So let me get this straight: if I record in 24 bit Hugh will come round and fit my carpets for me?

--------------------
"Ignorance more frequently begets confidence than does knowledge" Charles Darwin.

Quote simonplent:

---

So let me get this straight: if I record in 24 bit Hugh will come round and fit my carpets for me?

---

Thats pretty much the gist of it;)

Seablade

Quote:

---

So let me get this straight: if I record in 24 bit Hugh will come round and fit my carpets for me?

---

yes, but he'll probably put them on your wall -and then add some bass traps...


--------------------
planet nine
lincoln, uk.

Quote Nathan:

---

Quote:

---

So let me get this straight: if I record in 24 bit Hugh will come round and fit my carpets for me?

---

yes, but he'll probably put them on your wall -and then add some bass traps...

---

Don't forget - he won't be sticking the stuff on your wall -oh no - he wile be hanging them off a couple of hooks so that you don't end up with a sticky wall!

Len

--------------------
www.youtube.com/leonardngmusic

To get back on topic, am I going to trust the ears of a Grammy winner who has (IMHO) recorded some truly stunning work to a level I can only dream of, or some guy who comes along screaming that the math is questionable (and even then as Hugh has pointed out, not very accurately)?

Answers on a postcard, please.

Len

--------------------
www.youtube.com/leonardngmusic

Methinks the Grammy on the wall might be a killer argument...

--------------------
Dynamite with a laser beam...

Quote Len:

---

To get back on topic, am I going to trust the ears of a Grammy winner who has (IMHO) recorded some truly stunning work to a level I can only dream of, or some guy who comes along screaming that the math is questionable (and even then as Hugh has pointed out, not very accurately)?

Answers on a postcard, please.

---

Neither.

--------------------
Midipicks - the all new MIDI Guitar forum...

And I should point out that Hugh is completely wrong too. Carpets are discrete in nature (being composed of individual threads) and so should not be used in this discussion

--------------------
Midipicks - the all new MIDI Guitar forum...

Quote R2B: David Rogers:

---

There is absolutely no excuse for referring to a writer by their surname only - wherther you are a fan or a foe.

---

I don't think that the use of the surname only requires an excuse. It is common practice to refer to writers by their surname and does not imply any pro or anti viewpoint of their work.

--------------------
"Qui habet aures audiendi audiat"

Quote seablade:

---

That would be a misunderstanding on how binary works.

---

I was a bit hazy, but when we talk about 16-bit audio or 24 bit audio, we do talk about a model. An integer model where 16-bits can store 2^16-1 = 65535 discrete values. This model can be implemented by floating point, fixed point or integers as long as it behaves as the model, but the model is integers. So there are 16 and 24 bit resp.

I should have talked about precision and error instead. Would you agree that the error caused by the limited number of bits is in the least significant bit? Would you then agree that when you extend the representation with one bit the error due to limits in the representation is halved? The value you try to represent is the same, but the representation is done with greater precision. Wouldn't then errors from the last bit in a 16 bit representation be within the last 6dB of the 96 dB? If you add a bit to 17 bits, wouldn't the error then become smaller? 6dB smaller for every bit added? Isn't that what like the OP said?

Lars

There's is absolutely no substance in that statement.
Robbie Williams has one a few awards and I would trust quite a few pairs of ears before his, including my own.
This argument is more to do with the theory of it, not the sound.
People saying things sounds 'better' doesn't necessarily mean it IS technically better.
Remember that the human ear is a funny old thing and one man's Barbera Striesand is anoter man's Victoria Beckham.

Also, a decent producer who can record/produce good material and make it sound good doesn't have to know the mathematical theories of digital audio - not at all!

--------------------
Jack of all trades, master of some.

But... the *real* questions are:

"Which soundcard should I buy?"

"Which monitors for under $200?"

"Which is better for recording, a PC or a Mac?"


I do know that Mr. Nichols is well-respected in some circles, and not-so-well respected in others...just like the rest of us, eh? Go to some of the US forums like PSW or GearSlutz, and if you do a search for "Roger Nichols", you will find some very heated arguments over the credibility of Mr. Nichols.

Being in the position he is in, you would expect some scrutiny, but for some reason with Mr. Nichols, it is not a matter of "like vs. dis-like", but rather a matter of credibility. Most of what I have read has *nothing* to do with liking his style or his work, or whether folks agree with his methodology, etc., it seems to be centered on his credibility or lack therof.

Do a search for some other "big names" in the business, Terry Manning, Elloitt Schneiner, Chuck Ainley, Ross Hogarth, etc., and you will find differing opinions on the popularity of their styles, but you will not see many (if any?) credibility "issues"?

