Data sheets never lie.......
Of course not. Cough, cough.
Since I have nothing better to do today [joke], here some selected excerpts from various AD DACs, as it is the only databook I can find today.
AD1856 (16-bit): Makes no claim of S/N, or dynamic range. Only "15-bit monoticity".
AD1860 (18-bit): S/N ratio of 102 dB (16.94 bits), and 15-bit monotonicity.
AD1862 (20-bit): S/N ratio of 113 dB (18.77 bits), and a dynamic range of 102 dB.
AD1864 (18-bit): S/N of 102 dB, dynamic range of 94 dB (15.61 bits).
AD1865 (16-bit): S/N of 107 dB (17.77 bits), dynamic range of 94 dB.
AD1868 (18-bit): S/N of 95 dB (15.78 bits), dynamic range of 86 dB (14.28 bits). Yuk!
What is theorectical and practical aren't the same, and vary greatly. Guess we are lucky that someone doesn't have the last one in their soundcard, or this could go on for weeks.
Remember Bob Pease's caveat on data sheets (supplied by Phred). Speaking of which.........
Dynamic range was measured using something called "EIAJ Test Standard CP-307", whatever that is. Let's see if he can dig that one out. I can just hear him screaming: "Is that all you think I have time to do?"
No comment.
Jocko
Of course not. Cough, cough.
Since I have nothing better to do today [joke], here some selected excerpts from various AD DACs, as it is the only databook I can find today.
AD1856 (16-bit): Makes no claim of S/N, or dynamic range. Only "15-bit monoticity".
AD1860 (18-bit): S/N ratio of 102 dB (16.94 bits), and 15-bit monotonicity.
AD1862 (20-bit): S/N ratio of 113 dB (18.77 bits), and a dynamic range of 102 dB.
AD1864 (18-bit): S/N of 102 dB, dynamic range of 94 dB (15.61 bits).
AD1865 (16-bit): S/N of 107 dB (17.77 bits), dynamic range of 94 dB.
AD1868 (18-bit): S/N of 95 dB (15.78 bits), dynamic range of 86 dB (14.28 bits). Yuk!
What is theorectical and practical aren't the same, and vary greatly. Guess we are lucky that someone doesn't have the last one in their soundcard, or this could go on for weeks.
Remember Bob Pease's caveat on data sheets (supplied by Phred). Speaking of which.........
Dynamic range was measured using something called "EIAJ Test Standard CP-307", whatever that is. Let's see if he can dig that one out. I can just hear him screaming: "Is that all you think I have time to do?"
No comment.
Jocko
SY said:
Hm, I guess you are right that the acoustic wave from a trumpet
would have an offset, at least if the mic is close enough. On
the other hand, you might AC-couple or remove the offset
digitally, but then of course we actually introduce an artificial
offset to get rid of the non-artificial offset.
SY said:
The whole symmetry/offset thing is irrelevant. Losing one quantization step is NOT the same as losing your LSB. It only means that instead of describing the waveform with 65,536 levels, you're describing it with 65,535 levels. So instead of 96.329599dB, you're looking at 96.329466dB.
se
laws of phyics
The graph is from the April 2001 reveiw of the Naim CD. Full scale output is 2 volts. The referenced output level is 2 volts the same as the output full scale voltage of the Burr Brown DAC referenced (1 volt to -1 volt for an amplitude of two volts. The article refers to "three discrete voltage levels being clearly evident." From the graph it would appear to be about 90 microvolts, 0 Volts, and -90 microvolts. This looks like the LSB voltage to me with some noise on top of it. It looks like a nice straight transition between the LSB amplitude levels and zero volts. I don't see any transistions between to a lower weighted bit voltage and the sign wave appears to be truncated to transistions of the LSB.
http://www.stereophile.com/showarchives.cgi?730
If looks like a DAC, walks like a DAC...........
Sorry for all the legalize. Can you tell me what part was out of context? While speculating on professions and understanding of the topic, you might want to ask KYW and jwb what their education background and profession are as well.........
The graph is from the April 2001 reveiw of the Naim CD. Full scale output is 2 volts. The referenced output level is 2 volts the same as the output full scale voltage of the Burr Brown DAC referenced (1 volt to -1 volt for an amplitude of two volts. The article refers to "three discrete voltage levels being clearly evident." From the graph it would appear to be about 90 microvolts, 0 Volts, and -90 microvolts. This looks like the LSB voltage to me with some noise on top of it. It looks like a nice straight transition between the LSB amplitude levels and zero volts. I don't see any transistions between to a lower weighted bit voltage and the sign wave appears to be truncated to transistions of the LSB.
http://www.stereophile.com/showarchives.cgi?730
If looks like a DAC, walks like a DAC...........
