Hi, I know pd operates in 32bit precision, but that is for all numbers out there, which can be data control or audio streams. Now, output audio values is actually just from -1 to 1, so we can't say that the audio output in Pd is 32 bits.
So I wonder what is the actual audio bit resolution that we have in Pd.
Cheers
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1 and
See http://en.m.wikipedia.org/wiki/Floating_point best,
Jamie
-- http://jamiebullock.com On Wed, Apr 22, 2015 at 6:26 pm, Alexandre Torres Porres porres@gmail.com wrote: Hi, I know pd operates in 32bit precision, but that is for all numbers out there, which can be data control or audio streams. Now, output audio values is actually just from -1 to 1, so we can't say that the audio output in Pd is 32 bits. So I wonder what is the actual audio bit resolution that we have in Pd. Cheers
but how come, since you can also get audio stream with values over the boundaries of -1 & 1?
cheers
2015-04-22 15:04 GMT-03:00 Jamie Bullock jamie@jamiebullock.com:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1 and 1. See http://en.m.wikipedia.org/wiki/Floating_point
best,
Jamie
On Wed, Apr 22, 2015 at 6:26 pm, Alexandre Torres Porres porres@gmail.com wrote:
Hi, I know pd operates in 32bit precision, but that is for all numbers out there, which can be data control or audio streams. Now, output audio values is actually just from -1 to 1, so we can't say that the audio output in Pd is 32 bits.
So I wonder what is the actual audio bit resolution that we have in Pd.
Cheers
Internally you can haves values over the boundaries but it's likely that Pd and the operating system will clip the signal to [-1 ... 1] and/or scale it before sending it to the driver.
On 22 April 2015 at 19:12, Alexandre Torres Porres porres@gmail.com wrote:
but how come, since you can also get audio stream with values over the boundaries of -1 & 1?
cheers
2015-04-22 15:04 GMT-03:00 Jamie Bullock jamie@jamiebullock.com:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1 and 1. See http://en.m.wikipedia.org/wiki/Floating_point
best,
Jamie
On Wed, Apr 22, 2015 at 6:26 pm, Alexandre Torres Porres < porres@gmail.com> wrote:
Hi, I know pd operates in 32bit precision, but that is for all numbers out there, which can be data control or audio streams. Now, output audio values is actually just from -1 to 1, so we can't say that the audio output in Pd is 32 bits.
So I wonder what is the actual audio bit resolution that we have in Pd.
Cheers
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
On Wednesday, April 22, 2015, Jamie Bullock jamie@jamiebullock.com wrote:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1
and 1.
I don't think that's right.
The range of a single precision floating point number is from
-3.4028234 × 10E38 to 3.4028234 × 10E38 (not from -1 to 1)
There are only 23 bits of precision for the mantissa + 1 for sign in a single precision float.
On 22 April 2015 at 19:44:26, William Huston (williamahuston@gmail.com) wrote:
On Wednesday, April 22, 2015, Jamie Bullock jamie@jamiebullock.com wrote:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1 and 1.
I don't think that's right.
The range of a single precision floating point number is from
-3.4028234 × 10E38 to 3.4028234 × 10E38 (not from -1 to 1)
True, but I didn’t say the range of 32-bit float was -1 to 1!
There are only 23 bits of precision for the mantissa + 1 for sign in a single precision float.
Also true, but when I said “resolution” I didn’t mean “precision”. Because the exponent can be negative, resolution scales dynamically from 1..0 according to the value of the exponent, whilst precision stays fixed according to the number of bits in the mantissa. Thus for very small values the resolution (or quantisation step size) is far finer than can be represented with the mantissa alone.
What I was trying to put across (poorly!) in my original reply is that unlike fixed point where for lower order values fewer bits are available in the binary representation, with floating point, just because e.g. -1..1 is a smaller range than -3.4 x 10E38..3.4 x 10E38 it doesn’t imply “fewer are bits available”, e.g.
Sign Exponent Mantissa 0 01111110 11111111111111111111111 -> 0.99999994 0 00000001 11111111111111111111111 -> 2.3509886E-38 1 01000000 0000000000000000000000 -> -1.0842022E-19 1 011111110 0000000000000000000000 -> -1.0
Strictly speaking, I guess only 31 bits “count” in the range -1..1 due to a maximum of 7-bits being significant in the exponent.
best,
Jamie
WOW... I just learned something important! So, my whole point here was that I had the idea that DAWs like Ardour support 32-bit considering only values from -1 to 1. But that is just wrong!
I just learned you can put a sound file with values in the hundreds / thousands in 32 bit float, load them into your DAW, and scale it down.
I tried it by creating a sine wave in Pd with values from -100 to 100, exported as 32-bit float with writesf~, loaded into soundforge, then scaled it down 40 dB, and the sine wave was there!
So yeah, Pd's audio resolution is the same as DAW which say they handle 32-bit float sound files. My whole issue was that Pd had a different way of dealing with 32-bit float, but not at all. In Principle, other softwares out there also deal with 32-bit float outside the boundaries of -1 to 1!!!
