R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 3 [4]   Go Down

Author Topic: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse  (Read 9506 times)

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #45 on: October 07, 2005, 06:06:17 PM »

Johnny B wrote on Fri, 07 October 2005 20:36


IIRC, someone was just talking about using 80-bit or greater float in an EQ product.

IIRC, the poster thought that partcular EQ was among the best he'd ever used.



Are you being deliberately obtuse?

Why can you not understand that just because something produces improved results in one place, it does not improve results everywhere?

if you widen a busy road, the traffic flows easier.

If you widen a rarely used road, the traffic flows no easier, but you spend a lot of money which could have been spent on other roads, tarmac over a lot of nice green land, and probably have to knock down a couple of people's houses that were in the way, so far from improving things, you've made things worse to no purpose.

Equalizers are a mathematical equation, equalizers as implemented in hardware (digital OR ANALOGUE) are equations with errors, the task of the designer in both realms is to design an equalizer in which the errors are kept either so low that they can be disguarded, or kept as low as possible given physical or financial constraints.

In analogue some circuit configurations can sound great, but because of their complexity, they require the finest possible most expensive precidion made laser trimmed components for them to work, because the nature of the circuit means any errors in those components are massively magnified... other circuit topologies may have certain disadvantages (for example they may screw up the phase), but in situations where that is acceptable (or even desirable) they can be built to a very high audio standard using much cheaper components, simply because the nature of the circuit means that errors are not amplified.. and may perhaps even cancel out. In those circumstances better components are simply a waste of money, since they make no difference.

Similarly with digital processing, some configurations require high levels of accuracy to implement, others not only do not require it, BUT GAIN NOTHING FROM IT. It is true that there are some filter topologies which require high accuracies to work properly, there are also other which do not, much as with analogue circuitry, different filters sound good in different situations.

This is not black magic, it's engineering, and the domain, whether analogue or digital (I've worked in both), is irrelevant.
Logged

Johnny B

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1134
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #46 on: October 08, 2005, 12:17:28 AM »

Well you talk about "errors," but some errors can sound good, despite what the theory or generally accepted specs may say.  Some errors we like, some errors we hate.

As far as you bringing up the cost aspects, the trade-offs and whatnot, you talk just like a million double e's I've dealt with, but the ones I liked the best were always the ones who did not let cost considerations get in the way...Granted, that was a luxury that not everyone enjoys.

As for the "road analogy," I like your concern for environmental values and preserving people's homes (in real life I share those values and concerns with you) but the data path as roadway seems apt. You seem to speak of a wide-lane, wide open highway. Like the Autobahn, where there is no posted speed limit...the actual speed depends on a variety of environmental conditions and other factors...but doing all those intermediate steps look like road blocks and stop lights...Removal of road blocks, yellow caution lights, and red stop lights would seem to have the potential to increase overall efficiency and system performance.

BTW, it may have been Bob K. who brought up an EQ which uses an 80-bit float. Might have been a little more than 80-bit...I already forgot...sorry...



 
Logged
"As far as the laws of mathematics refer to reality,
they are not certain; as far as they are certain,
they do not refer to reality."
---Albert Einstein---

I'm also uncertain about everything.

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #47 on: October 08, 2005, 05:08:43 AM »

Johnny B wrote on Sat, 08 October 2005 05:17

Well you talk about "errors," but some errors can sound good, despite what the theory or generally accepted specs may say.  Some errors we like, some errors we hate.

As far as you bringing up the cost aspects, the trade-offs and whatnot, you talk just like a million double e's I've dealt with, but the ones I liked the best were always the ones who did not let cost considerations get in the way...Granted, that was a luxury that not everyone enjoys.

As for the "road analogy," I like your concern for environmental values and preserving people's homes (in real life I share those values and concerns with you) but the data path as roadway seems apt. You seem to speak of a wide-lane, wide open highway. Like the Autobahn, where there is no posted speed limit...the actual speed depends on a variety of environmental conditions and other factors...but doing all those intermediate steps look like road blocks and stop lights...Removal of road blocks, yellow caution lights, and red stop lights would seem to have the potential to increase overall efficiency and system performance.

BTW, it may have been Bob K. who brought up an EQ which uses an 80-bit float. Might have been a little more than 80-bit...I already forgot...sorry...



Johnny,

Firstly I'm guessing that you must be the only person reading these threads who has missed the number of times I've talked about pleasant sounding errors and unpleasant ones. Unlike you I also happen to know what errors can occur where... IN BOTH DOMAINS.

Secondly I am a real engineer working in the real world, but one who takes a great deal of pride in the quality of his work from all aspects, including the sound quality of what comes out, some systems I do to cost and performance constraints.
Some I do to no constraints other than quality

I happen to know how to do both, I also know how to get the best performance from a system even if quality is the prime consideration.

This is what I do, it is my job, I am very good at it, I am what some people call an expert at it.