So now, I have to wonder: "Why Roger Nichols?" Seems a bit odd, eh?

Recently, he (or should I say his company, "Roger Nichols Digital") announced that they were releasing some Pro Tools plugins, and oh man... The flame wars began once again calling his qualifications into question!!! His company web page is HERE .

Look at the names of a couple of his plugins: "Bitchin-izer" and "Wendel-izer"... these names make me want to run out an buy them... NOT!!! What the hell is a Bitchin-izer anyway???

I ask again: "Why Roger Nichols?" He is obviously successful and has a decent track record with more Grammys than most of us have! So what is it with him? Maybe he tells it like he feels without treading down the politically correct lane? Maybe he speaks his mind and bucks the generally-accepted techniques???

Who knows??? I surely don't!

I only post this as a 3rd party observer, not to stir up the pot with some unsubstantiated rumors, but rather to point out that for some reason, he seems to be quite often associated with controversy. I urge everyone to form their own opinions, and a quick search on any of the above-mentioned forums will provide a lot of details.

Personally, I know nothing of the man other than to say that I like some of his work.

I must admit, I *was* rather "surprised" to see that he was now going to be publishing a regular SOS article considering his "controversial" status.

Actually, on second thought, maybe it is a good idea *because* of his controversial nature! Controversay sells copy!!!

Personal opinion: I found both of his SOS articles to be a bit "ho-hum" and a bit of a let-down.

--------------------
James
"Never interrupt your enemy when he is making a mistake" ~Napoleon Bonaparte~

QUOTE:"Personal opinion: I found both of his SOS articles to be a bit "ho-hum" and a bit of a let-down. "

Ho-humdy-ho...
Now that IS controversial.

--------------------
Jack of all trades, master of some.

Quote Doublehelix:

---

Look at the names of a couple of his plugins: "Bitchin-izer"

What the hell is a Bitchin-izer anyway???

---

Clearly a pluging that makes your mixes 'Bitchin' regardless of your ability. Actually who needs ability, just use his auto mix first, then the bitchinator and your done.

I thought a "bitchin-izer" was what some of ran our posts through before submitting them to this forum

--------------------
~~~ A weasel hath not such a deal of spleen as you are tossed with! www.feline1.co.uk ~~~

Roger Nicholls is a weird one. Sometimes he comes across as a successful, talented and knowledgable guy, and other times he comes across as an opinionated amateur.

Who he *really* is is anybody's guess..

I guess his credibility issues stem from the fact he writes/reviews in many US mags, who are not as unbiased as some more reputable UK mages *ahem* and thus when he talks about gear (and to some extent, techniques) people are not quite sure whether he means what he says or he's doing the endorser/reviewer thing.

Edited by desmond (27/04/06 03:23 PM)

Quote desmond:

---

Who he *really* is is anybody's guess..

---

Rosemary, the telephone operator?

--------------------
~~~ A weasel hath not such a deal of spleen as you are tossed with! www.feline1.co.uk ~~~

Quote cc.:

---

And surely the 'bass benefits more' idea is just plain wrong - the noise floor is reduced for all frequencies equally (until you get into shaping anyway).

---

I think psychoacoustics raise their head at this point. I think we would probably all agree that recording with 24 bit converters generally sounds a lot better than recording with 16 bit converters. There have been countless threads on this aspect before on the forum.

So if one person subjectively feels that the bass benefits in particular, who am I to argue. The fact remains that 24 bits does generally sound better than 16 bits....

The search for an explanation might be a bit wobbly... but there's nothing surprising in that. Very few people have a completely infallible understanding of this subject, myself included.

hugh

--------------------
Technical Editor, Sound On Sound

Quote feline1:

---

I thought a "bitchin-izer" was what some of ran our posts through before submitting them to this forum

---

Personally I prefer 'Auto-Sneer' or 'Maxi-Sarc'.

Quote Sheriton:

---

There's a fundemental misunderstanding of digital audio right here. Back to basics:

Lets start with a 2 bit signal (easier), assuming full scale is 1 volt (and ignoring signing for the moment):

00 - 0v
01 - 0.33v
10 - 0.67v
11 - 1v

Add another bit:

000 - 0v
001 - 0.14v
010 - 0.29v
011 - 0.43v
100 - 0.57v
101 - 0.71v
110 - 0.86v
111 - 1v

---

This is my understanding of things and if I am wrong please correct me:

You seem to be taking a big endian binary word and adding the Least Significant bit at the Most Significant end. You need to add that extra bit to the right, not to the left.