That just answers my question then... once and for all and for good.
I guess the discussion I ended up promoting is a parallel issue... let me rephrase it then.
Regarding 24 bit DAC converters in sound cards, the 24 bits in there are just for values from -1 to 1, right?
If so, then 32 bit float isn't really "8 more bits". And you've been also saying 24 bit converters are fixed, not float. So there's a weird relationship between this conversion from 32 bit float files to the soundcard.
But then, I guess I'm happy with all I've learned so far.
thanks
2015-04-23 13:26 GMT-03:00 Jamie Bullock jamie@jamiebullock.com:
On 22 April 2015 at 19:44:26, William Huston (williamahuston@gmail.com) wrote:
On Wednesday, April 22, 2015, Jamie Bullock jamie@jamiebullock.com wrote:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1
and 1.
I don't think that's right.
The range of a single precision floating point number is from
-3.4028234 × 10E38 to 3.4028234 × 10E38 (not from -1 to 1)
True, but I didn’t say the range of 32-bit float was -1 to 1!
There are only 23 bits of precision for the mantissa + 1 for sign in a single precision float.
Also true, but when I said “resolution” I didn’t mean “precision”. Because the exponent can be negative, resolution scales dynamically from 1..0 according to the value of the exponent, whilst precision stays fixed according to the number of bits in the mantissa. Thus for very small values the resolution (or quantisation step size) is far finer than can be represented with the mantissa alone.
What I was trying to put across (poorly!) in my original reply is that unlike fixed point where for lower order values fewer bits are available in the binary representation, with floating point, just because e.g. -1..1 is a smaller range than -3.4 x 10E38..3.4 x 10E38 it doesn’t imply “fewer are bits available”, e.g.
Sign Exponent Mantissa 0 01111110 11111111111111111111111 -> 0.99999994 0 00000001 11111111111111111111111 -> 2.3509886E-38 1 01000000 0000000000000000000000 -> -1.0842022E-19 1 011111110 0000000000000000000000 -> -1.0
Strictly speaking, I guess only 31 bits “count” in the range -1..1 due to a maximum of 7-bits being significant in the exponent.
best,
Jamie
According to Wikipedia
"as of 2007 digital audio converter technology is limited to a SNR of about 124 dB (21-bit) because of real-world limitations in integrated circuit design. Still, this approximately matches the performance of the human auditory system"
http://en.wikipedia.org/wiki/Audio_bit_depth#Floating_point
So, yeah, apparently there's no circuitry technology to go anywhere near 32 bit, so a 32 bit DAC seems weird and pointless...
I guess a good way to approach this issue I raised is to be able to define what is the internal dynamic range in Pd for values between -1 and 1.
cheers
2015-04-23 15:02 GMT-03:00 Alexandre Torres Porres porres@gmail.com:
WOW... I just learned something important! So, my whole point here was that I had the idea that DAWs like Ardour support 32-bit considering only values from -1 to 1. But that is just wrong!
I just learned you can put a sound file with values in the hundreds / thousands in 32 bit float, load them into your DAW, and scale it down.
I tried it by creating a sine wave in Pd with values from -100 to 100, exported as 32-bit float with writesf~, loaded into soundforge, then scaled it down 40 dB, and the sine wave was there!
So yeah, Pd's audio resolution is the same as DAW which say they handle 32-bit float sound files. My whole issue was that Pd had a different way of dealing with 32-bit float, but not at all. In Principle, other softwares out there also deal with 32-bit float outside the boundaries of -1 to 1!!!
That just answers my question then... once and for all and for good.
I guess the discussion I ended up promoting is a parallel issue... let me rephrase it then.
Regarding 24 bit DAC converters in sound cards, the 24 bits in there are just for values from -1 to 1, right?
If so, then 32 bit float isn't really "8 more bits". And you've been also saying 24 bit converters are fixed, not float. So there's a weird relationship between this conversion from 32 bit float files to the soundcard.
But then, I guess I'm happy with all I've learned so far.
thanks
2015-04-23 13:26 GMT-03:00 Jamie Bullock jamie@jamiebullock.com:
On 22 April 2015 at 19:44:26, William Huston (williamahuston@gmail.com) wrote:
On Wednesday, April 22, 2015, Jamie Bullock jamie@jamiebullock.com wrote:
Pd is 32-bit *floating point*, so you have 32-bit resolution between -1
and 1.
I don't think that's right.
The range of a single precision floating point number is from
-3.4028234 × 10E38 to 3.4028234 × 10E38 (not from -1 to 1)
True, but I didn’t say the range of 32-bit float was -1 to 1!
There are only 23 bits of precision for the mantissa + 1 for sign in a single precision float.
Also true, but when I said “resolution” I didn’t mean “precision”. Because the exponent can be negative, resolution scales dynamically from 1..0 according to the value of the exponent, whilst precision stays fixed according to the number of bits in the mantissa. Thus for very small values the resolution (or quantisation step size) is far finer than can be represented with the mantissa alone.
