Newsletter #113 - July 22, 2006
The Audio Sample Rate
of DV-AVI Files - 32kHz or 48kHz?
There was an interesting exchange of posts about audio issues on
the Movie Maker newsgroup about a week ago. They were saying that audio
issues result from Movie Maker 'up-sampling' audio to 48kHz. The
posters agreed that MM2 does such a poor job of it that you
should up-sample with other software before using the source file
in a project.
I added the essence of the posts to my Problem Solving > Audio
Issues page, and it seemed like a good subject for a newsletter. Lots of users
including myself have regular or intermittent audio issues. They drive
me more and more to saving as high quality WMV files instead of
DV-AVI.
Maybe the posters are right about the up-sampling being the cause...
let's take a look. The properties of a typical type I DV-AVI source
file from my camcorder, as shown by MM2, has an audio sample rate
of 32 kHz... when captured with MM2 or WinDV.
Rendering it to a new DV-AVI file in MM2 shows the sample rate of
saved movies is 48 kHz. According to the posters, it's this part of the
process that introduces significant audio artifacts because Movie Maker doesn't
do it very well.
Let's use this newsletter to study it enough to see if the posters
are right... and what we can do about it.
FLASH
At the 11th hour toward the newsletter issue deadline, I got a clue
that the audio sample rate being reported by MM2 in XP and Movie
Maker in Vista might be in error, and there's really no up-sampling going on
when the camcorder records the audio at 48 kHz.
Opening the files with other software such as VirtualDub (the
screen shot at the right) shows them as being 48 kHz.
It appears to be a
reporting issue by Movie Maker, not a resampling, and Microsoft is checking
into it...
As the bottom line isn't conclusive yet, I'm issuing the newsletter with
hopes that it raises your awareness of the audio sampling rate.
... before getting into it
further, a few notes...
Notes...
My first
for-sale video on Google
Video... submitted May 3... still has the status of "Video
is verified; stay tuned - it will be live shortly"... it's been over 2
months in the queue to be online.
I've stopped checking it's status every day and now taking bets as
to whether or not it ever gets to the next step of 'being live'.
The
MaximumPC issue with the Photo Story 3
tutorial is on magazine stands in the US. The 7 page article starts
on page 68 in the Special Summer 2006 edition....
You
can automatically open a second video at the end of the
first one... this item is too short for a full newsletter issue,
but worth a mini-mini tutorial.
My
latest Video
Doodle is a 2-part one. The first part is a 10 second
countdown Photo Story that ends by opening another
story...
The
first one opens the second by an embedded 'script
command'.
Open the story file (or WMV movie), go to the frame where you want
to embed the command, press the Script Commands button, then the Add button, and
enter the full URL. In this case it's
for
the 2nd story.
You
can have it open a website page, a movie, a story, an image... any URL address
you usually enter in a browser.
With the high quality of photo stories and the average quality of online
videos, it might be a neat way to make a 1-2 combo without lowering
the quality of the story by re-rendering it as a clip in
the movie.
It's not for everyone, as a security setting in the Windows Media
Player may need tweaking from its default. Details are included at the end of
the first story... if the 2nd story doesn't automatically start.
Vista Corner... I was using it to
cross-check the audio sampling rate of captured files for this
newsletter. Its reports are consistent with MM2 in XP, but now looking to
be consistently wrong.
.... back to the main
topic...
The Audio Sample
Rate
Real world audio waves are analog,
not digital. The sample rate is like putting dots on the audio
waves to approximate them... like pixels approximates visual images.
Put 32,000 dots on the waves each second and you have a sample rate of 32 kHz.
48,000 dots gets you 48 kHz. Can you hear the difference?... probably not. Is it
an academic discussion?... you hope... but it might depend on the quality of the
process that's used to change from one rate to another.
I'll start with my digital camcorder,
a Sony TRV80. The manual says the
audio recording for 16 bit stereo (my normal setting) is at a 48 kHz
sample rate.
Captured
Files using various apps
get you either 32 or 48 kHz, as reported by the properties in Movie
Maker
Rather than having to rip the audio
from the type I files and do the up-sampling with another app, maybe
using MM1 or WinDV to capture to type II files gets you there easier and
with better results. (I said that before finding that the reporting
might be in error and the files are actually starting at 48
kHz)
Ripped or
Re-rendered files from the captured files - using default
settings
-
MM2 rips the audio of a DV -AVI file to a High Quality WMA file
with a sample rate of 44.1 kHz... but the WMA file is a frame shorter than the
original DV -AVI file. This isn't an issue as we know the dropped frame
is the last one, which won't effect the movie's audio/video
sync.
- MM1 renders to type II DV-AVI files
with 48 kHz.
- MM2 renders to type I DV-AVI files
with 48 kHz.
Test Project
I captured a one minute segment of a
Christmas concert... the same segment 4 times, using MM1, MM2, and WinDV for
each of types I and II. Here's the storyboard view of this simple
project.
I rendered it to new
DV-AVI files a number of times to check how the 'upsampling' from the
32 kHz source files makes it versus those that don't need upsampling. After each
render I made the project a bit more complex to see if and when
the audio starts to run into problems.
- 1st render - nothing done to any of
the clips other than add title overlays to note which of the files
was playing
- 2nd render - split each of the 4 one
minute source clips into 6 parts... every 10 seconds on the
timeline
- 3rd render - added a random effect to
each clip
- 4th render - added an MP3 music file
to the audio/music track... balanced it such that the audio of the
video clips is prominent... to see if mixing DV-AVI and MP3 audio results in
audio issues
- 5th render - added transitions between all
clips... just using the standard fade one
- 6th render - changed the fade transitions to
others, using a random assortment
- 7th render - added fade-in and fade-out to the
audio track of each video clip
- 8th render - shifted the balance from the audio
of the video clips to the MP3 music
Each of the rendered DV-AVI files
sounded as good as the original. I couldn't hear any differences
between the 32 kHz or the 48 kHz source file segments. I wasn't
successful at getting the audio to breakdown, not even a hiccup or
short blip.
That kind of supports the issue being one of
reporting the sample rate, not one of actual differences...
Conclusions and Closing
I asked the other digital media MVPs
and Microsoft for comments about this
subject, digested their collective thoughts, and mulled things over
for a day as I edited the issue for the final time after finding out about the
reporting issue.
My conclusion was heading this way
before the 11th hour finding...
- My Sony camcorder records 16 bit audio at a 48 kHz
sample rate
- Type II DV-AVI capture processes maintain the 48
kHz rate... but type I captures convert it to 32
kHz
- Movie Maker 2 and Vista save DV-AVI files at the 48 kHz
rate
Maintaining the sample rate at 48 kHz from
camcorder tape to captured file to rendered movie seems preferable to going
from 48 to 32 and then back to 48 kHz... because resampling processes
can reduce quality or introduce audio artifacts.
I'll let you know next week or future issues what else is found....
now we're heading off to the Ann Arbor Art Fair to look at pictures and
take some video... have a great weekend.
Have a great week...
PapaJohn