If what you say above is correct, when you convert a 16 bit file to a 24 bit file, the voltages of the equivalent sample points would change. We would have the following:


3 bits____________4 bits
000 - 0.00v_______0000 - 0v
001 - 0.14v_______0001 - 0.07v
010 - 0.29v_______0010 - 0.15v
011 - 0.43v_______0011 - 0.22v
100 - 0.57v_______0100 - 0.29v
101 - 0.71v_______0101 - 0.36v
110 - 0.86v_______0110 - 0.43v
111 - 1,00v_______0111 - 0.50v


(Sorry for the underscores. I couldn't get the formating right).

This would be a big problem. You wouldn't be able to convert a 16 bit file into a 24 bit file (without requantization) because on playback the volumes would have dropped by arround 48 dB.

In reality we have the following:


3 bits_____________4 bits
000 - 0.00v________0000 - 0.00v
--------____________0001 - 0.07v
001 - 0.14v________0010 - 0.14v
--------____________0011 - 0.21v
010 - 0.29v________0100 - 0.29v
--------____________0101 - 0.36v
011 - 0.43v________0110 - 0.43v
--------____________0111 - 0.50v
100 - 0.57v________1000 - 0.57v
--------____________1001 - 0.74v
101 - 0.71v________1010 - 0.71v
--------____________1011 - 0.78v
110 - 0.86v________1100 - 0.86v
--------____________1101 - 0.92v
111 - 1v___________1110 - 1.00v

Every word in the 3 bit word has an equivalent word with the exact same value in the 4 bit word. There are of course extra values in the 4 bit list.

So as I said, the extra bits are "only" giving extra amplitude resolution at the lower end of the amplitude resolution scale. Or put differently, the quantization errors are reduced in amplitude at the lowest resolution scale. The quantization errors are at a lower amplitude so the distortion produced by these quantization errors is lower.

I think the way I phrased my comments was what was confusing the issue. (I seem to have forgotten the word resolution at a few strategic locations. I guess that is what happens when posting at 10 to 4am after a 15 hour work day).

Quote:

---

The extra bit doesn't just add extra resolution at the bottom end of the scale - it adds it throughout.

---

Sorry if I wasn't clear. I am talking about the amplitude resolution scale. Not the full scale of the values themselves. In other words, I am mentaly plotting the resolution of a digital word against its bit size. I am jumping out of one way of thinking into another without making that completely clear.

Quote:

---

Just because the extra zeros are added to the least significant end of the data doesn't mean that it just represents the quietest part of the converted signal.

---

No it indeed doesn't only represent the quietest part of the converted signal. It represents the lowest amplitude resolution scale of the sample values.

This does indeed affect the whole signal but, to my understanding, not in the way that Nichols[1] describes. Quantization errors at the lowest bits (at any bits actually) will be heard as distortion spread out over the whole bandwidth (and aliasing back into the audible band when bit reducing by truncation without application of dither as there are no anti-aliasing filters applied when bit reducing).

Quote:

---

I really wouldn't put much weight on what screen grabs of cool edit show you

---

Actually it is very good compared to others. It shows the reconstructed waveform as it would look after going through the Digital to Analogue process unlike many other wave editors which just "join the dots". Also, because it is in a dB scale, it shows us things relative to how we hear them.

This is one fo the problems with Nichols'[1] article. He writes this:

Quote:

---

A sample point nearing 0dB full scale is 256 times more accurate than the same sample recorded at 16-bit.

---

Nichols[1] implies a logarithmic scale and then goes on to talk about a linear scale but this is incorrect relative to how humans hear things. We hear things on a logarithmic scale so the amplitude of errors and the increased resolution should also be calculated on a logarithimic scale. 24 bit audio is not 256 times more accurate to our ears.

Now, about my outburst. As I said before, I was shocked by reading the article. I was in a very stressfull session and escaped home to have dinner. Instead of happily relaxing while reading an SOS article, I encountered this article which is questionable to say the least.

UnderTow

[1] I'm treating Mr Roger Nichols' article as a technical article. In my book, in a technical discussion, refering to him as Mr Nichols would be a bigger insult than refering to him only by his sir name.

Quote Lars Farm:

---

This model can be implemented by floating point, fixed point or integers as long as it behaves as the model, but the model is integers. So there are 16 and 24 bit resp.

I should have talked about precision and error instead. Would you agree that the error caused by the limited number of bits is in the least significant bit?

---

To be perfectly honest I am not certain I understand the concept you are trying to explain to be able to answer, this shows more my experience of being largely self taught as a programmer and a realtive newcomer in the field of DSP programming;) Sorry. I probably have an opinion on it, I just dont understand the terminology used.

Quote:

---

Would you then agree that when you extend the representation with one bit the error due to limits in the representation is halved?

---

I would agree with this. That is a funciton of binary and the fact it is a base 2 system, you get twice as precie with each added digit.

