CD and 16 bits

[mtrycrafts]

Within the past week there was a response to a post about 16 bit CD having 65,536 levels but then one bit was lost to sign +/-, hence a CD has only 15 bit levels. Lost that message where it was, who, is not important

That information is incorrect. There is no sign loss and there is a full 16 bit capability.

 

[jonnythan]

Um, -32765 to 32766 is 65536 levels still.... the decimal representation doesn't matter, you still have:
0000000000000000 to 1111111111111111

 

[MDS]

It was me and the information is correct. I am a programmer and I can assure you I have a very good understanding of the binary number system. It is sign-magnitude representation. Besides, jonnythan explained it (although the range is -32768 to +32767).

Edit: As an aside, 8 bit audio is totally different. 2^8 = 256 and you would think it would work the same way and the range would be -128 to +127 but it doesn't work that way. The negative values are encoded as 0 to 127 and the positive values are 128 to 255. Arbitrary and stupid if you ask me, but 8 bit audio is a thing of the past anyway.

 

[mtrycrafts]

It was me and the information is correct. I am a programmer and I can assure you I have a very good understanding of the binary number system. It is sign-magnitude representation. Besides, jonnythan explained it (although the range is -32767 to +32768).

Edit: As an aside, 8 bit audio is totally different. 2^8 = 256 and you would think it would work the same way and the range would be -127 to +128 but it doesn't work that way. The negative values are encoded as 0 to 127 and the positive values are 128 to 255. Arbitrary and stupid if you ask me, but 8 bit audio is a thing of the past anyway.

Are you saying that a CD has only 15 bits of data capability? Or, did I misread your post?

 

[BMXTRIX]

It has 65,365 levels of data represented from -32765 to positive 32766...

Accurately it is 15 bit with a sign placement at the beginning/end to show what the value is. But, practically it is 16 bit.

 

[avnetguy]

(although the range is -32767 to +32768).

Actually its -32768 to 32767.

8 bit is no different other than you have 256 levels instead of 65536 for 16 bit.

Steve

 

[MDS]

Yes, you are correct. Forgive my typos. There is always one more negative value than positive in sign-magnitude representation.

8 bit is different though - the values are unsigned; 256 levels but they range from 0 to 255.

 

[mtrycrafts]

It has 65,365 levels of data represented from -32765 to positive 32766...

Accurately it is 15 bit with a sign placement at the beginning/end to show what the value is. But, practically it is 16 bit.

15 bits would give you half that data, 32768. Which way is it, since it has 65536 levels of value.

 

[BMXTRIX]

15 bits would give you half that data, 32768. Which way is it, since it has 65536 levels of value.

Was that last part a question?

I try to avoid being technical whenever possible - so I (personally) would call it 16-bit as to not confuse people.

But, then you get around the techies who insist upon using very accurate terminology. And would correct you and say that it is 15 bit and includes a signed bit.

When I give you the numbers from 0 to 65535 then from -32768 to 32767 it is easy to see that those are two entirely different ranges of numbers.

Yet, both represent 65536 different numbers. To the computer doing the processing it matters a great deal which format it is in. Signed 15 bit or unsigned 16 bit? It matters. But, the actual combinations still yield the same number of possible results. So, are you a programmer who cares about the details, or an end user who really doesn't care about the details and just wants to know how many values are really possible?

15 bit + 1 signed bit for you programmers.
16 bit for the rest of us sane people!

 

[jonnythan]

What matters is that the CD data has *16 bits of resolution*. Period. That statement does not require qualification of any sort.

It doesn't matter if it's a 16 bit integer or a 15 bit signed integer.

Whichever way you look at it, you have 16 bits to work with.

 

[MDS]

I try to avoid being technical whenever possible - so I (personally) would call it 16-bit as to not confuse people.

It is 16 bit and you would be accurate in saying so. Some people may actually be interested in the various ways data is encoded and some are not. So be it.

'16 bits of resolution' is sufficient unless someone asks what the actual values are and that was the purpose of the original question.

 

[avnetguy]

8 bit is different though - the values are unsigned; 256 levels but they range from 0 to 255.

Nope, 8 bit, 16 bit ... no difference except the resolution. Either can use unipolar (all positive numbers) or bipolar (positive and negative numbers) but you have to make sure you match what the A/D or D/A is expecting.

For the general consumer this means nothing, there is no difference in quality but if you are doing data acqusition yourself, like digitizing music, you'll need to know those things.

Steve

 

[MDS]

Random google search:

'On the PC, 8-bit sound samples are always stored as 'unsigned integers', from 0 to 255 (and centred around 127), while 16-bit samples are 'signed integers' from -32768 to +32767 (centred on 0).'

 

[avnetguy]

'On the PC, 8-bit sound samples are always stored as 'unsigned integers',

Nope ... its file format dependant, AIFF for example, stores its values in twos-complement (-128 to 127) as opposed to offset-binary encoding (0 to 255). Really these differences came about with A/D and D/A conversions and whether or not one had a negative power supply rail for the devices.

The point is no cause for concern anyway, simple math takes care of the conversion between the two.

Steve

 

[Buckle-meister]

1+1=10

 

[jonnythan]

1+1=10

There are 10 kinds of people in the world...

 

[mulester7]

.....my two cents on this topic.....


1 Cup all-purpose flour
1 Cup yellow cornmeal
1/4 Cup sugar
1 Tablespoon baking powder
1/2 Teaspoon salt
2 eggs
1 Cup whipping cream
1/4 Cup vegetable oil
1/3 Cup honey

Combine flour, cornmeal, sugar, baking powder, and salt. In a small bowl, beat the eggs. Add cream, oil, and honey to eggs; beat well. Stir into the dry mixture just until moistened. Pour into a greased 9 in. baking pan. Bake at 400 for 20-25 minutes or until a toothpick in the center comes clean.

 

[avnetguy]

.....my two cents on this topic.....


1 Cup all-purpose flour
1 Cup yellow cornmeal
1/4 Cup sugar
1 Tablespoon baking powder
1/2 Teaspoon salt
2 eggs
1 Cup whipping cream
1/4 Cup vegetable oil
1/3 Cup honey

Combine flour, cornmeal, sugar, baking powder, and salt. In a small bowl, beat the eggs. Add cream, oil, and honey to eggs; beat well. Stir into the dry mixture just until moistened. Pour into a greased 9 in. baking pan. Bake at 400 for 20-25 minutes or until a toothpick in the center comes clean.

Sounds interesting .... is there a name for what this is ?

Steve

 

[jonnythan]

Sounds interesting .... is there a name for what this is ?

Steve

You're not from the south, are you?

 

[mtrycrafts]

Was that last part a question?

I try to avoid being technical whenever possible - so I (personally) would call it 16-bit as to not confuse people.

But, then you get around the techies who insist upon using very accurate terminology. And would correct you and say that it is 15 bit and includes a signed bit.

When I give you the numbers from 0 to 65535 then from -32768 to 32767 it is easy to see that those are two entirely different ranges of numbers.

Yet, both represent 65536 different numbers. To the computer doing the processing it matters a great deal which format it is in. Signed 15 bit or unsigned 16 bit? It matters. But, the actual combinations still yield the same number of possible results. So, are you a programmer who cares about the details, or an end user who really doesn't care about the details and just wants to know how many values are really possible?

15 bit + 1 signed bit for you programmers.
16 bit for the rest of us sane people!

OK, I see it, thanks. Suppose a 0 up front represent the - signals and the 15 bits would yield that -32768 and a 1 up front would yield all the positives.