All posts with more than 3000 Hits, prior to 2008

Tue May 20, 2003 2:27 pm

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:02 pm, edited 1 time in total.

Tue May 20, 2003 2:40 pm

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:03 pm, edited 1 time in total.

Tue May 20, 2003 4:03 pm

http://www.silcom.com/~aludwig/EARS.htm

Don't just take my word for it. My point has never been that cd's suck!! It has always been the arguement for what is possible. Why not put the scope way beyond what a human can hear? People can adapt, and a trained ear I believe can POSSIBLY adapt as well. IF one person can hear a difference, then that should be the standard. I feel the same way with video!!

I can't believe that a lot of people are arguing symantacs with meon this board. I have more then made my point. Subjectively it comes down to opinion. MY opinion has always been....it is possible.

Wed May 21, 2003 12:10 am

[quote="vinyljunkie]
A trick question. How would you describe the difference in the final waveform for a square wave above 8kHz compared to a sawtooth wave at the same frequency after they are transfered to a DSD master and back to CD-DA? Can you describe the wave characteristics for both wave forms after this process?

When you have solved this puzzle you may see things a little bit different than you do now.[/quote]

Since gensim did not answer I will try to give an answer. It might a be trap, but I presume the final waveform is the same regardless of the two transfer methods described.

Wed May 21, 2003 2:31 am

Um what does that have to do with anything? The saw will give a wave form if it is one contiuous stroke. In that wave form all the same rules apply. I am done with this thread. Using logic I "proved" DSD technology can improve the representation on a dowsampled red book cd.

By the way, Vinylman you might want to check out the IT HURTS ME post. I brought up something you might find interesting.

Wed May 21, 2003 6:08 am

<Um what does that have to do with anything? The saw will give a wave form if it is one contiuous stroke.

I really hope that is a joke.

Wed May 21, 2003 8:06 am

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:08 pm, edited 1 time in total.

Wed May 21, 2003 9:47 am

I had to edit this post. Now I know what Greg was talking about. I saw a saw tooth reference and didn't even get what you were saying. I thought you were referring to a tool. Man I must be tired.

Anyway, Vinyl Junkie you are arguing the same crap that Vinyl man was arguing. Now that I have fixed this error, let me post one more to clarify.

Wed May 21, 2003 10:06 am

Ok first of all Vinyl Junkie what you say doesn't make sense to me. It is quite vague.

I have no idea what you are asking. Please explain it a little better. Why 8khz? Are you asking for the wave characteristics of each one separately? My guess that they each stay the same as they are, cause both fit in the range of a CDDA?

Really, just get to your point. If I didn't know better, this is still leading up to the crap that the samples don't matter after a certain frequency. Of which case, I send you straight over to The It Hurts Me thread, just like I did with Vinylman(though he seems to have ignored it as well after he laid a egg!! while disproving his own "superior format"!)

Wed May 21, 2003 5:31 pm

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:09 pm, edited 1 time in total.

Wed May 21, 2003 6:05 pm

First of all, I doubt if you were reading my posts that you would be posing this "riddle" to me right now. I told you once. I am not in to solving problems. If you want to discuss that is fine. I myself never claimed to be a sound expert. Look at my last couple of posts. Are you disproving what I said? Quit this bull and just get to your answer!

I have no idea what 8khz produces and I readily admit that. Vinylman doesn't seem to know either(or Greg for that matter-though I am only guessing).

Name one instance where I put words in Vinyman's mouth. I caught him and he has not answered for that reason. He "disproved" his own format in his last post on It Hurts Me. I may not be a sound expert but I am smart enough to know that there is more of a volume of accurate representation when using a DSD transfer provided it is not sampled in the same place. That is all.

If you can disprove this, then please let me know. I am anxiously awaiting. I want to know the answer to the riddle, cause I am too "dumb" to figure it out. Or not as "knowledgable" as you or vinyl man or greg. If they know the answer, I invite them to answer as well. Lets see if all the experts come up with the same answer. This should be fun.

Wed May 21, 2003 6:20 pm

Ahh I figured it out. It is a high frequency range you speak of. 8khz is for phone lines right?

So exactly what I was speaking of before. THIS HAS NOTHING TO DO WITH MY ARGUEMENT. Go back to the IT HURTS ME post and read clearly. THis has nothing to do with high frequency range. My arguement is the same range as a red book cd. USE YOUR HEAD AND THINK ABOUT WHAT I SAID. Doesn't mater how high or fast. God didn't you ever take a math course?