What I was trying to put across (poorly!) in my original reply is that unlike fixed point where for lower order values fewer bits are available in the binary representation, with floating point, just because e.g. -1..1 is a smaller range than -3.4 x 10E38..3.4 x 10E38 it doesn’t imply “fewer are bits available”, e.g.
Sign Exponent Mantissa 0 01111110 11111111111111111111111 -> 0.99999994 0 00000001 11111111111111111111111 -> 2.3509886E-38 1 01000000 0000000000000000000000 -> -1.0842022E-19 1 011111110 0000000000000000000000 -> -1.0
Strictly speaking, I guess only 31 bits “count” in the range -1..1 due to a maximum of 7-bits being significant in the exponent.
best,
Jamie
On 23/04/2015 20:02, Alexandre Torres Porres wrote: ...
If so, then 32 bit float isn't really "8 more bits". And you've been also saying 24 bit converters are fixed, not float.
My soundcard has a longer word* than yours :-P
On 04/22/2015 07:26 PM, Alexandre Torres Porres wrote:
Now, output audio values is actually just from -1 to 1,
where did you get that from? there is no -1/+1 in "audio".
instead "audio" will have many different values (and units), including:
- sound-pressure in the air (hmm, hard to make any actual numbers)
- voltage at your sound-card output (+- 1V, depending on the line level
in use)
- digital values fed to your soundcard (depending on the the bit
resolution of your soundcard, this might be +-32768 (16bit) or +-8388608 (24bit))
- digital values output by Pd (many audio backends (e.g. jack) will use
floating point as their intermediate format)
now your question really only makes sense when you consider fixed-point values such as used by your typical soundcard (where you divide a range (e.g. +-1) into N equally small fractions (e.g. 1/32678 steps)
but Pd uses floating point arithmetic, which does not have equal fractions but fractions depending on the current range of the value. so the quantization steps are rather largish when it comes to bigger numbers, but can be *very* small when it comes to small numbers.
IEEE-754 (which defines the exact format of single/double precision floating point values as used by most systems) uses 24bits for the significand, which will correspond to 24bit fixed point around the values +-1., but will already correspond to 25bit fixed point around the values +-0.5 and so on.
gfmsrd IOhannes
"*IEEE-754 (which defines the exact format of single/double precision*
*floating point values as used by most systems) uses 24bits for thesignificand, which will correspond to 24bit fixed point around thevalues +-1., but will already correspond to 25bit fixed point around the**values +-0.5 and so on.*"
I'm not sure if I got this, but this is the kind of information I'm after.
It's clear to me that the resolution for audio streams sent from pd (with values between -1 and 1) can't be considered as "32 bits". My guess is that it'd be more like 24 bit precision or around that, considering most top notch audio cards only deal with 24 bit audio.
Does your answer state that audio in Pd is 24 bits?
Obviously, the actual resolution depends on the audio card. Which also makes me wonder how does the downscale/bit reduction work if, say, it's only a cheap 8 bit audio card, or 16 bits. I assume it's up to the audio card and audio driver, not Pd at all.
Cheers
2015-04-22 15:30 GMT-03:00 IOhannes m zmölnig zmoelnig@iem.at:
On 04/22/2015 07:26 PM, Alexandre Torres Porres wrote:
Now, output audio values is actually just from -1 to 1,
where did you get that from? there is no -1/+1 in "audio".
instead "audio" will have many different values (and units), including:
- sound-pressure in the air (hmm, hard to make any actual numbers)
- voltage at your sound-card output (+- 1V, depending on the line level
in use)
- digital values fed to your soundcard (depending on the the bit
resolution of your soundcard, this might be +-32768 (16bit) or +-8388608 (24bit))
- digital values output by Pd (many audio backends (e.g. jack) will use
floating point as their intermediate format)
now your question really only makes sense when you consider fixed-point values such as used by your typical soundcard (where you divide a range (e.g. +-1) into N equally small fractions (e.g. 1/32678 steps)
but Pd uses floating point arithmetic, which does not have equal fractions but fractions depending on the current range of the value. so the quantization steps are rather largish when it comes to bigger numbers, but can be *very* small when it comes to small numbers.
IEEE-754 (which defines the exact format of single/double precision floating point values as used by most systems) uses 24bits for the significand, which will correspond to 24bit fixed point around the values +-1., but will already correspond to 25bit fixed point around the values +-0.5 and so on.
gfmsrd IOhannes
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
On 04/22/2015 10:10 PM, Alexandre Torres Porres wrote:
[...] but this is the kind of information I'm after.
i know, but this doesn't mean that your answer makes (much) sense¹.
It's clear to me that the resolution for audio streams sent from pd (with values between -1 and 1) can't be considered as "32 bits".
which is just wrong: Pd's audio samples are 32bit floating point. as such they must be considered as "32bit floating point" which is a) "32 bit" and b) something else as "16bit PCM" or "8bit ulaw" or what you call "32 bits" (which is likely "32bit fixed point").