Quote:

---

The value you try to represent is the same, but the representation is done with greater precision. Wouldn't then errors from the last bit in a 16 bit representation be within the last 6dB of the 96 dB?

---

Now this is where I disagree, though it may come down to misunderstanding what you are trying to express.

If I am understanding you correctly, the shortcoming in this line of thought is the fact that it is not only that single digit that represents that range, but rather many digits(All of them for the largest section)

Quote:

---

If you add a bit to 17 bits, wouldn't the error then become smaller? 6dB smaller for every bit added? Isn't that what like the OP said?

Lars

---

I am not sure that I would quantify the error reduction in terms of dB in that regards. We already decided that it would be half the errors, so I suppose it would be correct, but mentioning it in terms of dB is in my opinion a bit misleading,, even if it is technically correct. Though in this instance I think it would be a difference of 3dB anyways.

The reason I wouldnt do this is I consider it a bit misleading in that it makes it seem that you are trying to say it merely adds that much onto it(Either making the signal louder, or capable of being quieter, the second is technically correct, but does not describe the entire process).

Of course as I said, it is possible we are trying to express the same opinion, and I am just misunderstanding you.

Seablade

Quote Lars Farm:

---

Quote seablade:

---

That would be a misunderstanding on how binary works.

---

I should have talked about precision and error instead. Would you agree that the error caused by the limited number of bits is in the least significant bit? Would you then agree that when you extend the representation with one bit the error due to limits in the representation is halved? The value you try to represent is the same, but the representation is done with greater precision. Wouldn't then errors from the last bit in a 16 bit representation be within the last 6dB of the 96 dB? If you add a bit to 17 bits, wouldn't the error then become smaller? 6dB smaller for every bit added? Isn't that what like the OP said?

Lars

---

This is what I was driving at. We are talking about errors, precision and resolution. In these terms my comments were correct AFAIK.

Thanks for confirming that.

UnderTow

you obviously haven't filtered that post through the bitch-inizer...

-=sigh=-

Quote Sam Inglis:

---

Roger Nichols is talking about converting from 24-bit to 16-bit. You seem to be talking about converting the other way.

---

Not really. I am using my examples to show the bit relationship between 16 and 24 bit recording formats. In my example you will see that you can convert back and forth at will. (Note: This is only applicable in this example. Under normal conditions the signal needs to be dithered when bit reduced).

Quote:

---

What other 'fact' could there be about whether the bass sounded better to Roger Nichols on that particular recording?

---

For one he could have mentioned what he is comparing to. Judging by the rest of the article he seems to be comparing modern 24 bit converters to early 3M digital machines (with a combination of a 12-bit converter with an additional four bits of an 8-bit converter for gain ranging). I could be wrong but as he isn't clear, I can't be sure.

So it would appear that the conclusion of his whole article is: Modern 24 bit converters sound better than 24 year old converters with 12 bit resolution and 4 bit gain scaling.

No [ ****** ] sherlock!

UnderTow

"This is incomplete to say the least. You could have a recording with a noise floor way above -96 dB FS. It won't suddenly go down to -144 dB FS just by converting to 24 bits.
" - Undertow

Undertow, you seem to be fixated with converting a 16 bit file to 24 bit, as demonstrated by your 3 bit - 4bit example and the quote above.
What is it that you are arguing? That 24 bits isn't more accurate than 16? If you do the crude convertion that you propose, you don't gain any resolution, but it wasn't audio recorded at 16 bits, and you haven't started processing it yet.
There was also comments mixing up quantisation noise (roughly no. bits * 6 dB), with the studio signal chain noise floor. They are not the same thing.

--------------------
www.wildhope.com - pike in a lilypond

Quote UnderTow:

---

This is my understanding of things and if I am wrong please correct me:

---

Er... in the nicest possible way, myself and others have already attempted to correct you...

Quote:

---

If what you say above is correct, when you convert a 16 bit file to a 24 bit file, the voltages of the equivalent sample points would change.

---

Quite. That is an error on his part in providing those particualr sample value examples. Clearly, sample values don't change when importing a 16 bit file into a 24 bit environment.

Quote:

---

In reality we have the following:

---

Yes, that table looks right (as far as I can see with the tricky formatting...)

Quote:

---

Every word in the 3 bit word has an equivalent word with the exact same value in the 4 bit word. There are of course extra values in the 4 bit list.

---

Correct.

Quote:

---

So as I said, the extra bits are :

---

nly" giving extra amplitude resolution at the lower end of the amplitude resolution scale.

---

No, I still can't agree. The extra amplitude resolution applies at all signal amplitudes. The amplitude of a sample can be defined with greater resolution whatever it's amplitude if you have more quantising levels (more bits in the wordlength).