Sorry for all the legalize. Can you tell me what part was out of context? While speculating on professions and understanding of the topic, you might want to ask KYW and jwb what their education background and profession are as well.........
Re: Data sheets never lie.......
No one's denying that practical implementations will not be the same as the theoretical ideal. That's because it wasn't Kuei's claim that practical implementations are limited to 90.3dB. His claim was that 90.3dB was the ABSOLUTE THEORETICAL maximum. That you simply cannot encode any information below 90.3dB. Period.
And THAT'S what's at issue here.
So there's no need to refer to any data sheets.
se
Jocko Homo said:
No one's denying that practical implementations will not be the same as the theoretical ideal. That's because it wasn't Kuei's claim that practical implementations are limited to 90.3dB. His claim was that 90.3dB was the ABSOLUTE THEORETICAL maximum. That you simply cannot encode any information below 90.3dB. Period.
And THAT'S what's at issue here.
So there's no need to refer to any data sheets.
se
Re: laws of phyics
Suppose you're referring to my post. I don't think I said it was
out of context, unless you mean that table you posted
yesterday. However, you still obviously don't have a clue what
an LSB is. In your world the LSB can obviously have three
different values. LSB means "the least significant bit". Now
we don't even need to bother about the middle two words,
so let's just say "the bit". Note the "the", which to a non-native
speaker of English like me, presumes there is only one bit,
not two, not three, not 4711, but one. One bit, by definition,
can only encode two values.
Ah, there were some Soviet experiments with
ternary computers back in the 60's, using base three, but I
don't think they are used for PCM and the word "bit" is anyway
no longer appropriate in this context.
Ah, yes, I skipped the LS for simplicity. The LSB is simply
the one bit having the least weight in a binary encoded
number. However, it is not allowed to compensate this somewhat
humiliating role by having three possible values instead.
(As for the profession and background of Kuei you will find himself
telling that a while back in this thread, or in the cap thread,
or wherever it went from there. And for myself, in case you
are curios, I think about the same procedure will apply.)
Fred Dieckmann said:
Suppose you're referring to my post. I don't think I said it was
out of context, unless you mean that table you posted
yesterday. However, you still obviously don't have a clue what
an LSB is. In your world the LSB can obviously have three
different values. LSB means "the least significant bit". Now
we don't even need to bother about the middle two words,
so let's just say "the bit". Note the "the", which to a non-native
speaker of English like me, presumes there is only one bit,
not two, not three, not 4711, but one. One bit, by definition,
can only encode two values.
Ah, there were some Soviet experiments with
ternary computers back in the 60's, using base three, but I
don't think they are used for PCM and the word "bit" is anyway
no longer appropriate in this context.
Ah, yes, I skipped the LS for simplicity. The LSB is simply
the one bit having the least weight in a binary encoded
number. However, it is not allowed to compensate this somewhat
humiliating role by having three possible values instead.
(As for the profession and background of Kuei you will find himself
telling that a while back in this thread, or in the cap thread,
or wherever it went from there. And for myself, in case you
are curios, I think about the same procedure will apply.)
Re: laws of phyics
Perhaps you need to go back and re-read Kuei's original cliam too.
These are the ABSOLUTE limits of the CD Format. Below this (-90.3db) or above this (0db) there is no way to record any Information.
Again, he's saying that there is NO WAY TO RECORD any information that's below 90.3dB.
That's simply incorrect. Information below 90.3dB WILL BE RECORDED. If the signal to be recorded is equal to or greater than +/-1/2LSB, it will be recorded in an undithered system. In dithered systems (which describes virtually all of them) you can record much lower than this but the original context was for an undithered system.
So in order to show that Kuei's claim is correct, you have to show that a signal below -90.3dB cannot be recorded.
se
Fred Dieckmann said:
Perhaps you need to go back and re-read Kuei's original cliam too.
These are the ABSOLUTE limits of the CD Format. Below this (-90.3db) or above this (0db) there is no way to record any Information.
Again, he's saying that there is NO WAY TO RECORD any information that's below 90.3dB.
That's simply incorrect. Information below 90.3dB WILL BE RECORDED. If the signal to be recorded is equal to or greater than +/-1/2LSB, it will be recorded in an undithered system. In dithered systems (which describes virtually all of them) you can record much lower than this but the original context was for an undithered system.