Before you keep trying to dazzle people with your tricks, first COMPREHEND what I wrote.

P.s. The phone line bit is wrong, I give up!
Last edited by genesim on Wed May 21, 2003 6:32 pm, edited 1 time in total.

Wed May 21, 2003 6:26 pm

Hi -

As a technological layman, I'd like to thank all you guys for explaining the intricacies of DSD, curvform, PCM transfer, CDDA, and spacial separation.

All this is now as clear as mud !

Colin B
Last edited by ColinB on Wed May 21, 2003 8:07 pm, edited 1 time in total.

Wed May 21, 2003 6:38 pm

I wish I could draw on this board because I would make my point clearer. Though at least Greg agreed with me to some extent. I admit that I am arguing about SUBTLE SUBTLE differences.

Truthfully, this subject is getting old and it really isn't exhillerating for me to prove anybody wrong. Some people get off on it. I think I have a firm grasp on what happens with cd transfer. In the end, I have my opinion. I have read tons of articles arguing back and forth. I would like this to end soon and get back to talking about Elvis. Though I am "anxiously" waiting for the answer to the riddle!

Wed May 21, 2003 8:55 pm

Genesim,

I and a couple others have tried to explain all this, in fact in the last week I have spent over an hour total time composing responses for this topic. It is getting a bit old. This is really a very complex issue, and trying to get all the important parts in one or two tidy messageboard posts without graphics is not really possible. Now, I am not trying to be nasty when I say this, just calling it the way I see it, but I think perhaps you don't have a great enough understanding of audio than you really need to make sense of all this, though you know some portions of the story very well, and have proved your skill at doing research on some of the topics, like the links you posted to the articles on human hearing and stuff. So the last thing I want to do is trivialize your knowledge, that is not my point at all. I just don't think you have the tools right now to see the big picture. For one last attempt at the topic of whether or not a 16/44.1k transfer is better or worse than a 24/96k trasfer downsampled to redbook for CD release, I offer the following:

1. I think we can say that, for all practical purposes, the master analog tape has an infinite number of samples/second.

2. 16/44.1k Redbook will have 44100 samples a second

3. 24/96k will have 96000 samples a second

4. A 24/96k transfer will kick the butt of a 16/44.1k redbook digital transfer in terms of dynamic range and frequency response

5. However, if you downsample a 24/96k transfer to 16/44.1k redbook, like they did on the Japanese Paper Sleeves and new Heart & Soul, etc, you will very likely lose data because instead of taking a 16/44.1k redbook transfer of a master with an infinite number of samples/second, you are taking a 16/44.1k redbook transfer of a master that has already had a ton of information removed from it when it was transferred at 24/96k. However, this is academic as far as old analog recordings are concerned because the equipment of the day didn't capture enough high frequencies anyway (look up the specs of the mics Elvis used, Bill Porter details all of them in his write up in the DCC CD artwork, and check the specs of the tape machines that were used in the 60s).

To apply this idea to another situation, think of copy machines. Let's say you go down to the local copy center and they have an old XEROX machine from the 1970s, and a brand new machine with much better copy quality. If you first make a copy of your document on the good machine, then make a copy of that copy on the old machine, you'd be worse off than if you'd just made a copy of the original on the old machine in the first place. There's no value added to first making a copy on the good machine and then making a copy of that on the bad machine. However, this is exactly what happens when one makes a 24/96K transfer then downsamples to redbook standard for CD release. The end result is a second generation copy, because both the first generation (24/96k) and second generation (16/44.1k) transfers discarded information. If the remastered CDs that were done at 24/96k sound better, it's not because they were mastered at 24/96k, it's because they were from a better source, used better encoders, used better noise reduction, or whatever.

I think I've about exhausted all I care to say on this topic for the moment.

Please take this at face value, it is just information and no malice is intended. Just because I have disagreed with you and said you have a ways to go to fully understand this doesn't mean I am attacking you. On the contrary, because of you, everyone has a lot of cool posts to browze, some of which may even help them learn a thing or two.

Greg

Wed May 21, 2003 9:08 pm

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:09 pm, edited 1 time in total.

Wed May 21, 2003 9:17 pm

Greg, I do agree that this topic has been exhausted.

You seem to understand the first part of my point, but not the second. Sometime I am going to post a graphic explaining exactly what I am saying. I tried to describe it, but some people don't understand. That is alright.