Quote:

---

Or put differently, the quantization errors are reduced in amplitude at the lowest resolution scale.

---

No again. The quantisation errors are reduced in amplitude full stop.

Quote:

---

The quantization errors are at a lower amplitude so the distortion produced by these quantization errors is lower.

---

If the converter is correctly dithered, there won't be any distortion, just random noise -- but yes, the quantisation errors are smaller.

Quote:

---

In other words, I am mentaly plotting the resolution of a digital word against its bit size. I am jumping out of one way of thinking into another without making that completely clear.

---

"Resolution of a digital word in relation to its bit size"... huh? You are still not making yourself clear to me!

Quote:

---

Quantization errors at the lowest bits (at any bits actually) will be heard as distortion spread out over the whole bandwidth

---

No, if properly dithered there will be no distortion, just noise.

Quote:

---

We hear things on a logarithmic scale so the amplitude of errors and the increased resolution should also be calculated on a logarithimic scale. 24 bit audio is not 256 times more accurate to our ears.

---

You are correct in that a 24 bit system isn't 256 times more accurate to our ears than 16 bit system, and yes, our sense of hearing is broadly logarithmic.

However, the quantisation scale is linear, and the size of quantisation errors is linearly related to the quantisation resolution. In that context, I think it is fair to say that the amplitude of a sample nearing 0dBFS can theoretically be determined with 256 times greater precision in a 24 bit system than is possible in a 16 bit system.

Quote:

---

In my book, in a technical discussion, refering to him as Mr Nichols would be a bigger insult than refering to him only by his sir name.

---

I don't think he has been knighted...

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Quote:

---

This 256 times higher resolution of a 24-bit sample is in effect everywhere on the waveform, from the lowest levels to the highest peaks. A sample point nearing 0dB full scale is 256 times more accurate than the same sample recorded at 16-bit.

---

Quote UnderTow:

---

This is incorrect. The resolution is only added at the lower end of the amplitude scale.

---

Again, not strictly true. Quantising with more bits reduces the level of the noise floor and allows quieter signal to be heard more easily, but the resolution is actually improved across the entire dynamic range.

---

Indeed. Sorry for confusing things. The terminology I used was inadequate to convey my thoughts. I hope I have clarified things in my later posts.

Quote:

---

Think of it this way. You have an analogue signalto quantise that runs between 0mV at its quietest and 1000mV at its loudest. If you quantise with 16 bits, you effectively have to draw a grid over the audio waveform with 65536 equally spaced horizontal lines, each line representating the level of a valid quantising value.

If you quantise with 24 bits, that grid has to contain 16,777,216 lines -- in the same space. We don't only add the extra 16,711,680 lines underneath the bottom of the 16 bit grid! So yes, the resolution across the entire dynamic range is improved. We can define the position of a sample peak with considerably greater resolution regardless of its particular amplitude. That is why the quantising error is smaller, and that's why the noise floor is lower.

---

Agreed. Again, my terminology was wanting. The errors at lower bits are much smaller than errors at higher bits.

Quote:

---

It certainly was a, shall we say, interesting article...

hugh

---

I think this sums it up.

UnderTow

Quote UnderTow:

---

The errors at lower bits are much smaller than errors at higher bits.

---

I think you are confusing quantisation errors with bit errors.

A quantisation error has the same maximum amplitude regradless of the amplitude of the audio signal being quantised.

A bit error would produce an erroneous sample amplitude, in which an error of the LSB would produce a far smaller amplitude error than an error in the MSB. But this particular issue is not relevant in any way to the topic being discussed.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Quote:

---

If what you say above is correct, when you convert a 16 bit file to a 24 bit file, the voltages of the equivalent sample points would change.

---

Quite. That is an error on his part in providing those particualr sample value examples. Clearly, sample values don't change when importing a 16 bit file into a 24 bit environment.

---

Ok, we agree.

Quote:

---

Quote:

---

In reality we have the following:

---

Yes, that table looks right (as far as I can see with the tricky formatting...)

---

Ok, we agree.

Quote:

---

Quote:

---

Every word in the 3 bit word has an equivalent word with the exact same value in the 4 bit word. There are of course extra values in the 4 bit list.

---

Correct.

---

Ok, we agree.

Quote:

---

Quote:

---

So as I said, the extra bits are :

---

nly" giving extra amplitude resolution at the lower end of the amplitude resolution scale.

---

No, I still can't agree. The extra amplitude resolution applies at all signal amplitudes. The amplitude of a sample can be defined with greater resolution whatever it's amplitude if you have more quantising levels (more bits in the wordlength).

