[Frank Stearns]

A client handed me several mono files of a chorus recorded with a
Canon T3i DSLR camera.

Yeah, I know... clients do the cutest things. (But it's all they
had. The client did want to disable the camera AGC but apparently the
AGC did everything, including input trim -- no way to adjust to the
range needed.)

Client said that the clipping was terrible without the AGC engaged,
so he left it on. Fortunately, at least, these tracks are clip-free;
and this was done with an external microphone.

I have repaired most of the tonality issues, even added a believable
stereo field. But completely undoing the AGC is a challenge. I've
been able to fix much of its damage by cascading two expanders with
different time constants and ratios, but by no means are all the
dynamics fully repaired.

As a point of reference, so far the repaired tracks sound more like
music and they can be tolerated (barely), while the original tracks
are unlistenable -- sounds like they were encoded twice by the old
DBX tape noise reduction system but then never decoded.

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

Thanks in advance for any information,

Frank
Mobile Audio
--

[Don Pearce[_3_]]

On Sun, 15 Sep 2013 06:55:07 -0500, Frank Stearns
wrote:

A client handed me several mono files of a chorus recorded with a
Canon T3i DSLR camera.

Yeah, I know... clients do the cutest things. (But it's all they
had. The client did want to disable the camera AGC but apparently the
AGC did everything, including input trim -- no way to adjust to the
range needed.)

Client said that the clipping was terrible without the AGC engaged,
so he left it on. Fortunately, at least, these tracks are clip-free;
and this was done with an external microphone.

I have repaired most of the tonality issues, even added a believable
stereo field. But completely undoing the AGC is a challenge. I've
been able to fix much of its damage by cascading two expanders with
different time constants and ratios, but by no means are all the
dynamics fully repaired.

As a point of reference, so far the repaired tracks sound more like
music and they can be tolerated (barely), while the original tracks
are unlistenable -- sounds like they were encoded twice by the old
DBX tape noise reduction system but then never decoded.

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

Thanks in advance for any information,

Frank
Mobile Audio

Can't help with this, but from now on, be sure to run Magic Lantern
firmware on the camera. Absolutely brilliant for the video, and gives
full control over audio, including disabled AGC.

d

[Scott Dorsey]

Frank Stearns wrote:

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

I don't, but borrow the camera and try applying tone bursts of varying
lengths to it and seeing what the envelope of what comes out looks like.

I'd start with just a 1 KC tone burst, 10 ms, 100 ms, 1000 ms, and 10,000 ms
long.

Then I'd try 100 Hz and 5 KHz just to see if they were doing anything that
might be frequency-specific in there. If they're smart they're probably
compressing the bass separately to prevent so much pumping under rumbly
street noise situations. Then again, if they were smart you wouldn't be
having to do this in the first place.

But I think getting the camera and trying to characterize it might be your
best bet.

Still, I am inclined to think eliminating AGC is close to impossible; once
information is lost, it's not going to come back. If there was a miracle
available, someone would use it on those Toscanini recordings....
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Don Pearce[_3_]]

On Sun, 15 Sep 2013 06:55:07 -0500, Frank Stearns
wrote:

A client handed me several mono files of a chorus recorded with a
Canon T3i DSLR camera.

Yeah, I know... clients do the cutest things. (But it's all they
had. The client did want to disable the camera AGC but apparently the
AGC did everything, including input trim -- no way to adjust to the
range needed.)

Client said that the clipping was terrible without the AGC engaged,
so he left it on. Fortunately, at least, these tracks are clip-free;
and this was done with an external microphone.

I have repaired most of the tonality issues, even added a believable
stereo field. But completely undoing the AGC is a challenge. I've
been able to fix much of its damage by cascading two expanders with
different time constants and ratios, but by no means are all the
dynamics fully repaired.

As a point of reference, so far the repaired tracks sound more like
music and they can be tolerated (barely), while the original tracks
are unlistenable -- sounds like they were encoded twice by the old
DBX tape noise reduction system but then never decoded.

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

Thanks in advance for any information,

Frank
Mobile Audio

This may help. I've just hit my Canon with a 440Hz tone and this is
the result with the AGC enabled. It isn't pretty - the first few
cycles are well clipped, so you may not find the results brilliant
even if you can neutralize it.

http://www.soundthoughts.co.uk/look/Canon%20AGC.png

d

[Don Pearce[_3_]]

On Sun, 15 Sep 2013 15:19:19 GMT, (Don Pearce) wrote:

On Sun, 15 Sep 2013 06:55:07 -0500, Frank Stearns
wrote:

A client handed me several mono files of a chorus recorded with a
Canon T3i DSLR camera.