I actually consulted with my Dad(who is a very intelligent person and the king of all debaters). He did a drawing and and showed me how it could be better. I understand it fully and actually related it to my knowledge in Calculus. He is one of those old school guys that had a quad stereo, knows programming computers backwards and forwards(pre windows and on), can fix anything, and actually spent years on a compression scheme(if you looked at his work then you would think of the part in the Shining "all work and No play makes Jack a dull boy"). My dad had it figured out in 5 minutes. That is after he screamed at me to get my terms straight! Though to be fair, he said the difference is small.

We looked up every number and he confirmed what I felt all along. I even asked "how can that be?" He showed me example after example of how having more points is benificial EVEN if the same number is sampled. BUT it must be at a different place!! The filters are a factor, but the hard numbers shows that after a million samples(using the same RED BOOK VALUE IN DIFFERENT PLACES), a truer representation of the sound curve is gained.


How much is heard is another story all together. Comparing it to a copy machine is oversimplifying things. Though I decribed it accurately, one has to see it to believe it.

Anyway, good discussion. I will only peak at the answer to the quiz.

Wed May 21, 2003 9:28 pm

VInyl Junkie, You speak of ALL samples being filtered out that are past the PCM. UH I am not speaking of that!!!!!!

You said exactly what I though you would. IT is quantity!!! There is a limit to how much of the sound curve can be represented. I am not talking decibals, frequency or whatever.

SIMPLE SIMPLE.

Take a sound curve(THAT IS IN THE INTENDED RANGE OF THE CD!!!) cut it up.

Take the same sound curve, cut it up the same BUT also add 64 times as many cuts.

Now take equal intervals samples, twice.
EXCEPT on one take same intervals but starting from a different place. Say one bit before or after. Guess what in the one that is cut up more there are going to be better samples FOR EACH SAMPLE BEYOND THE FIRST CUTS. What is so hard to understand about this? You keep speaking about range but you fail to see the mathematical logic. I got news for you, when compared to a analog curve and digital sample can NEVER truly represent it.

THis has nothing to do with limitations it has to do with fundamental highschool knowledge!!

Wed May 21, 2003 9:31 pm

I still do not agree. Resampling something that was already sampled loses more data than sampling at the lower rate the first time, because the original sampling discarded data. Sampling a lossy copy can't possible provide a better result than sampling the original. The copy machine analogy did not over simplify it, just simplified it. A second generation copy is never more detailed than a first generation. No calculus can prove otherwise, although I had a calculus instructor prove that the turtle would eventually win the race against the turtle. Which we all know is a load. Feel free to post your calculus to prove this, instead of just talking about it.

Greg

Wed May 21, 2003 9:33 pm

"When sampling this at 44.1 you have to add a filter at 22.05. This leaves just an ordinarie sinusoidal curve with a frequency at 8kHz. The 3rd harmonic (24kHz) and higher will be filtered out. If these are not filtered out they will be above the limits of the Nyquist theorem and will aliasing in the 8kHz sinuswave. So they must be filtered out that is the basics for the whole PCM scheme. This is how a PCM tranfer works, not the way you describe it. "

This proves you have not read my post. I am not speaking of sampling beyond 44.1 khz!!!!!!!!!!!!!!!!! ONLY ON THE MASTER THAT HAS NOTHING TO DO WITH WHAT A CD SAMPLES. IT TREATS THE NEW TRANFER LIKE IT IS A REGULAR TRANFER. WHERE IT BEGINS SAMPLING IS THE IMPORTANT PART. LIMITING THE RANGES THAT ARE BEYOND THE FREQUENCY RANGE OF A CD. That still doesn't change the fact that 16 bit sampling at one place is not the same as sampling 16 bit at another starting point!!

Experiment

Wed May 21, 2003 9:45 pm

I just did a little test. I took a 16/48k (the highest rate my rig can do) transfer of a Steely Dan analog source. Then I took another transfer of the same exact song using 16/44.1k. I performed a spectral analysis of the two, and as expected the 16/48k showed much more high frequency components than the 16/44.1k.

Then I resampled the 16/48k to 16/44.1k. Spectral analysis of the resultant product demonstrated the worst high frequency response of all 3. A real world experiment, that verifies what I have said all along, that there is no value added to downsampling. If your weakest link is a redbook CD, there is no point in using a higher sampling rate earlier in the process*. Any gains made in encoding at the higher rate will be discarded during downsampling.