---

Arf! I am not talking about the amplitude scale nor am I talking about the amplitude resolution! I am talking about an amplitude resolution scale

Imagine the following plot: Y-axis = amplitude resolution of LSB. i.e. the amplitude precision of the LSB value. X-axis = number of bits.

The plot line starts on the left with Y=-6dB FS and X=1 bit. The next dot on the line is Y=-12dB FS X=2 bit, then Y=-18 dB FS and X=3 bits. At the end we have Y=-144 dB FS and X=24 bits.

So the line decreases. The amplitude resolution of the LSB of each digital word decreases as the word length increases.

Quote:

---

Quote:

---

Or put differently, the quantization errors are reduced in amplitude at the lowest resolution scale.

---

No again. The quantisation errors are reduced in amplitude full stop.

---

Arf! By "at the lowest resolution scale" I mean at the least significant bits. As you go through the bits in a digital word, starting at the MSB and ending at the LSB, the amplitude of the error caused by flipping a particular bit decreases.

Quote:

---

Quote:

---

The quantization errors are at a lower amplitude so the distortion produced by these quantization errors is lower.

---

If the converter is correctly dithered, there won't be any distortion, just random noise -- but yes, the quantisation errors are smaller.

---

One of my points about the article is that the word dither is not mentioned once. But I agree with what you say. The dither linearises the quantae.

But as I am talking about quantization errors, you could have given me the benefit of the doubt and assumed I was talking about inproper or absent dither.

Quote:

---

Quote:

---

In other words, I am mentaly plotting the resolution of a digital word against its bit size. I am jumping out of one way of thinking into another without making that completely clear.

---

"Resolution of a digital word in relation to its bit size"... huh? You are still not making yourself clear to me!

---

Oh come on, put a bit of effort into it.

As the word length increases, the resolution increases. I know that you know and understand that so you must be able to understand that sentence above.

Quote:

---

Quote:

---

Quantization errors at the lowest bits (at any bits actually) will be heard as distortion spread out over the whole bandwidth

---

No, if properly dithered there will be no distortion, just noise.

---

Arf! If you properly dither, you won't have quantization errors! So if you have quantization errors, you havn't properly dithered!

You wouldn't be trying to be disingenuous would you?

Quote:

---

Quote:

---

We hear things on a logarithmic scale so the amplitude of errors and the increased resolution should also be calculated on a logarithimic scale. 24 bit audio is not 256 times more accurate to our ears.

---

You are correct in that a 24 bit system isn't 256 times more accurate to our ears than 16 bit system, and yes, our sense of hearing is broadly logarithmic.

---

Good.

Quote:

---

However, the quantisation scale is linear, and the size of quantisation errors is linearly related to the quantisation resolution. In that context, I think it is fair to say that the amplitude of a sample nearing 0dBFS can theoretically be determined with 256 times greater precision in a 24 bit system than is possible in a 16 bit system.

---

Only if we qualify that statement be precising that we are talking about the linear scale as you did. I have no objection to what you are stating.

Quote:

---

Quote:

---

In my book, in a technical discussion, refering to him as Mr Nichols would be a bigger insult than refering to him only by his sir name.

---

I don't think he has been knighted...

Hugh

---

---

Lol! Typo.


UnderTow

Quote Hugh Robjohns:

---

Quote UnderTow:

---

The errors at lower bits are much smaller than errors at higher bits.

---

I think you are confusing quantisation errors with bit errors.

---

Arf! Not at all. I am doing some mental acrobatics to explain a point!

I think you are working under the assumption that I don't understand what I am talking about instead of working under the assumption that I do and trying to follow my logic.

Of course, the first does require less effort on your part.

Quote:

---

A quantisation error has the same maximum amplitude regradless of the amplitude of the audio signal being quantised.

---

Agreed and stated by me in previous posts.

Quote:

---

A bit error would produce an erroneous sample amplitude, in which an error of the LSB would produce a far smaller amplitude error than an error in the MSB. But this particular issue is not relevant in any way to the topic being discussed.

Hugh

---

Not directly but it is as a bit of mental exercise to understand the subject. You know as well as me that digital audio is not intuitive at first glance.

As I have explained in earlier posts, I was describing my mental processes to clarify that my terminology was wrong but not my thought processes.

Again, try and follow my logic instead of assuming that I don't understand. Then! you can say I am wrong.

UnderTow

Quote Spyder:

---

"This is incomplete to say the least. You could have a recording with a noise floor way above -96 dB FS. It won't suddenly go down to -144 dB FS just by converting to 24 bits.
" - Undertow

Undertow, you seem to be fixated with converting a 16 bit file to 24 bit, as demonstrated by your 3 bit - 4bit example and the quote above.