Yeah, I know... clients do the cutest things. (But it's all they
had. The client did want to disable the camera AGC but apparently the
AGC did everything, including input trim -- no way to adjust to the
range needed.)

Client said that the clipping was terrible without the AGC engaged,
so he left it on. Fortunately, at least, these tracks are clip-free;
and this was done with an external microphone.

I have repaired most of the tonality issues, even added a believable
stereo field. But completely undoing the AGC is a challenge. I've
been able to fix much of its damage by cascading two expanders with
different time constants and ratios, but by no means are all the
dynamics fully repaired.

As a point of reference, so far the repaired tracks sound more like
music and they can be tolerated (barely), while the original tracks
are unlistenable -- sounds like they were encoded twice by the old
DBX tape noise reduction system but then never decoded.

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

Thanks in advance for any information,

Frank
Mobile Audio

This may help. I've just hit my Canon with a 440Hz tone and this is
the result with the AGC enabled. It isn't pretty - the first few
cycles are well clipped, so you may not find the results brilliant
even if you can neutralize it.

http://www.soundthoughts.co.uk/look/Canon%20AGC.png

d

Sorry - less than useful. Back with something definitive in a minute.
You won't like it.

d

[Frank Stearns]

(Scott Dorsey) writes:

Frank Stearns wrote:

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

I don't, but borrow the camera and try applying tone bursts of varying
lengths to it and seeing what the envelope of what comes out looks like.

I'd start with just a 1 KC tone burst, 10 ms, 100 ms, 1000 ms, and 10,000 ms
long.

Then I'd try 100 Hz and 5 KHz just to see if they were doing anything that
might be frequency-specific in there. If they're smart they're probably
compressing the bass separately to prevent so much pumping under rumbly
street noise situations. Then again, if they were smart you wouldn't be
having to do this in the first place.


Thanks, Scott. This is a great idea. Unfortunately, the beast is 1000 miles away,
but maybe I can borrow one locally, or have the owner run some tests and send me the
resulting files.

Still, I am inclined to think eliminating AGC is close to impossible; once
information is lost, it's not going to come back. If there was a miracle
available, someone would use it on those Toscanini recordings....

That was my first take on things, but then there was that memory of the old DBX NR.
I now surely have a much fuller appreciation of why calibration was so important in
such systems.

Interestingly, allowing only 1 or 2 dB of range on the expanders seems to help a
lot, even though one can guess that much more dynamic range than that was removed
during recording. As I set the expanders to a wider range, the reconstruction falls
apart.

I'm cautiously optimistic that if I can dial in the correct times and ratios, I can
increase the expansion range another few dB and recover a little more (and improve
the illusion); though it is probably a sure thing that full dynamic range recovery
is not possible.

This is one area where I am very thankful for digital dynamics processing. I have
look-ahead and very precise controls.

Thanks again for the great idea.

Frank
Mobile Audio
--

[Don Pearce[_3_]]

On Sun, 15 Sep 2013 06:55:07 -0500, Frank Stearns
wrote:

A client handed me several mono files of a chorus recorded with a
Canon T3i DSLR camera.

Yeah, I know... clients do the cutest things. (But it's all they
had. The client did want to disable the camera AGC but apparently the
AGC did everything, including input trim -- no way to adjust to the
range needed.)

Client said that the clipping was terrible without the AGC engaged,
so he left it on. Fortunately, at least, these tracks are clip-free;
and this was done with an external microphone.

I have repaired most of the tonality issues, even added a believable
stereo field. But completely undoing the AGC is a challenge. I've
been able to fix much of its damage by cascading two expanders with
different time constants and ratios, but by no means are all the
dynamics fully repaired.

As a point of reference, so far the repaired tracks sound more like
music and they can be tolerated (barely), while the original tracks
are unlistenable -- sounds like they were encoded twice by the old
DBX tape noise reduction system but then never decoded.

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

Does anyone know the specific ratios and time constants used in the
Canon T3i? I'd like to get a little closer "rough in" of the
expansion before I spend a lot more time finessing. Are any of the
dynamics time and ratio characteristics published? I can't seem to
find anything other than turning AGC on or off, and that horse has
already left the barn.