My guess is that it'd be more like 24 bit precision or around that, considering most top notch audio cards only deal with 24 bit audio.
please don't. when i started with Pd it used 32bit floating point values for audio processing, and "most top notch audio cards" had 16 bit dac. this is why Pd was then announced as doing "full CD quality audio", which i always found very unfortunate, as Pd was doing way better than just boring 16bit.
the bit resolution of your audio hardware (top notch, or low end, or whatever) has *absolutely nothing* to do with the bit resolution of Pd. your hearing experience might be directly related to the quality of your soundcard, though.
Does your answer state that audio in Pd is 24 bits?
no, it means that audio in Pd is "32bit floating point" (all right; i think i made my point; i think i'm not very good at explaining the difference between floating point and fixed point and what that means to bit resolution; in any case you *should* read about floating point if you are interested in those kind of things).
Obviously, the actual resolution depends on the audio card.
depends what you mean by "actual". but yes, what comes out of your soundcard heavily depends on your soundcard.
Which also makes me wonder how does the downscale/bit reduction work if, say, it's only a cheap 8 bit audio card, or 16 bits. I assume it's up to the audio card and audio driver, not Pd at all.
yes. luckily Pd is blissfully unaware of audio cards used. it only queries the audio backend about the name, number of available channels and samplerate and then passes the samples to the audio backend which is free to do whatever with the samples, but somehow takes care that they end up correctly in the sound cards DAC)
fgmads IOhannes
It's clear to me that the resolution for audio streams sent from pd (with values between -1 and 1) can't be considered as "32 bits".
which is just wrong: Pd's audio samples are 32bit floating point. as such they must be considered as "32bit floating point" which is a) "32 bit" and b) something else as "16bit PCM" or "8bit ulaw" or what you call "32 bits" (which is likely "32bit fixed point").
Maybe I'm not being clear, let me try again. When we talk about "audio in Pd" we can mean two things:
One is that audio signals in Pd can be just about any number according to the bit resolution, which is 32-bit float. Cool. I don't remember the limits of this, but it's just quite a lot.
Another thing is the actual audio that comes out , and the restrains are numbers from -1 to 1! So, What goes to the soundcard is not all the possible values that you can get with 32-bit float, and just "less" of that, which is whatever you can get between -1 and 1 in 32-bit float.
So I start with this idea that the audio (values from -1 to 1) can't be in full 32 bit float resolution, it's less. I don't see why that is "wrong". And then, from it, my first question here was: "what is the audio resolution then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Cheers
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres porres@gmail.com wrote:
So I start with this idea that the audio (values from -1 to 1) can't be in full 32 bit float resolution, it's less. I don't see why that is "wrong". And then, from it, my first question here was: "what is the audio resolution then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Let me try:
32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller.
However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision.
The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system.
So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC).
Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1.
Chuck
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry czhenry@gmail.com wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres porres@gmail.com wrote:
So I start with this idea that the audio (values from -1 to 1) can't be
in
full 32 bit float resolution, it's less. I don't see why that is "wrong". And then, from it, my first question here was: "what is the audio
resolution
then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Let me try:
32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller.
However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision.
The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system.
So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC).
Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1.
Chuck
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit precision for many years to come, so it's just way more than enough.
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks jbeezez@gmail.com:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry czhenry@gmail.com wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres porres@gmail.com wrote:
So I start with this idea that the audio (values from -1 to 1) can't be
in
full 32 bit float resolution, it's less. I don't see why that is
"wrong".
And then, from it, my first question here was: "what is the audio
resolution
then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Let me try:
32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller.
However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision.
The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system.
So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC).
Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1.
Chuck
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Le 23/04/2015 16:41, Alexandre Torres Porres a écrit :
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit precision for many years to come, so it's just way more than enough.
The human ear is usually consider to be sensible from 0dB to 120dB, so a range of 10^(12/2) between the smallest and biggest amplitude. i.e from 1 to 1 000 000, or from 1 to 2^13.8 so, the human ear sensitivity can be considered to be about 14 bits. 16 bits diffusion should be enough. 24 bits diffusion is already overkill.
cheers c
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks <jbeezez@gmail.com mailto:jbeezez@gmail.com>:
Nice. Thanks Chuck, I learnt something. On 22 April 2015 at 23:45, Charles Z Henry <czhenry@gmail.com <mailto:czhenry@gmail.com>> wrote: On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres <porres@gmail.com <mailto:porres@gmail.com>> wrote: > So I start with this idea that the audio (values from -1 to 1) can't be in > full 32 bit float resolution, it's less. I don't see why that is "wrong". > And then, from it, my first question here was: "what is the audio resolution > then?". I'm still clueless here about this answer. > > Moreover, is it more or less than what 24 bit audio cards handle? Let me try: 32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller. However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision. The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system. So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC). Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1. Chuck _______________________________________________ Pd-list@lists.iem.at <mailto:Pd-list@lists.iem.at> mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
I get 1 000 000 = 2^19.9 so a 20 bit dynamic range.