*Where the higher sampling rate does help is in editing. You can do far more effective trimming, pasting, noise removal, filtering and other similar operations at higher rates, and this is a great reason to encode at higher rates for CD release, but that's not what we are debating, and if you are only talking about basic digital transfers, there is no added value in oversampling first and downsampling later. Your product will be inferior to what you would get encoding at redbook standard in the first place.

Greg

Wed May 21, 2003 10:10 pm

The worst? That doesn't make sense! How is a sample from a Higher quality sample gonna look worse!

Whatever, your real world experiment "proved" me wrong. Though I highly doubt the software is a perfect example.

Incidently, shouldn't the 2 wave forms be the same? If nothing is gained or lost, the why would it differ. This goes against any computer program I ever used! Does't make sense at all.

I look at it the same as a picture. True when you get a big picture and then cut it down on a scanner, it does look worse then just scanning it outright. Too bad this is because the software is not meant for this!!! Horrible example.

Wed May 21, 2003 10:36 pm

DEL
Last edited by vinyljunkie on Fri Jan 23, 2004 3:10 pm, edited 1 time in total.

Wed May 21, 2003 10:49 pm

Greg in your example I was looking and you forgot one cruicial step. YOU DID NOT SAMPLE THE 16bit/44.1khz cd from the 44.1 khz PCM transfer. Instead you played it outright with nothing reading it. This is not a fair representation of a Red Book CD. You cannot skip this step!! Also you have to have a 64 times as much sample and not just a little bit better.

This is the point you are missing all along.

The calculus is what I described. Like using intregrals with a trapezoid rule. For ever shape withing the curve there is still a small point that is not caculated. It goes to infinite for every trapezoid!! This is one example. The more Trapezoids the closer to the original function. If you take any equal series of points out of the function it will be a better representation, compared with a function of less magnitude.

Vinyl Junkie,

"From what I read out of your posting it seems like the interpolation made is random. But it isn't."

Who said random? PRECISE 16 bit at the same intervals starting at different places!!!

"As long as the sampled curve is inside the limitations of the dataword. Increasing the lenght of the dataword is meaningless if your signal doesn't need the higher resolution. You cannot increase the dynamic range of a signal by increasing the lenght of the dataword. You will just add a lot of zero's!! "

There again who stated increasing the resolution of the RED BOOK CD!! I never said increase the dynamic range of the RED BOOK CD either!!!

This is futile. Both of you miss my point completely!! Greg as a little bit of a grasp, but Vinyl Junkie you are way off.

"You say that an anlogue curve cannot be truely represented digitally, this is untrue."

Are you serious? Maybe I should ammend what I said since I paraphrased and the meaning was lost. Yes a curve can be represented in some inferior way, BUT never 100% accurate!!!!! That is what I meant by "truely accurate". Take some trigonomotry and work on the basics. The functions of trig explain it all!!

"What to compare is a 1st generation 16bit/44.1kHz PCM master taken directly from the analogue tape. This must be compared with a DSD transfer converted to 16bit/44.1kHz. "

Why is this true? They are completely unrelated. What you should compare is a 16 bit master sampled from a pcm 44.1 khz(or 96) master).

In other words the transfer is there and now it has to read the bits!!!! 16 bit at a time. WHere it reads determines how accurate it is. If it deviates from any bit then the bit on a DSD transfer is going to be more accurate(provided it falls within the paremeters and it is not the same bit as the lower quality). That said then for every bit there is a 64 times more likely chance!!!!

You know what is funny, I keep saying the same things and nobody has really disproved this. It coming back to frequency or dynamic range, which has no bearing on the examples I have given because they fall within the the sampled paremeters!!
Last edited by genesim on Wed May 21, 2003 11:22 pm, edited 4 times in total.

Wed May 21, 2003 10:54 pm

"You forgot to convert the 64 bit curve to CD-DA!! You actually compare a 64bit master with a 16 bit. You must convert before you compare. What to compare is a 1st generation 16bit/44.1kHz PCM master taken directly from the analogue tape. This must be compared with a DSD transfer converted to 16bit/44.1kHz. "

NO DID NOT FORGET. Did you even think about what I said. Please do me this one favor VINYLJUNKIE.

Repeat what I think is true. I have been posting the same thing over and over. Please just repeat what I theorize then we can get somewhere. Through all these posts, you do not seem to know what I am even saying!!!!