Thanks in advance for any information,

Frank
Mobile Audio

OK, this will show you what you want to know. Good luck correcting it.

http://www.soundthoughts.co.uk/look/canon%20agc.htm

d

[Scott Dorsey]

Frank Stearns wrote:
(Scott Dorsey) writes:
Still, I am inclined to think eliminating AGC is close to impossible; once
information is lost, it's not going to come back. If there was a miracle
available, someone would use it on those Toscanini recordings....

That was my first take on things, but then there was that memory of the old DBX NR.
I now surely have a much fuller appreciation of why calibration was so important in
such systems.

The thing about the DBX system is that it was monotonic. There was a direct
1:1 mapping of input level to output level, regardless of time. It was just
a straight nonlinear function.

With an AGC, the gain of the system depends on the previous input. It's not
a simple monotonic transfer function because there are a bunch of different
possible inputs that can all create the same output. This is why I say that
information is being lost.

Interestingly, allowing only 1 or 2 dB of range on the expanders seems to help a
lot, even though one can guess that much more dynamic range than that was removed
during recording. As I set the expanders to a wider range, the reconstruction falls
apart.

Right, because the expansion you're applying isn't the same as the compression
that was there.

With the expander you should be able to undo limiting pretty well, but you
can't undo the slow gainriding. The thing is, if your knee is at a
different level or the curve is a different shape than the original limiter,
what you're getting might be sounding worse. So it becomes a matter of
turd polishing to see what you can do that will maybe not undo it but at
least make it less nasty.

I'm cautiously optimistic that if I can dial in the correct times and ratios, I can
increase the expansion range another few dB and recover a little more (and improve
the illusion); though it is probably a sure thing that full dynamic range recovery
is not possible.

You might. Also, I bet you can add some manual gain riding. Even though
the computer can't undo the slow AGC, because the computer doesn't know
what the original levels going into the AGC are, a human being might be
able to undo a little bit of it because a human being knows that the
crescendo is supposed to be louder than the harp solo.

This is one area where I am very thankful for digital dynamics processing. I have
look-ahead and very precise controls.

I hope you're getting a lot more than scale for this job.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Frank Stearns]

(Don Pearce) writes:

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

snips

OK, this will show you what you want to know. Good luck correcting it.

http://www.soundthoughts.co.uk/look/canon%20agc.htm

Thanks, Don, for taking the time to put this together.

Very interesting (and somewhat depressing), but the information is most helpful.

Frank
Mobile Audio

--

[Don Pearce[_3_]]

On Sun, 15 Sep 2013 11:44:18 -0500, Frank Stearns
wrote:

(Don Pearce) writes:

It sounds as though the camera AGC has three dynamics curves in play:

1. fast attack/fast release limiter

2. moderate attack/moderate release compressor

3. true AGC with a slower attack, exaggerated gain reduction, and an
extremely long hold time (on the order of several seconds); but when
it finally lets go, it is a moderate release.

I'll address #3 with gain lines, and the cascaded expanders help on
#1 and #2.

snips

OK, this will show you what you want to know. Good luck correcting it.

http://www.soundthoughts.co.uk/look/canon%20agc.htm

Thanks, Don, for taking the time to put this together.

Very interesting (and somewhat depressing), but the information is most helpful.

Frank
Mobile Audio

You're welcome. Any specific scenario you want to know about, just
say.

d

[Mike Rivers[_2_]]

On 9/15/2013 11:19 AM, Don Pearce wrote:

've just hit my Canon with a 440Hz tone and this is
the result with the AGC enabled. It isn't pretty - the first few
cycles are well clipped, so you may not find the results brilliant
even if you can neutralize it.

It's not that bad, actually. It figured out where it wanted to be in 10
ms and then held the level pretty stable once it got there. Chances are
the clipping was on the analog side of the processing. I suspect that
it's not a real analog gain control right at the front end. There are
some de-clipping algorithms that work fairly well that might get rid of
those four clipped cycles.

But then a steady tone isn't really all that good of a test for an AGC
or any dynamic processor. The test signal needs to be, well, dynamic.

--
For a good time, call http://mikeriversaudio.wordpress.com

[Scott Dorsey]

William Sommerwerck wrote:
"Scott Dorsey" wrote in message ...

