd-a-lazingonasundayafternoon

EDIT: Just want to mention.. I've gone though more songs, listening with FFDShow. Then, I just decided to play the same thing with AC3Filter and my eyes finally opened. Everything is clearer and sharper. Whatever difference exists, it makes AC3Filter sound better for sure. I am convinced now.

You need to ABX otherwise your test means nothing.

[tetsuo55]

To do this test correctly:

Hook up your system, so you can record the output from the player.

Then record the stream for each decoder (3 in total).
Compare the 3 streams with a reference decoder (one that is guaranteed to be accurate)

The one that matches the reference decoder best is the most accurate.

I'm very interested in the results of such a test (theoretically libavcodec should win)

[Ditto666]

You need to ABX otherwise your test means nothing.

What's ABX?

To do this test correctly:

Hook up your system, so you can record the output from the player.

Then record the stream for each decoder (3 in total).
Compare the 3 streams with a reference decoder (one that is guaranteed to be accurate)

The one that matches the reference decoder best is the most accurate.

I'm very interested in the results of such a test (theoretically libavcodec should win)

AC3Filter uses libavcodec?

In regards to the test, I don't think I can do it >.<

[leeperry]

as much as I agree that libdts sounds awful, libavcodec in ffdshow sound great..do you also fault it?

but libdts decodes in 32float, where libavcodec is only 16int...which is too bad considering DTS is said to be 18bit.

[tebasuna51]

Sorry, but the AC3Filter v1.63b DTS decoder have a bug.

The output is amplified by 3 dB and the peaks are clipped.
I have all the controls to 0 or Off:
Gain
DRC
Auto Gain
Auto Matrix
Equalizer

The sample was generated by me and Dts Pro Encoder.
Decoded with ffdshow (without differences betwen libdts and libavcodec) and ArcSoft the volume level is correct.

One more time higher volume is considered by ones like more quality, but here is on the contrary.

[Ditto666]

Sorry, but the AC3Filter v1.63b DTS decoder have a bug.

The output is amplified by 3 dB and the peaks are clipped.
I have all the controls to 0 or Off:
Gain
DRC
Auto Gain
Auto Matrix
Equalizer

The sample was generated by me and Dts Pro Encoder.
Decoded with ffdshow (without differences betwen libdts and libavcodec) and ArcSoft the volume level is correct.

One more time higher volume is considered by ones like more quality, but here is on the contrary.

How typical for someone here to go by technical numbers... :/ When it comes to music, that's just something that isn't done. That's unfortunate to hear though. Nevertheless, it doesn't change whatever it is that makes it sound better. Maybe I'm wrong though, dunno again... Sigh...

Two questions though, because maybe I need to retest a few things.. well, the old fashioned way. Does AC3Filter use libavcodec? And, when putting a number between BackLeft and SideLeft in the Matrix Mixer, for example, where BackLeft has a value of 1, does that number affect the Backleft in any way or does it leave it untouched and just adds to the side channel?

[Inspector.Gadget]

How typical for someone here to go by technical numbers... :/ When it comes to music, that's just something that isn't done. That's unfortunate to hear though. Nevertheless, it doesn't change whatever it is that makes it sound better. Maybe I'm wrong though, dunno again... Sigh...

"Technical numbers" are the only valid way one has to verify results. And you're flat wrong about music: Hydrogenaudio has rigorous rules about making evidence-based claims and generally insists on ABX testing. If you're looking for handwaving and unsupported claims, perhaps a forum that at bottom is essentially about mathematics isn't the right venue.

[THX-UltraII]

yeah, libdts sounds terrible(hissy/distorted dialogs)..did not notice this at all. I ll test tonight. I checked 16,24 and 32 bit in ffdshow as output because I noticed that float is not supported by my ATI HD4350. ffdshow outputs 32bit cause I can see that in the info of Reclock. How do I know if Reclock also OUTputs 32bit? IF reclock outputs 32bit, is this then 'downgraded' to 18bit cause this is the max of DTS?

libavcodec is cool, too bad libavcodec decodes to 16int when libdts is 32float.Why would someone need 32float? (what IS float btw?)

I personally like Sonic 4.2(at least they have an official DTS licence, it's no dodgy reverse engineering....but you can't completely disable DRC :(0what is DRC? Is this a filter that can work together with ffdshow?

[Ditto666]

did not notice this at all. I ll test tonight. I checked 16,24 and 32 bit in ffdshow as output because I noticed that float is not supported by my ATI HD4350. ffdshow outputs 32bit cause I can see that in the info of Reclock. How do I know if Reclock also OUTputs 32bit? IF reclock outputs 32bit, is this then 'downgraded' to 18bit cause this is the max of DTS?

Why would someone need 32float? (what IS float btw?)

what is DRC? Is this a filter that can work together with ffdshow?

Good questions... Questions that I can't fully answer myself and would like to know the answers to as well.

[PatlaborForce]

Why would someone need 32float? (what IS float btw?)

http://en.wikipedia.org/wiki/Floating_point

what is DRC? Is this a filter that can work together with ffdshow?

http://en.wikipedia.org/wiki/Dynamic_range_compression

[Ditto666]

http://en.wikipedia.org/wiki/Floating_point



http://en.wikipedia.org/wiki/Dynamic_range_compression

Yes, because that's why he asked. To read an entire Wikipedia article instead of a quick explanation of what it is...

[PatlaborForce]

Yes, because that's why he asked. To read an entire Wikipedia article instead of a quick explanation of what it is...

Then that's why you read the summary sentences if all you want is a brief summary of the topic.

In computing, floating point describes a system for numerical representation in which a string of digits (or bits) represents a rational number.

The term floating point refers to the fact that the radix point (decimal point, or, more commonly in computers, binary point) can "float"; that is, it can be placed anywhere relative to the significant digits of the number. This position is indicated separately in the internal representation, and floating-point representation can thus be thought of as a computer realization of scientific notation. Over the years, several different floating-point representations have been used in computers; however, for the last ten years the most commonly encountered representation is that defined by the IEEE 754 Standard.

Dynamic range compression, also called DRC (often seen in DVD and car CD player settings) or simply compression, is a process that reduces the dynamic range of an audio signal. Compression is used during sound recording, live sound reinforcement, and broadcasting to control the level of audio. A compressor is the device used to apply compression.

They are linked because he or someone else may be interested. I didn't realize that providing people with information on a topic they are interested in was some how a bad thing.

[leeperry]

32float is better because of this: http://mp3decoders.mp3-tech.org/24bit.html

lossy audio should always be decoded in 32float, and I do quite a bit of post-processing in ffdshow(custom mixing matrix, etc) + Reclock resampling(that works in 32float internally)...so it's a nice addition.

and DRC is not something you want, as it will compress the dynamics.

liba52 in ffdshow decodes in 32float(I checked w/ Ozone4) and SQ is great I think...but the AC3 specs are public, so no surprise here....but what about DTS? what's a good spec-compliant 32float DRC-free decoder? doesn't exist, right?