From the things you have said in this forum I can only conclude that it is not your job, you know nothing about it. You don't know what the audio algorithms are, you don't know the different ways they could be done, and you don't know about processor architectures, pipelines, caches, multiple execution units, SIMD processing or any of the other things involved.

You are guessing, I am not.

I am not perfect, I do make mistakes, I do have new things to learn, but the level at which you are implying my knowledge is, is just plain insulting.

Now the thing is, you don't have to take my word on this, you could go off and talk with other people who know and confirm the things I've said, you could even learn more about the subject yourself and confirm the things I've said.

Instead you simply repeat the same viewpoint again and again without reconsidering it once, and in doing so imply that I don't know what I'm talking about.

Your stance is that if we go to higher sample rates and wider data words, then things will magically improve. Despite the fact that I and others who know infinitely more about the subject than you do have pointed out the falacies in your thinking a thousand times, you continue to repeat the same view point, because you "have a feeling".

This is not a discussion, it is not a debate, we're not even beginning to drift into the areas of audio processing and conversion where there is a debate to be had.

If it wasn't for the fact that I care enough about this subject, and the industry, to not want people reading who are still learning (and unlike some are still open to learning) to actually believe and propogate the various flawed hypothesis you put forward, I wouldn't even bother.
Logged

C-J

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 80
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #48 on: October 08, 2005, 12:35:18 PM »

Jon Hodgson wrote on Sat, 08 October 2005 12:08

If it wasn't for the fact that I care enough about this subject, and the industry, to not want people reading who are still learning (and unlike some are still open to learning) to actually believe and propogate the various flawed hypothesis you put forward, I wouldn't even bother.

Jon,

I want to give you a big, public THANK YOU for taking the time to write your very informative and long posts here, and on other threads. I'm sure that everybody who has read them feel the same!

Thanks,
C.J., Finland
Logged

Johnny B

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1134
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #49 on: October 08, 2005, 12:47:11 PM »

Jon,

I did not mean to insult you...sorry if you were offended.

I thought I was merely asking some questions...in fact, I believe I asked you what kinds of proto's and experiments you might try in light of the new tech coming our way.

IIRC, someplace you agreed that the float could be expanded or improved due to the potential for errors when constrained with using the currently popular systems.

Logged
"As far as the laws of mathematics refer to reality,
they are not certain; as far as they are certain,
they do not refer to reality."
---Albert Einstein---

I'm also uncertain about everything.

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #50 on: October 11, 2005, 09:33:17 PM »

Johnny B wrote on Fri, 07 October 2005 15:36


IIRC, someone was just talking about using 80-bit or greater float in an EQ product.

IIRC, the poster thought that partcular EQ was among the best he'd ever used.






But this particular EQ uses frequency-domain processing to accomplish a linear phase EQ. The cumulative errors over many stages can get pretty bad. One never knows if you need that many bits unless you study the architecture of the equalizer. Other models of EQ get away with fewer bits and sound good...   Please don't quote me as saying or implying that EVERY equalizer would sound better with 80 bits just because the Algorithmix Red does and its designer believes that it needs it.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Tomas Danko

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 4733
Re: A Challenge To Nyquist? --- A "Better Than" Nyquist Pulse
« Reply #51 on: October 16, 2005, 08:43:45 AM »

Maybe it'd ease Johnny B's mind a tad bit if we just reminded him that in order to improve digital audio at this moment we need to improve the analog interfacing between the physical world and the digital representation of the audio. Digital theory is just fine as it is, already. Just compare, say, a cheap Creative Labs Soundblaster audio card to, say, a Lavry Gold and you'll see that's where the big difference is to be found.

Next, another reminder about where we currently are in terms of processing power might help. With the available computing resources of today we need to weigh higher sampling rates against lower with respect to the actual decimation process.

A higher sample rate will give you a lot less computational power to decimate the incoming audio stream with accuracy when creating the actual digital values.

So with what we have today in terms of computers, hardware and what not, it's a safe bet to spend the big bucks on the analog part of the ADC and keep sample rates fairly "low" (ie 96 kHz, I guess Johnny B calls that a low value) so that the decimation process can be done with the highest precision possible.

This will most likely render the best digital audio possible today. Don't go blaming digital theory in itself.


I'd humbly suggest Johnny B to wait a while until we have got recources enough to actually be able to perform even better decimation processing than today, at the higher sampling rates he suggests. Meanwhile, let's hope that the analog engineering also improves the front- and backend of the systems.

Sincerely,

Tomas Danko
Logged
http://www.danko.se/site-design/dankologo4s.gif
"T(Z)= (n1+n2*Z^-1+n2*Z^-2)/(1+d1*z^-1+d2*z^-2)" - Mr. Dan Lavry
"Shaw baa laa raaw, sidle' yaa doot in dee splaa" . Mr Shooby Taylor
Pages: 1 2 3 [4]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.051 seconds with 17 queries.