---

Arf! No! *sigh* These are just examples to clarify a point. Unfortunately, people are not trying to following the logic and are assuming I am wrong from the start.

UnderTow

Quote Lars Farm:

---

AFAIK… In the digital domain the interpretation of the bitpattern is:
- bit 1/MSB carries info up to 0dBFS
- bit 2 (that happens to be LSB in example 1) carries info about what happens up to -6dBFS
- bit 3 (that happens to be LSB in example 2) carries info about what happens up to -12dBFS

So, the added bit has added four new values at the bottom of the amplitude range. Logarithmic, not linear. Shouldn't it be so in the analouge domain too?

Lars

---

If expressed in dBV or dbm, yes. Of course if expressed in absolute voltages, then no.

You seem to be the only one that understands what I am driving at ...

UnderTow

Quote Len:

---

To get back on topic, am I going to trust the ears of a Grammy winner who has (IMHO) recorded some truly stunning work to a level I can only dream of, or some guy who comes along screaming that the math is questionable (and even then as Hugh has pointed out, not very accurately)?

Answers on a postcard, please.

Len

---

And this is the crux of the problem. He has a Grammy (or more) and people believe him blindly. That is why he should either, check his facts or have his articles checked by somone at SOS that is more knowledgable about digital audio.

I am not saying that Mr Nichols should be fired. I am saying that he should not write deep technical articles about digital audio. I feel it would best serve the subscribers of SOS if he shared his knowledge about recording and mixing. That knowledge that has, amongst other things, earned him a Grammy award and respect in (part of) the recording industry.

UnderTow

I stumbled upon this thread before reading the article and it made me want to read it to find out what all the ‘fuss’ is about.
To quote from the article, ‘As a percentage of the difference between adjacent samples, the quantisation error is much greater for a low-frequency signal than a high-frequency one’.

After reading this I had to double check the cover of the magazine to make sure that it wasn’t the April issue.
But seriously, is this comparison actually meaningful? Can’t a low frequency signal be much more accurately represented than a high frequency one using the same sample rate?

e.g. a 48 Hz sine wave sampled at 48 kHz contains 1,000 samples per sine wave.
Whereas, a 4.8 kHz sine wave sampled at 48 kHz contains 10 samples per sine wave.
I’m not saying that this means that lower frequency sounds will ‘sound better’, I’m just trying to make another comparison, to juxtaposition it against Roger’s.
Is it anymore meaningful? My mind says yes, but I really don't know!

I found Roger’s style of writing difficult to follow; it reminded me that I use certain words in conversation that I’d find impossible to define accurately, but when I look them up in the dictionary I usually see that I’ve been using them in the correct context. There’s a big difference between being able to do something and to be able to talk about what you are doing in a clear and concise way. I’m looking forward to his next column; although I'm not sure that my motives are pure.

Quote Steve Hill:

---

Methinks the Grammy on the wall might be a killer argument...

---

That’s assuming you can still see it after Hugh’s popped round and carpeted your lounge wall from floor to ceiling

Edited by Crow (27/04/06 07:20 PM)

everybody step back !!!

the engineers are arguing !

this could turn nasty.



--------------------
http://profile.myspace.com/abitfunkdub/

I'll repeat part of my old post:

" [Mr Nichols] thinks long and hard, and comes up with a good, logically consistent theory about bass sounds with 24 bit. And writes an interesting and brain-engaging article."

So? What's wrong?

His argument is logically self-consistent.

SOS is not a peer-reviewed science magazine.

If someone stimulates some thinking, that's a good thing. Theories will always provoke criticism - it's the way science and knowledge progress. The article is obviously personal effort at understanding a phenomenon.

G

--------------------
PA stuff on FB

Quote Crow:

---

After reading this I had to double check the cover of the magazine to make sure that it wasn’t the April issue.

---

That was my reaction too ...

Quote:

---

But seriously, is this comparison actually meaningful? Can’t a low frequency signal be much more accurately represented than a high frequency one using the same sample rate?

e.g. a 48 Hz sine wave sampled at 48 kHz contains 1,000 samples per sine wave.
Whereas, a 4.8 kHz sine wave sampled at 48 kHz contains 10 samples per sine wave.
I’m not saying that this means that lower frequency sounds will ‘sound better’, I’m just trying to make another comparison, to juxtaposition it against Roger’s.
Is it anymore meaningful? My mind says yes, but I really don't know!