I don't think A/D/A hardware ever gets better than about 110 dB dtnamic range though.
cheers Miller
On Thu, Apr 23, 2015 at 05:20:51PM +0200, Cyrille Henry wrote:
Le 23/04/2015 16:41, Alexandre Torres Porres a écrit :
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit precision for many years to come, so it's just way more than enough.
The human ear is usually consider to be sensible from 0dB to 120dB, so a range of 10^(12/2) between the smallest and biggest amplitude. i.e from 1 to 1 000 000, or from 1 to 2^13.8 so, the human ear sensitivity can be considered to be about 14 bits. 16 bits diffusion should be enough. 24 bits diffusion is already overkill.
cheers c
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks <jbeezez@gmail.com mailto:jbeezez@gmail.com>:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry <czhenry@gmail.com mailto:czhenry@gmail.com> wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres <porres@gmail.com <mailto:porres@gmail.com>> wrote: > So I start with this idea that the audio (values from -1 to 1) can't be in > full 32 bit float resolution, it's less. I don't see why that is "wrong". > And then, from it, my first question here was: "what is the audio resolution > then?". I'm still clueless here about this answer. > > Moreover, is it more or less than what 24 bit audio cards handle? Let me try: 32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller. However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision. The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system. So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC). Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1. Chuck _______________________________________________ Pd-list@lists.iem.at <mailto:Pd-list@lists.iem.at> mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
I know there's a parallel discussion about human hearing and what audio cards can get there. But I'm just trying to get one simple fact clear, that is the bit depth of audio in Pd :)
Leaving the human hearing or audio cards aside, some DAW (like Pro Tools or Ardour) do operate on audio files that are actual 32 bit resolution. I guess the idea is to keep quantization error as low as possible when mixing, normalizing, processing, filtering, mastering and everything. Then you can convert it to, say, 24 bit high quality audio afterwards for distribution - I guess this is the standard for highest digital audio these days, meaning that it's pointless to have a final audio that's higher than that, but then, 32 bit dac seem to be showing up already as chuck pointed out, but I digress. Moreover, you can also convert it to 16 bit CD quality afterwards, or even just make some MP3 or whatever...
Not to get into the discussion if dealing with 32 bits internally on a DAW is really important or worth the hassle (and not even getting into the deal with new 32 bit dac), the fact is that 32 bit audio exists out there for some time now. They do have this 32 bits option, or even more maybe... (not sure if they're pushing it to 64 yet, but it doesn't matter).
So, I always knew Pd was "32 bit", and knowing the above, I was misled to think Pd was just like Pro Tools or Ardour, that it could process audio in 32 bit. But I was thinking about it these days and it hit me that you just can't say Pd processes audio like Ardour and Pro Tools do on 32 bit precision.
Bottom line, you can't!
So this made me wonder what the heck that precision would be...
Chuck made me think it was a bit more than 24 bits, now it seems Miller says it's more likely to be 20 bits :)
cheers
2015-04-23 12:25 GMT-03:00 Miller Puckette msp@ucsd.edu:
I get 1 000 000 = 2^19.9 so a 20 bit dynamic range.
I don't think A/D/A hardware ever gets better than about 110 dB dtnamic range though.
cheers Miller
On Thu, Apr 23, 2015 at 05:20:51PM +0200, Cyrille Henry wrote:
Le 23/04/2015 16:41, Alexandre Torres Porres a écrit :
Yep, nice indeed, I guess I learned - in short and in layman's
undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit
precision for many years to come, so it's just way more than enough.
The human ear is usually consider to be sensible from 0dB to 120dB, so a
range of 10^(12/2) between the smallest and biggest amplitude.
i.e from 1 to 1 000 000, or from 1 to 2^13.8 so, the human ear sensitivity can be considered to be about 14 bits. 16 bits diffusion should be enough. 24 bits diffusion is already overkill.
cheers c
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks <jbeezez@gmail.com <mailto:
jbeezez@gmail.com>>:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry <czhenry@gmail.com
mailto:czhenry@gmail.com> wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres <porres@gmail.com <mailto:porres@gmail.com>> wrote: > So I start with this idea that the audio (values from -1 to
- can't be in
> full 32 bit float resolution, it's less. I don't see why thatis "wrong".
> And then, from it, my first question here was: "what is theaudio resolution
> then?". I'm still clueless here about this answer. > > Moreover, is it more or less than what 24 bit audio cardshandle?
Let me try: 32-bit floating point numbers have 24 bits of precision.Always. The
remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precisionin
floating-point. The value of the least significant bit (LSB)gets
proportionally smaller. However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smallernumbers have
less precision. The mismatch occurs when converting from the 32-bit floats tothe
24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the24-bit
fixed point system. So, yes, the resolution (of small numbers) in floating point(internal
to Pd) is finer than the resolution of those numbers when output (driver/DAC). Also, the 24-bit fixed point format is for values between -1and 1.
