Speeding up GOLDWAVE processing.

GoldWave general discussions and community help
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

:D More fun with "hex" files... (This is just for your general interest in digital audio. It won't help with your noise problem*.)

In GoldWave, you can save your file as Numerical Text. You'll get a .TXT file showing the data as normalized decimal values (between +1.0 and -1.0). **

Here are the first few samples from a couple of very short files:
[ASCII 44100Hz, Channels: 2, Samples: 6613, Flags: 0]
0.000000
0.000000
-0.000031
0.000000
-0.000031
0.000000
-0.000031
0.000000
-0.000031
0.000000
-0.000031
...
[ASCII 44100Hz, Channels: 2, Samples: 80, Flags: 0]
-0.194941
-0.020009
-0.302844
-0.020580
-0.378686
-0.018579
-0.425135
-0.017150
-0.466962
....


Those are both stereo files, and the left & right data are interleaved.

GoldWave can re-open these files as audio, so you could actually edit the TXT file with Notepad or Wordpad, and then save the changes. You might even be able to manipulate the file with Excel, but I'm not sure if Excel can re-save the data in .TXT format.

You can also save your audio file in MatLab format, which would allow you to do true DSP.

Another thing that might be interesting is to create a square wave or sine wave, and look at it as "numerical text" and/or the hex data. (Tools - Expression Evaluator -> ->Presets ->Waves -> Sine)

If you do that, I suggest you make very short files (maybe 1 second or less), and if you make a sine wave, don't go too high in frequency, or you won't have enough data-points per-cycle to visualize the shape of the waveform (1kHz is fine).



* I haven't listened to your files, but I understand DewDude's description of the problem, and I trust his assessment of the situation.

** GoldWave always uses (normalized) 32-bit floating point for internal processing, no matter what the original file format. So, this is just a decimal representation of the raw data in GoldWave.
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

DewDude420 wrote:I've told you OVER AND OVER you're not going to be able to completly fix those files...
Coriolanus,

A couple of thoughts that might provide some perspective...

Everything in the professional recording world is aimed at preventing noise and distortion. Because, you can't always "fix it up" with software.

Even with modern software and modern professional software, professional recording studios are still soundproof, and they still use very expensive, high-quality, low-noise, low-distortion equipment. (You really don't have to go to that extent... You can get good results with a home (or simi-pro) studio and moderately priced equipment. The key is to get a good recording to start with and prevent problems.)

Almost all "on-location" movie dialog is re-recorded in the studio. Radio and TV news broadcasts are done from soundproof studios, with good micropones placed close to the speaker's mouth to minimize noise pickup.

When news reporters go "live on the scene", you can usually hear the background noise, and sometimes you hear wind noise, even with a "wind sock" on the mic.

One of my "standard comments":
DougDbug wrote:- An excellent recording doesn't need any processing.

- A program like GoldWave can turn a good recording into an excellent recording.

- If you have a bad recording, the cure can be worse than the disease.
I hope this helps you to understand and accept what DewDude is telling you.
Coriolanus
Posts: 181
Joined: Tue Apr 29, 2008 3:50 pm

Post by Coriolanus »

DougDbug wrote:
DewDude420 wrote:I've told you OVER AND OVER you're not going to be able to completly fix those files...
Coriolanus,

A couple of thoughts that might provide some perspective...

Everything in the professional recording world is aimed at preventing noise and distortion. Because, you can't always "fix it up" with software.

Even with modern software and modern professional software, professional recording studios are still soundproof, and they still use very expensive, high-quality, low-noise, low-distortion equipment. (You really don't have to go to that extent... You can get good results with a home (or simi-pro) studio and moderately priced equipment. The key is to get a good recording to start with and prevent problems.)

Almost all "on-location" movie dialog is re-recorded in the studio. Radio and TV news broadcasts are done from soundproof studios, with good micropones placed close to the speaker's mouth to minimize noise pickup.

