Topic: Is AC-3 any good?

[KronKyrios]

I would like some discussion on the AC-3 format. 

Here is my dilemma:

I am trying to encode videos of concerts at the highest possible quality considering both audio and video.  I have a good capture card and can grab raw audio and video without dropping more than a frame or two in an hour.  But of course raw files are HUGE, especially the video.  So I have chosen MPEG II for video compression, using the highest possible bitrate CBR.  (the only reason to use VBR is to save on space, and I simply don't care about space usage). 

Now, I wanted to use PCM audio at 1536K (the only setting for PCM), since it is uncompressed, thereby keeping the highest possible quality audio, and using 1536K bitrate.  The alternative is to use AC-3 448K bitrate (the maximum setting allowed for AC-3). 

Most Video encoding for MPEG II at 720x480 resolution does fine at 8000K.  But the main exception to that is wide audience shots since it is a large amount of fine - fast moving - detail.  And of course concert videos typically have wide audience shots.  The MPEG standards dictate that 8000K is the maximum bitrate allowed when using PCM audio encoding.  If I use AC-3 encoding, I can increase that to 9200K, thereby preventing any potential problems with the high-motion portions of the video. 

After much searching on the web, the most commonly reported problem with AC-3 is the volume level/normalization problem.  I have not seen any consistent arguments about the quality of AC-3 encoding.  I know it is a lossy format, and I understand the reasons that is bad in and of itself... but just how bad is it?  Might the type of loss involved going to be a problem in future applications?  Personally I can't hear the differences between the lossless and the AC-3, but that is just me.  If anyone else can hear the differences, I would welcome comments from that angle as well.  Remember, I am talking aboiut the maximum setting for AC-3, so if you want to do your own tests, use that setting.  I am also referring to 2-channel audio, since that is the source of most of the items I have.  I know AC-3 is capable of multi-channel formats, and that is much of it's appeal. 

Any discussion is welcome. 

Kenneth

[hyperplane]

first off, i apologize for going on a tangent from your question/point...

you CAN get VERY GOOD video quality at a bitrate of 6000 to 7000. your assumption/assertion that VBR is 'only good for saving space' is quite innacurrate. using a software encoder such as CCE (Cinemacraft Encoder), i can do a 5 to 9-pass VBR encoding at 4000 that easily beats the majority of 8000 CBR encodings i've seen.

are you familiar with Photoshop? if so, take an image - any image - and compress it to JPEG at say Quality 6 (out of 12) with the default settings... then compress the same image using Progressive 5-pass compression. it's EASY to see the difference. the more "scans' the software does, the more it can tell where to allocate higher bitrates, and hence, the better the quality--ESPECIALLY, with video, in scenes with fast-paced motion.

what type of capture card are you using? at the consumer price level of equipment, about the best capture quality you can get is using a miniDV/digital8 camcorder "pass-through" to the computer via Firewire/IEEE1394/i.Link connection and capturing in DV .AVI format. using an All-In-Wonder type card is not so great of quality in comparison...

now, as for AC3 audio. yes, it is lossy. but most people will not be able to tell the difference between a properly encoded AC3 audio file and an uncompressed PCM file. i use SoftEncoder (which is very good, as is Minnetonka software) to encode AC3 audio - and set the dialog normalization correctly (which makes the audio level appropriately "loud" compared with the PCM track), at 448 kbps.

but even using PCM audio for a DVD, i can use CCE at 5 to 9 pass VBR encoding and have a PCM audio track, and get 69 minutes per disc. (if you lower the video bitrate, and increase the number os passes, the video quality will still be excellent and you can squeeze around 90 minutes of video at 4000 kbps onto a single layer DVDR with PCM audio).

[KronKyrios]

Thank you for your insight.  It does help to know that most people can't hear the difference on AC-3.  My larger question was whether the loss that takes place will be a problem in the future.  But I suppose nobody can really answer that.  One time I got a new soundcard and half my sound files were corrupt on the new card because the old one was doing a slop job of encoding, but it would not recreate the problems on playback.  With the quality of conversions available today, I suppose I won't have any significant problems in the future. 

I wasn't planning on discussing the Video part of this, but I do have some points:

I must challenge you on your notion that your multi-pass encoding can be any better than CBR, on a level playing field.  It is my understanding that -according to the DVD specs- the max bitrate is the max bitrate no matter whether you use VBR or CBR.  So if your VBR max is 8000K, then any single GOP (Group of Pictures) cannot be any higher than 8000K, which the CBR at 8000K does at every GOP.  Yes, the video may be fine at 6000K, but if a particular GOP needs to be encoded at 9200K, then your max bitrate setting will deny you that bandwitdh.  You can't use more than your max on that GOP just because you used less in other GOPs.

In order to be consistent, your JPEG analogy should be comparing 5-pass compression against a single pass compression of 12 (out of 12).  You aren't going to get any better than the max allowed by the format, no matter how many passes you do. 

I still say the ONLY reason to use VBR is to save space. 

Anyway, I am not capturing DV.  I am getting this from VHS (VHS sucks, I know).  I do have an ATI 9200 based card, and am using VirtualDub for the initial capture and filtering.  I am able to capture 640x480 at 30 fps progressive - direct to HUFFYUV lossless with PCM while losing about 1 to 2 frames per hour.  LOL, Maybe I will just leave everything uncompressed...at 23 Gigs per hour.  With the future of data storage, this should not be a problem for long. 

Anyway, then I am using the latest TMPGEnc for MPEG 2 encoding. It finally fully supports MPEG-2.  But I will take a look at Cinemacraft.  I had not seen that one first hand, yet.

At this point of the discussion, I am leaning toward using the AC-3 to gain the better consistency in the video.

Looking forward to your reply.

Kenneth