---

Actually no. You only need two sample points to accurately represent a sine wave. Any sine wave up to half the sampling rate (the so called nyquist frequency) can be accurately reproduced. For a full explanation of this, please read this article:

http://www.lavryengineering.com/documents/Sampling_Theory.pdf

UnderTow

I'll repeat part of my old post:

" [Mr Nichols] thinks long and hard, and comes up with a good, logically consistent theory about bass sounds with 24 bit. And writes an interesting and brain-engaging article."

So? What's wrong?

His argument is logically self-consistent.

SOS is not a peer-reviewed science magazine.

If someone stimulates some thinking, that's a good thing. Theories will always provoke criticism - it's the way science and knowledge progress. The article is obviously personal effort at understanding a phenomenon.

G

BTW: Nice comment about the banjo . . . though I like them, myself, and Bela Fleck is of course, a god.

--------------------
PA stuff on FB

Quote UnderTow:

---

Quote Crow:

---

But seriously, is this comparison actually meaningful? Can’t a low frequency signal be much more accurately represented than a high frequency one using the same sample rate?

e.g. a 48 Hz sine wave sampled at 48 kHz contains 1,000 samples per sine wave.
Whereas, a 4.8 kHz sine wave sampled at 48 kHz contains 10 samples per sine wave…

---

Actually no. You only need two sample points to accurately represent a sine wave. Any sine wave up to half the sampling rate (the so called nyquist frequency) can be accurately reproduced.

---

My choice of sine waves in my example was unfortunate; replace that with a complex waveform and my point will be clearer.

Quote UnderTow:

---

For one he could have mentioned what he is comparing to. Judging by the rest of the article he seems to be comparing modern 24 bit converters to early 3M digital machines (with a combination of a 12-bit converter with an additional four bits of an 8-bit converter for gain ranging). I could be wrong but as he isn't clear, I can't be sure.

So it would appear that the conclusion of his whole article is: Modern 24 bit converters sound better than 24 year old converters with 12 bit resolution and 4 bit gain scaling.

No [ ****** ] sherlock!

UnderTow

---

It's been a few weeks since I read that article, but WOW: that statement alone makes it clear you missed the forest for the trees, and got so hung up on tech details that you entirely missed his point for writing the article!

The SUBJECT of his article is to point out issues encountered when transferring between digital formats. OF COURSE HE'S TALKING ABOUT ANTIQUATED GEAR, since that's the same gear that was used to make the original recordings he's archiving!

In his case, he's archiving digital recordings made in the early 1980's on 3M multitrack recorder, and he's simply saying that understanding the workings of the gear that the original music was recorded on will only aid in ending up with a more accurate rendition of the original material in current digital format. Yes, he recorded on that gear back when it WAS state-of-the-art, and knows it's inner workings quite intimately.

Now if he was unaware of problems with the earlier gear, or was a young whipper-snapper engineer with no first-hand experience of the 3M machine, he might have been tempted to use a digital interface board to do the transfers digitally, thinking a digital transfer would be preferable. Bit no, he decided NOT to do that, since it would've induced non-linear errors that the 3M's original D-A converters compensate for.

So in his words, "analogue transfers it was to be". He recorded to 24-bit from the analog outs.... Why? To make a more accurate archive by recording in 24-bit with modern gear....


You also referred to his last article, where he mentioned hard drives. You completely missed the point there, too: he was providing anecdotal experience that leaving data on hard drives in NOT the most reliable archiving method. This is hardly just his theory: computer scientists will tell you hard drives are subject to data corruption from stray cosmic radiation, and the damage is cumulative. Have an error induced in the wrong sector, and the entire drive may be left unuseable. I think he mentioned the concern that some of the moving parts inside of hard drives tend to seize/bind if the drives are allowed to sit without use for long periods (decades), which is entirely true due to bearing grease and other issues.

Do a Google on "hard drive bearing grease".


@@@@

Dude (and you come off as a Dude, BTW), I'd think long and hard before calling Roger Nichols to the carpet on either his knowledge or experience: he's worked for the major labels for at least 4 decade, and has FORGOTTEN more about digital audio and recording engineering than some wanna-be anorak freshman straight out of of basic audio theory classes (earning a "C", no doubt) could hope to know. The "Immortal" Roger Nichols has been working in the studios as an engineer even before many of us were waddling about in diapers, and the major audio manufacturers were banging on his door for his technical assistance to help them usher in digital audio gear in the early 1980s!

If nothing else, he has the respect of Donald Fagan/Walter Becker, and if he's good enough to be the engineer for Steely Dan records, then that's good enough for me.

BTW, SOS, Kudos on adding Roger Nichols to the roster: a definite feather in the cap.

Chris

--------------------
1.5GHz Al 17" Powerbook G4 (2.0GB RAM, Hitachi 60GB 7,200 rpm drive), running Logic Pro 7 under OSX 10.4.5