That means that numbers between 0 and 1 have just 23 bits. In32-bit
math, the numbers between 0.5 and 1 still have 24 bits ofprecision
(the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for allnumbers
between -1 and 1. Chuck _______________________________________________ Pd-list@lists.iem.at <mailto:Pd-list@lists.iem.at> mailing list UNSUBSCRIBE and account-management ->http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management ->
http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management ->
http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Chuck made me think it was a bit more than 24 bits, now it seems Miller says it's more likely to be 20 bits :)
or I just got that way wrong and miller wasn't talking about this at all... so I stick to chuck's answer.
cheers
2015-04-23 12:55 GMT-03:00 Alexandre Torres Porres porres@gmail.com:
I know there's a parallel discussion about human hearing and what audio cards can get there. But I'm just trying to get one simple fact clear, that is the bit depth of audio in Pd :)
Leaving the human hearing or audio cards aside, some DAW (like Pro Tools or Ardour) do operate on audio files that are actual 32 bit resolution. I guess the idea is to keep quantization error as low as possible when mixing, normalizing, processing, filtering, mastering and everything. Then you can convert it to, say, 24 bit high quality audio afterwards for distribution - I guess this is the standard for highest digital audio these days, meaning that it's pointless to have a final audio that's higher than that, but then, 32 bit dac seem to be showing up already as chuck pointed out, but I digress. Moreover, you can also convert it to 16 bit CD quality afterwards, or even just make some MP3 or whatever...
Not to get into the discussion if dealing with 32 bits internally on a DAW is really important or worth the hassle (and not even getting into the deal with new 32 bit dac), the fact is that 32 bit audio exists out there for some time now. They do have this 32 bits option, or even more maybe... (not sure if they're pushing it to 64 yet, but it doesn't matter).
So, I always knew Pd was "32 bit", and knowing the above, I was misled to think Pd was just like Pro Tools or Ardour, that it could process audio in 32 bit. But I was thinking about it these days and it hit me that you just can't say Pd processes audio like Ardour and Pro Tools do on 32 bit precision.
Bottom line, you can't!
So this made me wonder what the heck that precision would be...
Chuck made me think it was a bit more than 24 bits, now it seems Miller says it's more likely to be 20 bits :)
cheers
2015-04-23 12:25 GMT-03:00 Miller Puckette msp@ucsd.edu:
I get 1 000 000 = 2^19.9 so a 20 bit dynamic range.
I don't think A/D/A hardware ever gets better than about 110 dB dtnamic range though.
cheers Miller
On Thu, Apr 23, 2015 at 05:20:51PM +0200, Cyrille Henry wrote:
Le 23/04/2015 16:41, Alexandre Torres Porres a écrit :
Yep, nice indeed, I guess I learned - in short and in layman's
undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit
precision for many years to come, so it's just way more than enough.
The human ear is usually consider to be sensible from 0dB to 120dB, so
a range of 10^(12/2) between the smallest and biggest amplitude.
i.e from 1 to 1 000 000, or from 1 to 2^13.8 so, the human ear sensitivity can be considered to be about 14 bits. 16 bits diffusion should be enough. 24 bits diffusion is already overkill.
cheers c
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks <jbeezez@gmail.com <mailto:
jbeezez@gmail.com>>:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry <czhenry@gmail.com
mailto:czhenry@gmail.com> wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres <porres@gmail.com <mailto:porres@gmail.com>> wrote: > So I start with this idea that the audio (values from -1 to
- can't be in
> full 32 bit float resolution, it's less. I don't see whythat is "wrong".
> And then, from it, my first question here was: "what is theaudio resolution
> then?". I'm still clueless here about this answer. > > Moreover, is it more or less than what 24 bit audio cardshandle?
Let me try: 32-bit floating point numbers have 24 bits of precision.Always. The
remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose anyprecision in
floating-point. The value of the least significant bit (LSB)gets
proportionally smaller. However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smallernumbers have
less precision. The mismatch occurs when converting from the 32-bit floats tothe
24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number inthe 24-bit
fixed point system. So, yes, the resolution (of small numbers) in floating point(internal
to Pd) is finer than the resolution of those numbers whenoutput
(driver/DAC). Also, the 24-bit fixed point format is for values between -1and 1.
That means that numbers between 0 and 1 have just 23 bits. In32-bit
math, the numbers between 0.5 and 1 still have 24 bits ofprecision
(the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for allnumbers
between -1 and 1. Chuck _______________________________________________ Pd-list@lists.iem.at <mailto:Pd-list@lists.iem.at> mailinglist
UNSUBSCRIBE and account-management ->http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management ->
http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management ->
http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Le 23/04/2015 17:25, Miller Puckette a écrit :
I get 1 000 000 = 2^19.9 so a 20 bit dynamic range.
yes, your right, thanks for the correction c
I don't think A/D/A hardware ever gets better than about 110 dB dtnamic range though.
cheers Miller
On Thu, Apr 23, 2015 at 05:20:51PM +0200, Cyrille Henry wrote:
Le 23/04/2015 16:41, Alexandre Torres Porres a écrit :
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit precision for many years to come, so it's just way more than enough.
