ultimate intelligent AGC?
Posting
Hi,
I do some volunteer recordings for a couple of local non-profit
organizations to keep a record of their activities. A small
intellectual church, for example, has a variety of sounds during
a service, including speech both at a podium mike and away from
a mike, congregation, choir and solo singing, organ, harpsicord,
cello, etc. It's a typical live recording, so I have essentially
no foreknowledge of what a piece will sound like, whether someone
will speak close to the podium mike, when and where the peaks in
a musical piece will be, etc. We currently use a normal consumer
cassette deck without AGC, and I manually adjust the gain to keep
the levels in the appropriate linear range. Of course, when an
initially quiet music piece eventualy thunders out at full blast,
I have to lower the gain on the fly. And when someone is speaking
far from a mike, I set the gain to the maximum.
Eventually these tapes are to be digitized, post-processed, and
made available as streaming audio on the web. I would, of course,
like to skip the tape digitization step, so I have been thinking
about what to substitute as a digital recorder. But I really,
really want to get away from having to hang out for essentially
the entire performance watching a level meter and adjusting the
gain manually.
Of course, there is AGC and it makes its corrections over some time
scale. Ideally I'd like some mechanism to adjust the gain on the
fly during a changing passage and then be able to correct for the
changing gain after the fact. And so it occurred to me, why not?
Certainly this can't be a new idea, so I was wondering if someone
could tell me what is the official name in the audio recording
field for digitizing both the musical signal and the gain setting
and then deconvolving to regain the constant gain musical signal?
After some time searching on on the web and Usenet, I can't find
any references to equipment with this capability. The extra
storage for the gain data is small, and I presume programmatically
controlled gain is not hard to do. I'm going to digitally process
the signals anyway, so adding a deconvolution step is trivial.
I figure this general idea has been around the audio industry
for decades, and there must be equipment that already does this.
Can anyone enlighten me?
Thanks alot!
--
A. Lester Buck buck <that at sign> compact <.> com
John Cafarella
Posting <pos...@compact.com> wrote in message
news:200206140158....@biagi.compact.com...
>
> Hi,
>
> I do some volunteer recordings for a couple of local non-profit
> organizations to keep a record of their activities. A small
<<snip>>
> and then deconvolving to regain the constant gain musical signal?
<<snip>>
> A. Lester Buck buck <that at sign> compact <.> com
>
If you ultimately want the same dynamic range as the performance, why don't
you just keep it simple.
Get a good low noise 24bit recording system, set the gain to cope with your
loudest peaks and just record. Keeping the recording level in " the
appropriate linear range" as you have put it just isn't an issue any more.
You can then process the files in post, boosting the levels of the quite
bits to suit.
I recommend you hide the gain knob after you have it set, if its going to be
unattended.
--
John Cafarella
EOR Studio
Melbourne, Australia
Scott Dorsey
>
>Eventually these tapes are to be digitized, post-processed, and
>made available as streaming audio on the web. I would, of course,
>like to skip the tape digitization step, so I have been thinking
>about what to substitute as a digital recorder. But I really,
>really want to get away from having to hang out for essentially
>the entire performance watching a level meter and adjusting the
>gain manually.
With a digital recorder you have a huge amount of dynamic range. You
can just record straight and leave the gain set, and then do all of your
processing in post. The noise floor on the good digital gear is very low
which means you can record something with extreme dynamics without having
to ride the gains.
>Of course, there is AGC and it makes its corrections over some time
>scale. Ideally I'd like some mechanism to adjust the gain on the
>fly during a changing passage and then be able to correct for the
>changing gain after the fact. And so it occurred to me, why not?
>Certainly this can't be a new idea, so I was wondering if someone
>could tell me what is the official name in the audio recording
>field for digitizing both the musical signal and the gain setting
>and then deconvolving to regain the constant gain musical signal?
Something like this was done with Perspecta in the fifties, with an
out-of-band signal that adjusted the levels of a channel. In some ways,
it's similar to some noise reduction systems. Thank God we don't have to
do junk like that any more.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Bob Cain
Scott Dorsey wrote:
>
> With a digital recorder you have a huge amount of dynamic range. You
> can just record straight and leave the gain set, and then do all of your
> processing in post. The noise floor on the good digital gear is very low
> which means you can record something with extreme dynamics without having
> to ride the gains.
I would agree with that with 24 bit converters and a 120 dB analog
channel but I run into quantization noise real quick at 16 bits without
the levels being strong.
>
> Something like this was done with Perspecta in the fifties, with an
> out-of-band signal that adjusted the levels of a channel. In some ways,
> it's similar to some noise reduction systems. Thank God we don't have to
> do junk like that any more.
I wish we did. Updating to use today's technology such a scheme would
greatly benefit 16 bit systems and allow for meaningful conversion to
higher word widths after recording as well as storage of wide dynamics
audio in narrower words. There are many ways it could be accurately
coded in the digital stream which would be inaudible. If you can
subtract as much information as is possible without serious degredation
using Meridian (none) or even ATRAC or high bitrate MP3 it should be
possible to replace a much smaller amount of the original stream with a
lookahead dynamics control stream without perceived alteration after
re-expansion. You could easily get the benefits of 24 bit dynamics
stored and transmitted as a 16 bit encoding.