When news reporters go "live on the scene", you can usually hear the background noise, and sometimes you hear wind noise, even with a "wind sock" on the mic.

One of my "standard comments":
DougDbug wrote:- An excellent recording doesn't need any processing.

- A program like GoldWave can turn a good recording into an excellent recording.

- If you have a bad recording, the cure can be worse than the disease.
I hope this helps you to understand and accept what DewDude is telling you.
My problem does not concern something I recorded myself. It is from an audiobook and I think the transcription was made from Tape.

I have played around with ISOTOPE RX ( interesting program-- don't like the way they handle selections though)

I have had some luck with improving the files and then boosting the volume afterwards since agressive correction deteriorates the sound quality but boosting the volume seems to get around it.

I certainly feel that they can be improved over their current state - and I work on them a little bit everyday. I need to learn more about the various parameters available on various forms of noise reduction in Audition and Sound Forge. I just work on a very small piece of the audio and I do it in mono only - is this a mistake?

Your information on the digital files is useful and certainly looking at them digitally I find helpful. A lot of people, if they are not used to reading HEX or OCTAL, don't full grasp the real potention of working with the files in the digital format.

Audition sufferes from too much integration of the Photoshop UI & Tools- of course I think ADOBE tries to integrate all of their UIs but I have found some of the recent changes to the Photoshop UI not too great and I have never liked the way they handle OPACITY in Photoshop
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

Your information on the digital files is useful and certainly looking at them digitally I find helpful. A lot of people, if they are not used to reading HEX or OCTAL, don't full grasp the real potention of working with the files in the digital format.
But the question is do YOU understand enough about audio to be able to see sample information for the positioning on the amplitude scale it is and how to redraw it to what you want?

You don't seem to grasp that you're modifying this stuff in digital to begin with...going into HEX or OCTAL just complicates the processing becuase now YOU have to manually manipulate samples rather than some software that's been programmed how to do it.
Coriolanus
Posts: 181
Joined: Tue Apr 29, 2008 3:50 pm

Post by Coriolanus »

DewDude420 wrote:
Your information on the digital files is useful and certainly looking at them digitally I find helpful. A lot of people, if they are not used to reading HEX or OCTAL, don't full grasp the real potention of working with the files in the digital format.
But the question is do YOU understand enough about audio to be able to see sample information for the positioning on the amplitude scale it is and how to redraw it to what you want?

You don't seem to grasp that you're modifying this stuff in digital to begin with...going into HEX or OCTAL just complicates the processing becuase now YOU have to manually manipulate samples rather than some software that's been programmed how to do it.
Correct but reading HEX or OCTAL is about seeing error patterns -- I am used to reading 2-3 feet high data dumps and spotting error patterns. A pattern is a pattern -- it is all about taking a very small sample - converting it to HEX and then looking for patterns and editing those patterns and seeing what results you get when you play the sample back.

You see you wanted to learn all about Audio and thus you have and learned how to use the various tools in AUdition, GOldwave etc. and what all the littler parameters mean. I find that a long route. I want to just do a few simple things. I don't want to know all about audio -- only enough to fix my problems with the minimum of time. Good presets save me time.

Reading HEX saves me time -- because when you get used to reading HEX you don't really have to understand what the data input it - you have to learn to recognize error patterns. It is a bit like scanning a page of English text --and picking out spelling errors without actually reading every single word on the page or even understanding the content on the page.

You are able to see the errors in waveforms because you have been looking at them for years. I have been looking a HEX data for years too -- it really is a question of what we are used to doing and how are cognitive facilities react to what is before our eyes.

