22 Dec
2005
22 Dec
'05
9:37 p.m.
Hi,
I have a question for some audio professionals out there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
Some people mentioned earlier on a previous thread that there was
something called soft gain control, where the user moves the gain up a
step, but the mixer gradually(fairly quickly) adjusts the volume to the
new level, so no clicks are heard on the speakers. How does these soft
gain controls prevent the clicking? Do they wait for the zero crossing
point to adjust the gain?
James
22 Dec
22 Dec
10 p.m.
James Courtier-Dutton <James(a)superbug.co.uk> writes:
Hi,
I have a question for some audio professionals out there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
Well, the minimum possible difference for 24bit is between -2^23+1 and
-2^23, and that would be about 0.000001dB.
Some people mentioned earlier on a previous thread
that there was
something called soft gain control, where the user moves the gain up a
step, but the mixer gradually(fairly quickly) adjusts the volume to
the new level, so no clicks are heard on the speakers. How does these
soft gain controls prevent the clicking? Do they wait for the zero
crossing point to adjust the gain?
No, why should they? You just ramp up the volume linearly. If you
want to be even nicer, you use a shape that is even smoother.
Basically, you put the step pulse from the jump gain through a low
pass filter, preferably one that does not overswing.
--
David Kastrup, Kriemhildstr. 15, 44793 Bochum
10:28 p.m.
On Fri, 2005-12-23 at 03:00 +0100, David Kastrup wrote:
James Courtier-Dutton <James(a)superbug.co.uk>
writes:
He said sensible not possible.
Lee
Hi,
I have a question for some audio professionals out there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
Well, the minimum possible difference for 24bit is between -2^23+1 and
-2^23, and that would be about 0.000001dB.
10:39 p.m.
Lee Revell <rlrevell(a)joe-job.com> writes:
On Fri, 2005-12-23 at 03:00 +0100, David Kastrup
wrote:
If you are using ramp gains, then you don't want to use courser steps
than available because you are buying additional noise in that
manner. If you wanted that noise, you'd not be using 24bit. So there
is an application where this kind of step is "sensible".
Since he does not tell what application he has in mind, one can't
choose what would be sensible for that.
--
David Kastrup, Kriemhildstr. 15, 44793 Bochum
James Courtier-Dutton
<James(a)superbug.co.uk> writes:
He said sensible not possible. Hi,
I have a question for some audio professionals out there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
Well, the minimum possible difference for 24bit is between -2^23+1 and
-2^23, and that would be about 0.000001dB. 
11:22 p.m.
On Fri, Dec 23, 2005 at 03:39:33AM +0100, David Kastrup wrote:
Lee Revell <rlrevell(a)joe-job.com> writes:
My guess is that you don't actually disagree: the O.P. didn't say
whether he meant "gain control" in the sense of user interface or in the
sense of DSP. Lee seems to be assuming the former. David seems to be
assuming the latter.
--
Paul Winkler
http://www.slinkp.com
He said sensible not possible.
If you are using ramp gains, then you don't want to use courser steps
than available because you are buying additional noise in that
manner.
11:42 p.m.
On Thu, 2005-12-22 at 22:22 -0500, Paul Winkler wrote:
On Fri, Dec 23, 2005 at 03:39:33AM +0100, David
Kastrup wrote:
Yes we are talking about user interfaces (specifically adding db gain
controls to the ALSA mixer).
Lee
Lee Revell <rlrevell(a)joe-job.com> writes:
My guess is that you don't actually disagree: the O.P. didn't say
whether he meant "gain control" in the sense of user interface or in the
sense of DSP. Lee seems to be assuming the former. David seems to be
assuming the latter.
He said sensible not possible.
If you are using ramp gains, then you don't want to use courser steps
than available because you are buying additional noise in that
manner. 
23 Dec
23 Dec
2:23 a.m.
On Fri, Dec 23, 2005 at 01:37:31AM +0000, James Courtier-Dutton wrote:
I have a question for some audio professionals out
there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
Some people mentioned earlier on a previous thread that there was
something called soft gain control, where the user moves the gain up a
step, but the mixer gradually(fairly quickly) adjusts the volume to the
new level, so no clicks are heard on the speakers. How does these soft
gain controls prevent the clicking? Do they wait for the zero crossing
point to adjust the gain?
Steps of 0.5 dB would be smalll enough from a operational POV (except
mayby in equipment used to do A/B tests, where you'd want better than
0.1 dB).
The simplest way to avoid clicks is to interpolate the change over one
ore more periods (linear is OK if the step is small). This could still
give perceptible 'zipper noise' on some signals, so one step further
would be to lowpass the gain change at period rate, then interpolate
linearly.
--
FA
5:24 a.m.
On Fri, 2005-12-23 at 01:37 +0000, James Courtier-Dutton wrote:
Hi,
I have a question for some audio professionals out there.
What is the smallest sensible gain control step in dB.
Is it 0.5dB ?
I am asking, because if one is using a digital gain control in a 24bit
fixed point DSP, once could use almost any step size, so I am looking
for the smallest sensible size to use.
This would be the resolution of your fader. So next up: What is a
sensible resolution for a fader? Expensive well renomated equipment
appear to settle for 10bit (or approximately 0.1db if you prefer.)
Some people mentioned earlier on a previous thread that there was
something called soft gain control, where the user moves the gain up a
step, but the mixer gradually(fairly quickly) adjusts the volume to the
new level, so no clicks are heard on the speakers. How does these soft
gain controls prevent the clicking? Do they wait for the zero crossing
point to adjust the gain?
Assuming 48K samplerate and a fader value (at most) every second ms
Left-shift your fader values as much as you can without loosing bits.
Interpolate the fader values over 96 samples.
Square them and average the most current 256 samples for smoothness
Multiply your signal.
James
--
mvh // Jens M Andreasen
6846
days inactive
6847
days old
linux-audio-dev@lists.linuxaudio.org
7 comments
6 participants
participants (6)
-
David Kastrup -
fons adriaensen -
James Courtier-Dutton -
Jens M Andreasen -
Lee Revell -
Paul Winkler