Bob
--
"Things should be described as simply as possible, but no simpler."
A. Einstein
////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
To contribute your unused processor cycles to the fight against cancer:
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\///////////////////////////////////////
Scott Dorsey
>Scott Dorsey wrote:
>>
>> With a digital recorder you have a huge amount of dynamic range. You
>> can just record straight and leave the gain set, and then do all of your
>> processing in post. The noise floor on the good digital gear is very low
>> which means you can record something with extreme dynamics without having
>> to ride the gains.
>
>I would agree with that with 24 bit converters and a 120 dB analog
>channel but I run into quantization noise real quick at 16 bits without
>the levels being strong.
Huh? What kind of converters are you running?
After years of running 1/4" analogue tape, the dynamic range from 14-bit PCM
was just incredible to me... it took me a while to get over how quiet it was.
>> Something like this was done with Perspecta in the fifties, with an
>> out-of-band signal that adjusted the levels of a channel. In some ways,
>> it's similar to some noise reduction systems. Thank God we don't have to
>> do junk like that any more.
>
>I wish we did. Updating to use today's technology such a scheme would
>greatly benefit 16 bit systems and allow for meaningful conversion to
>higher word widths after recording as well as storage of wide dynamics
>audio in narrower words.
But now it's not 16 bit data any more... you have 16 bits of data plus some
more bits out of band with the scaling factor.
Of course, you can always record mono on a 2-channel recorder, with one channel
20 dB higher than the other. Doing that, you can switch between the two in
post and effectively get 20 dB more dynamic range. But now you're making an
inefficient 32-bit system.
Bob Smith
>
> Scott Dorsey wrote:
> >
> > With a digital recorder you have a huge amount of dynamic range. You
> > can just record straight and leave the gain set, and then do all of your
> > processing in post. The noise floor on the good digital gear is very low
> > which means you can record something with extreme dynamics without having
> > to ride the gains.
>
> I would agree with that with 24 bit converters and a 120 dB analog
> channel but I run into quantization noise real quick at 16 bits without
> the levels being strong.
What kind of converters are you using that you run into perceptable
quantization noise so quickly?
I've got a remote recording of a jazz choir in front of me at this
moment that has been recorded 18 bit A/D into 16 bit storage, -5 dB FS
peaks for the music, and the HVAC noise before and after which is at or
below -55 dB FS sounds pretty faithful with a whopping 8 bit
representation (least significant 8 bits of the 16 bit storage for those
signals below -48 dB FS).
--
bobs
we organize chaos
Bob Smith - BS Studios
http://www.bsstudios.com/
rsm...@bsstudios.com
Bob Cain
Scott Dorsey wrote:
>
> >I would agree with that with 24 bit converters and a 120 dB analog
> >channel but I run into quantization noise real quick at 16 bits without
> >the levels being strong.
>
> Huh? What kind of converters are you running?
16 bit Sony PCM-M1 DAT.
>
> After years of running 1/4" analogue tape, the dynamic range from 14-bit PCM
> was just incredible to me... it took me a while to get over how quiet it was.
But to get from 16 bits to 12 you only need to be 24 dB down. It is not
at all unusual in live performance to have peaks that are that much
above the mean level and the solution is to clip those peaks or
accomodate them at the expense of 12 bit noise over the meat of the
performance. A limiter in the chain would solve that but I'd rather
have the full flexibility of dealing with it later during the
engineering.
Orchestral music can have passages that are at least that far down in
the mean over the mean of the louder passages. I cannot stand to
constantly be turning the volume up or down on a recording to accomodate
those wide ranges so use volume envelopes to bring them closer
together. A purist I am not in that regard. The original dynamics may
be great in a hall where that is all that is happening but not in the
most common environments recordings are listened to. In accomodating
that I can often hear noise in the softest passages. I suppose the bulk
of that could be mic self noise, though, and wider digital dynamic range
ain't gonna help that.
Whatever justification is used for 24 bit recording to begin with can be
applied to what I have suggested for approaching those kinds of dynamics
within 16 bits. But 24 bit is where it is at now and we've already been
persuaded to make room for it so the point of packing more information
into 16 bits than straight PCM has become moot I guess.
>
> But now it's not 16 bit data any more... you have 16 bits of data plus some
> more bits out of band with the scaling factor.
Not in the kind of scheme I'm thinking of. I want to see a system where
audio data compression, which can be truly inaudible and possibly even
lossless at the modest levels I think are required, provides room within
the band for the dynamics information rather than having it out of band
and stored as an adjunct. Remove redundant information from the stream
and replace it with information that enhances the dynamic range such
that the resulting data rate and storage requirements are the same.
Meridian provides a lossless encoding that halves the data rate.
Imagine using that and replacing the redundant half with dynamics info.
There is a bit more to it than that as I envision it but that's the
basic idea. An idea whose time, it seems, has passed.
Scott Dorsey
>Scott Dorsey wrote:
>>
>> >I would agree with that with 24 bit converters and a 120 dB analog
>> >channel but I run into quantization noise real quick at 16 bits without
>> >the levels being strong.
>>
>> Huh? What kind of converters are you running?
>
>16 bit Sony PCM-M1 DAT.
That would do it.
Don't judge all 16-bit systems by that standard, please. The noise on the
M-1 is mostly from the analogue side.