I know that cleaning up audio can be done to a very high level -- I have some old opera recording made from 1900 wax cylinders and from 1930's radio it is remarkable how they have been cleaned up to be playable on CD's -- they sound remarkably good. So obviously it can done -- the question is really what tools to use, and how effectively you can use these tools. I suspect that there are audio editing tools out there that cost thousands of dollars and put to shame Audition and all the others.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

no, see, i understand what you're saying

but what you're NOT understanding is you CANNOT see patterns in audio for error.

it's like this...noise is a RANDOM waveform...therefore you're going to have samples jumping ALLLLLLL over the amplitude axis....ok, now add in to this some voice and it's going to be ENTIRELY different to disginuish from the noise and actual audio...you have no reference points, you jsut have a bunch of numbers.

i've SEEN raw sample outputs in goldwaves TXT and, seriously, there's no way of being able to see the noise. sure, you MIGHT be able to find a pop and click once in a while, but if you're dealing with a rather consistant noise and a waveform on top, you're just going to get totally lost. do YOU have the understanding of human hearing to know how to manipulate the samples to make the speakers produce the appropiate vibrations you want?

you're running around in circles. wave form DOES NOT EQUAL HEX.

Seriously, you can yell at me till you're blue in the face about how you'll be able to see the errors in hex and i'll sit here and tell you till i'm blue in the face. "Noise" does NOT represent itself as noise...you ahve to understand you're dealing with analog information stored digitally in a sampled system.....the system has NO IDEA what in that sampled audio is noise and what's not....you're not going to be able to read the hex and do anything with it.

you seriously don't seem to understand anything we're telling you. your files are GARBAGE. for whatever reason they are. you CAN'T fix garbage. at all. YOU might think you can, but the real test is, is someone else going to be able to?

you seriously need to give up this HEX and OCTAL stuff...it's not going to work, you're just going to screw things up.

stick to programming.
Coriolanus
Posts: 181
Joined: Tue Apr 29, 2008 3:50 pm

Post by Coriolanus »

DewDude420 wrote:no, see, i understand what you're saying


i've SEEN raw sample outputs in goldwaves TXT and, seriously, there's no way of being able to see the noise. sure, you MIGHT be able to find a pop and click once in a while, but if you're dealing with a rather consistant noise and a waveform on top, you're just going to get totally lost. do YOU have the understanding of human hearing to know how to manipulate the samples to make the speakers produce the appropiate vibrations you want?
Just ran a test -- small 10 seconds of a mono file -- put output to Goldwave text and output from a professional Hex Editor. The file do NOT LOOK ALIKE AT ALL.

What Goldwave puts out is unintelligible - but what a professional hex editor puts out is something totally differentl. Hex editor output is in bytes and I have no idea what Goldwave is doing with this text file -- I suspect it is an ASCII file which is NOT the same thing at all as a HEX File.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

look... the point is....and let me make this as big as i can so you can see it clearly:

YOU CANNOT COMPLETELY FIX AUDIO USING A HEX EDITOR. YOU NEED TO KNOW *HOW* TO MANIPULATE THE SAMPLES TO RECREATE THE PROPER WAVEFORM. NOT ONLY THAT, BUT YOU NEED TO KNOW HOW TO IDENTIFY NOISE WHICH YOU CANNOT DO BECAUSE *NOISE IS A RANDOM WAVEFORM THAT HAS NO PATTERN* JUST BECAUSE THE AUDIO IS IN A DIGITAL FORMAT YOU *CANNOT* TREAT IT LIKE A NORMAL DIGITAL FORMAT, IT IS THE MOST UNDIGITAL-DIGITAL CREATURE YOU'LL EVER CROSS. I DON'T CARE WHAT DEGREES YOU HAVE, YOU WILL NEVER ACCOMPLISH WHAT YOU'RE DOING. WORKING WITH SMALL FILES WILL TELL YOU NOTHING.

but hey, if you REALLY want to give it a shot. go for it. I'm just telling you you're using the ENTIRELY wrong method of doing it. You want to edit hex but you don't realize is you have to know how to edit the samples...recreate the waveform...look at the sample data and go "this is noise, this isn't. these two samples need to be here and this one needs to be here and that series should be here and this and that" and, seriously, most software has to use frequency analysis...which doesn't even generally cover a sample by sample basis. it just can't be done.