The human ear is usually consider to be sensible from 0dB to 120dB, so a range of 10^(12/2) between the smallest and biggest amplitude. i.e from 1 to 1 000 000, or from 1 to 2^13.8 so, the human ear sensitivity can be considered to be about 14 bits. 16 bits diffusion should be enough. 24 bits diffusion is already overkill.
cheers c
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks <jbeezez@gmail.com mailto:jbeezez@gmail.com>:
Nice. Thanks Chuck, I learnt something. On 22 April 2015 at 23:45, Charles Z Henry <czhenry@gmail.com <mailto:czhenry@gmail.com>> wrote: On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres <porres@gmail.com <mailto:porres@gmail.com>> wrote: > So I start with this idea that the audio (values from -1 to 1) can't be in > full 32 bit float resolution, it's less. I don't see why that is "wrong". > And then, from it, my first question here was: "what is the audio resolution > then?". I'm still clueless here about this answer. > > Moreover, is it more or less than what 24 bit audio cards handle? Let me try: 32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller. However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision. The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system. So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC). Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1. Chuck _______________________________________________ Pd-list@lists.iem.at <mailto:Pd-list@lists.iem.at> mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
It's already started.... 32bit DACs are available from AKM and XMOS, for example. Although I don't know what software/hardware platform you'd use to actually make use of this precision, you can build your own 32-bit sound playback interface with a few boards from HK: http://www.yuan-jing.com/dacs-decoder/32bit-192khz-usb-dac-decoder-ak4399-wm... http://www.yuan-jing.com/dacs-decoder/xmos-usb-audio-32bit-384khz-dac-decode...
On Thu, Apr 23, 2015 at 9:41 AM, Alexandre Torres Porres porres@gmail.com wrote:
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit precision for many years to come, so it's just way more than enough.
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks jbeezez@gmail.com:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry czhenry@gmail.com wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres porres@gmail.com wrote:
So I start with this idea that the audio (values from -1 to 1) can't be in full 32 bit float resolution, it's less. I don't see why that is "wrong". And then, from it, my first question here was: "what is the audio resolution then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Let me try:
32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller.
However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision.
The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system.
So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC).
Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1.
Chuck
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
What? 32Bit 384KHz dac? And for 50 bucks? what the $*#%@?
2015-04-23 12:43 GMT-03:00 Charles Z Henry czhenry@gmail.com:
It's already started.... 32bit DACs are available from AKM and XMOS, for example. Although I don't know what software/hardware platform you'd use to actually make use of this precision, you can build your own 32-bit sound playback interface with a few boards from HK:
http://www.yuan-jing.com/dacs-decoder/32bit-192khz-usb-dac-decoder-ak4399-wm...
http://www.yuan-jing.com/dacs-decoder/xmos-usb-audio-32bit-384khz-dac-decode...
On Thu, Apr 23, 2015 at 9:41 AM, Alexandre Torres Porres porres@gmail.com wrote:
Yep, nice indeed, I guess I learned - in short and in layman's undetailed terms - that audio output is ~24bits (a bit higher, but much higher for smaller numbers).
Moreover, digital audio cards won't likely have more than 24 bit
precision
for many years to come, so it's just way more than enough.
thanks
2015-04-23 6:43 GMT-03:00 Julian Brooks jbeezez@gmail.com:
Nice. Thanks Chuck, I learnt something.
On 22 April 2015 at 23:45, Charles Z Henry czhenry@gmail.com wrote:
On Wed, Apr 22, 2015 at 5:11 PM, Alexandre Torres Porres porres@gmail.com wrote:
So I start with this idea that the audio (values from -1 to 1) can't
be
in full 32 bit float resolution, it's less. I don't see why that is "wrong". And then, from it, my first question here was: "what is the audio resolution then?". I'm still clueless here about this answer.
Moreover, is it more or less than what 24 bit audio cards handle?
Let me try:
32-bit floating point numbers have 24 bits of precision. Always. The remaining 8 bits are just for the sign and exponent. When the amplitude of the signals decrease, you don't lose any precision in floating-point. The value of the least significant bit (LSB) gets proportionally smaller.
However, the output of a 24-bit soundcard always has a fixed quantization. The LSB is always the same size. Smaller numbers have less precision.
The mismatch occurs when converting from the 32-bit floats to the 24-bit fixed point numbers. Now, the smaller numbers aren't as precise anymore. They get rounded to the nearest number in the 24-bit fixed point system.
So, yes, the resolution (of small numbers) in floating point (internal to Pd) is finer than the resolution of those numbers when output (driver/DAC).