The thing about the DBX system is that it was monotonic. There was
a direct 1:1 mapping of input level to output level, regardless of time.
It was just a straight nonlinear function.

Actually, it was linear over a very wide dynamic range, with no thresholds.
But let's not start discussing dbx NR.

If the encoding is completely linear, then you can't get any noise reduction
because any given input level becomes that same level on tape.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[[email protected]]

Best approach is to look for a tear-down, find the chip that does the agc, and then download the Datasheet.

Bob

[[email protected]]

to the OP..

don't know if this is worth the time and effort, but here is another approach.

Look for any hum or narrow band background signal maybe from the HVAC or the H sync from a monitor or something that was at a constant level at the recording session. Then you can analyze the level changes of that tone using a real time analyzer (spectrum analyzer) and deduce the gain changes.

Otherwise I'd do what you are doing, first use some fast time constant expansion and then add slow gain changes by hand.
Adjust to taste.

Mark

[Trevor]

"Scott Dorsey" wrote in message
...
William Sommerwerck wrote:
"Scott Dorsey" wrote in message ...

The thing about the DBX system is that it was monotonic. There was
a direct 1:1 mapping of input level to output level, regardless of time.
It was just a straight nonlinear function.

Actually, it was linear over a very wide dynamic range, with no
thresholds.
But let's not start discussing dbx NR.

If the encoding is completely linear, then you can't get any noise
reduction
because any given input level becomes that same level on tape.

Rubbish, linear doesn't have to be 1:1, just a straight line for input gain
Vs output gain. The early DBX units were straight line 2:1
compression/expansion and gave better noise reduction than Dolby, at the
expense of greater artifacts.

Trevor.

[Trevor]

wrote in message
...
Best approach is to look for a tear-down, find the chip that does the agc,
and then download the Datasheet.

Even if you found the chip, it probably has either external components to
adjust the time constants and compression range, or a source of digital code
to do it. Of course the data sheet would tell you that, but not necessarily
what they are in this case.

Trevor.

[Marc Wielage[_2_]]

On Sun, 15 Sep 2013 07:41:51 -0700, Scott Dorsey wrote
(in article ):

Still, I am inclined to think eliminating AGC is close to impossible; once
information is lost, it's not going to come back.
------------------------------snip------------------------------

That's my opinion as well. I'm appalled by the number of people who go out
and shoot video with DSLR's and use the internal audio connection for sound.
To me, that's a scratch track *at best*.

--MFW

[hank alrich]

Scott Dorsey wrote:

Still, I am inclined to think eliminating AGC is close to impossible; once
information is lost, it's not going to come back. If there was a miracle
available, someone would use it on those Toscanini recordings....

While it works only sometimes, Izotope's RX package includes a dynamic
restoration function. I don't think it will save this camera track, but
I also don't know what else one could try.

I used it in mastering of our album. It restored a couple of dB's worth
of range on one mixed track, very nicely, and completely failed to help
another track. There wasn't time in the circumstance to try to figure
out why.

Importantly, we were dealing with well-controlled material, and all
upstream limiting or compression came via lovely outboard hardware, not
a camera's AGC circuit.

As with many audio finessing apps, a little bit at a time, repeatedly,
begets much better results than trying to unmash it all in one go.

--
shut up and play your guitar * HankAlrich.Com
HankandShaidriMusic.Com
YouTube.Com/WalkinayMusic

[S. King]

On Wed, 18 Sep 2013 12:54:22 -0500, hank alrich wrote:

Scott Dorsey wrote:

Still, I am inclined to think eliminating AGC is close to impossible;
once information is lost, it's not going to come back. If there was a
miracle available, someone would use it on those Toscanini
recordings....

While it works only sometimes, Izotope's RX package includes a dynamic
restoration function. I don't think it will save this camera track, but
I also don't know what else one could try.

I used it in mastering of our album. It restored a couple of dB's worth
of range on one mixed track, very nicely, and completely failed to help
another track. There wasn't time in the circumstance to try to figure
out why.

Importantly, we were dealing with well-controlled material, and all
upstream limiting or compression came via lovely outboard hardware, not
a camera's AGC circuit.

As with many audio finessing apps, a little bit at a time, repeatedly,
begets much better results than trying to unmash it all in one go.

Hank. Sorry to hear about your shop accident. Hope recovery goes quickly.

Steve King