So in order to show that Kuei's claim is correct, you have to show that a signal below -90.3dB cannot be recorded.
se
Oh well Steve, I guess it's the way it is. While some of us may
be accused of being too theoretical, sometimes not entirely
unfair, there are others who if asked what 2+3 is would hook
up two bench PSUs in series, set one at 2 volts, the other at
3 volts and then measure the total voltage and say that
2+3 is 4.998 +/2 digits. Don't know how they count much
further than 50 or so, though?
be accused of being too theoretical, sometimes not entirely
unfair, there are others who if asked what 2+3 is would hook
up two bench PSUs in series, set one at 2 volts, the other at
3 volts and then measure the total voltage and say that
2+3 is 4.998 +/2 digits. Don't know how they count much
further than 50 or so, though?
Re: Re: ...Yes M'Lud, the evidence...
Unbelievable. We're back at square one. That's what I thought to begin with: we have half the bits for plus and half the bits for minus. So, we have 65,536/1 or 96 dB peak to peak, but only 32,768/1 or 90 dB peak dynamic range.
And Christer is right, to this theoretical discussion what matters is the definition of dynamic range. DC offset or whatever other technical implementation details don't matter.
My conclusions:
A) The data have a dynamic range of 96 dB. Can we agree on that?
B) But in the context of audio where dynamic range is usually expressed as "LSB" (minimum recordable/encodable/whatever) -to - peak value, we have 90 dB.
Steve Eddy said:
Unbelievable. We're back at square one. That's what I thought to begin with: we have half the bits for plus and half the bits for minus. So, we have 65,536/1 or 96 dB peak to peak, but only 32,768/1 or 90 dB peak dynamic range.
And Christer is right, to this theoretical discussion what matters is the definition of dynamic range. DC offset or whatever other technical implementation details don't matter.
My conclusions:
A) The data have a dynamic range of 96 dB. Can we agree on that?
B) But in the context of audio where dynamic range is usually expressed as "LSB" (minimum recordable/encodable/whatever) -to - peak value, we have 90 dB.
Christer said:Oh well Steve, I guess it's the way it is. While some of us may
be accused of being too theoretical, sometimes not entirely
unfair, there are others who if asked what 2+3 is would hook
up two bench PSUs in series, set one at 2 volts, the other at
3 volts and then measure the total voltage and say that
2+3 is 4.998 +/2 digits. Don't know how they count much
further than 50 or so, though?
Hehehe.
The really sad thing here is that what we're discussing here isn't something that's way out on the fringes like the latest theories in quantum physics. Linear PCM (the modulation scheme used for CD audio) was first devised back in 1937 and is quite well understood. It's not as if it just happened to pop up with the advent of the CD and we're still trying to figure it out.
Isn't it just a bit odd (no pun intended) that even though Kuei has been making this claim for at least 6 years (and having been the one to make the claim the onus is on him to substantiate it) that he hasn't been able to cite a single reference source (not even any East German textbooks) that says the theoretical maximum dynamic range of 16 bit PCM is 90.3dB?
He's only been able to keep this going with sleight of hand mathematics and indirect inferrences.
Yet any web search will pull up thousands of references to the 96dB figure.
se
Nope.
It's not offset by one bit, it must be counted as if offset by 32768 bits in order to reach a valid comparison with the definition (as I understand it) of dynamic range in audio, namely, the smallest possible encoded signal above the largest possible encoded signal and this PEAK ABOVE GROUND, not peak to peak.
The smallest possible signal is one bit, toggled on or off either plus from gnd or minus from gnd (possible asymmetrical hence). The maximum distance from ground can only be half the maximum distance between ALL encodable values, i.e. 65536/2=32768.
QED
(or CQFD for the French)
It's not offset by one bit, it must be counted as if offset by 32768 bits in order to reach a valid comparison with the definition (as I understand it) of dynamic range in audio, namely, the smallest possible encoded signal above the largest possible encoded signal and this PEAK ABOVE GROUND, not peak to peak.
The smallest possible signal is one bit, toggled on or off either plus from gnd or minus from gnd (possible asymmetrical hence). The maximum distance from ground can only be half the maximum distance between ALL encodable values, i.e. 65536/2=32768.
QED
(or CQFD for the French)
SY said:
Well, fair enough. For a valid discussion then, please, let us first agree on a definition of dynamic range. I gave mine. What is yours? No sarcasm. I may have misunderstood something. But we won't get anywhere if we can't agree on the terms we're talking about.
Christer said:
I see that you have already catched up with the latest rocket
science from Texas and use the Dieckmann ternary bit, instead
of the usual binary one.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.