Also, the 24-bit fixed point format is for values between -1 and 1. That means that numbers between 0 and 1 have just 23 bits. In 32-bit math, the numbers between 0.5 and 1 still have 24 bits of precision (the sign is held elsewhere). That means that Pd's internal resolution is finer than the soundcard resolution for all numbers between -1 and 1.
Chuck
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
On 24/04/15 01:52, Alexandre Torres Porres wrote:
What? 32Bit 384KHz dac? And for 50 bucks? what the $*#%@?
analogue isn't just audio, higher frequencies might be very useful in some applications, higher resolutions perhaps also .... but even so you can only make use of resolutions that are not lost in the analogue circuit noise.
Simon
On 04/23/2015 05:52 PM, Alexandre Torres Porres wrote:
What? 32Bit 384KHz dac? And for 50 bucks? what the $*#%@?
2015-04-23 12:43 GMT-03:00 Charles Z Henry <czhenry@gmail.com mailto:czhenry@gmail.com>:
It's already started.... 32bit DACs are available from AKM and XMOS, for example. Although I don't know what software/hardware platform you'd use to actually make use of this precision, you can build your own 32-bit sound playback interface with a few boards from HK: http://www.yuan-jing.com/dacs-decoder/32bit-192khz-usb-dac-decoder-ak4399-wm8805-pcm2706-opa627au-optical-coaxial http://www.yuan-jing.com/dacs-decoder/xmos-usb-audio-32bit-384khz-dac-decoder-board-pcm5102-tda1308-headphone-amp
Yeah, but 11 bits of noise sound so much more clear and precise than 3 bits of noise. Also, noise from DC to 192k sounds much more brilliant than noise from DC to only 24k.
(To clarify, these noise bits are integer, not floating point. Floating point noise bits sound even more wonderful, but they require double precision.)
Also, air has a theoretical dynamic upper bound of 194 dB(SPL floating point) above hearing threshold before we hit hard vacuum in the low-pressure zone (leaving little pesky non-linearities aside), so we totally need those 32 bits (vacuum transients without headroom tend to sound mushy and lack punch).
Thanks to Yuan-Jing.com for finally enabling us to hear those in-engine recordings of the Saturn V launches like they were meant to be heard (at least for a split second, before dying of blunt trauma to the head).
(And yeah, I totally believe those specs and all their babble is absolutely plausible. Pd needs to be more future proof.)
On Mon, Apr 27, 2015 at 6:02 AM, Jörn Nettingsmeier nettings@stackingdwarves.net wrote:
On 04/23/2015 05:52 PM, Alexandre Torres Porres wrote:
What? 32Bit 384KHz dac? And for 50 bucks? what the $*#%@?
2015-04-23 12:43 GMT-03:00 Charles Z Henry <czhenry@gmail.com mailto:czhenry@gmail.com>:
It's already started.... 32bit DACs are available from AKM and XMOS, for example. Although I don't know what software/hardware platform you'd use to actually make use of this precision, you can build your own 32-bit sound playback interface with a few boards from HK:http://www.yuan-jing.com/dacs-decoder/32bit-192khz-usb-dac-decoder-ak4399-wm...
http://www.yuan-jing.com/dacs-decoder/xmos-usb-audio-32bit-384khz-dac-decode...
Yeah, but 11 bits of noise sound so much more clear and precise than 3 bits of noise. Also, noise from DC to 192k sounds much more brilliant than noise from DC to only 24k.
(To clarify, these noise bits are integer, not floating point. Floating point noise bits sound even more wonderful, but they require double precision.)
Now, there's still nothing I know of to let you record sound at 32-bit, 384kHz--so it's synthesis only (don't ask me how either).
So to be absolutely non-sarcastic to the audiophiles reading this: the noise from this system will not sound any different whatsoever. In reality, this dac is merely capable of producing signals that are below the noise floor of the playback system and will be entirely unheard.
Also, air has a theoretical dynamic upper bound of 194 dB(SPL floating point) above hearing threshold before we hit hard vacuum in the low-pressure zone (leaving little pesky non-linearities aside), so we totally need those 32 bits (vacuum transients without headroom tend to sound mushy and lack punch).
Thanks to Yuan-Jing.com for finally enabling us to hear those in-engine recordings of the Saturn V launches like they were meant to be heard (at least for a split second, before dying of blunt trauma to the head).
(And yeah, I totally believe those specs and all their babble is absolutely plausible. Pd needs to be more future proof.)
Same thing I tell people about tube amps. Nothing sounds exactly like a Saturn V rocket except a Saturn V rocket. To get the same experience over headphones, you'd better be rockin' a pair of actual rockets.
-
Alexandre Torres Porres -
Charles Z Henry -
Cyrille Henry -
IOhannes m zmölnig -
Jamie Bullock -
Joe White -
Julian Brooks -
Jörn Nettingsmeier -
Lorenzo Sutton -
Miller Puckette -
Simon Wise -
William Huston