it is SERIOUSLY much easier to look at some bad waveforms and learn what you're looking for. noise is almost impossible to see, you need to use the spectrogram to really find out where it is. but seriously, stop trying to defend your method, i've already written you off as one of those people that have absolutely crazy ideas on how to do something because you can't accept that you just CAN'T do it. digital audio isn't perfect, how many times do we have to tell you that...it's the ONLY thing in the digital world that will have interference...ok? it's a digital repsentation of an analog waveform and really should be treated like one...you work on the wave form, NOT the digital data...that's like trying to write something like windows in pure machine language.

and i love how you say your hexeditor is "professional", as far as i knew a hex editor could only do so many things and most of the "professional" fetures had been covered by free open-source offerings. obviously if you're trying to elevate your software to a level it shouldn't then i'm not even sure WHY i'm arguing with you at this point..you won't get it, you'll NEVER get it.
I suspect it is an ASCII file which is NOT the same thing at all as a HEX File.
right, it's an ASCII output of sample data. your hex editor is simply opening the original wav. it's the SAME DATA in different formats.

and AFAIK, there's no such thing as a HEX file...any file can be a "hex" file when opened in a hex editor......
have fun anyway.

[/quote]
piano nick
Posts: 423
Joined: Wed Jun 16, 2004 8:33 pm

Post by piano nick »

I probably shouldn't wade into this conversation but I will. I don't have nearly the knowledge of digital of you three, but there may be something I can add.

There is a problem removing noise and there always will be. There is also a great deal of effort by some in separating all the intruments of an orchestral piece into their individual tracks and this is essentially the same problem. If you can remove noise, then you can remove the oboe from the orchestra.

Example: Let's record the A440 note from a saxophone and break it into all its partials (overtones). Each partial will be a perfect sine wave, and each will have a different frequency and amplitude relative to the other partials, which gives each instrument its unique sound.

Now record an oboe playing A440, and break the A440 note into its partials, and compare the two fundamentals (A440) of the two instruments. They will be indistinguishable because they are both perfect sine waves with a freqency of A440. One may have a greater amplitude (loudness), but their shapes are both sine waves and are identical in shape.

Now play the two intruments together and record them, and break the sounds down into pure sine waves. The first will be the fundamental (A440 frequency), but there is only one wave of A440. How much do you attribute to each horn? Well, you really can't tell how much is oboe and sax.

Now try to separate the second overtone, etc. If the fundatmental is impossible to sort out, and the overtones compound the problem, how far will you get?

Considering that noise is almost certain to have some frequencies that are identical to some of the frequencies of the music, how easy is it to remove the sound without removing some of the music?

If you know "exactly" the frequency pattern of the noise, and the exact volumes of each it may be possible, but herein lies the problem - noise is seldom perfectly uniform in frequency and volume.

Once noise is recorded into music, it's virtually impossible to completely and perfectly remove. This is why soundcard makers strive to get SN ratios over 100 - the noise will not be noticeable, but it will be there.

Every component in your signal chain is vital, including cables, connections, soundcard, microphones, the whole works. One weak link, and the chain is broken.
Coriolanus
Posts: 181
Joined: Tue Apr 29, 2008 3:50 pm

Post by Coriolanus »

piano nick wrote:I probably shouldn't wade into this conversation but I will. I don't have nearly the knowledge of digital of you three, but there may be something I can add.

There is a problem removing noise and there always will be. There is also a great deal of effort by some in separating all the intruments of an orchestral piece into their individual tracks and this is essentially the same problem. If you can remove noise, then you can remove the oboe from the orchestra.

Example: Let's record the A440 note from a saxophone and break it into all its partials (overtones). Each partial will be a perfect sine wave, and each will have a different frequency and amplitude relative to the other partials, which gives each instrument its unique sound.

Now record an oboe playing A440, and break the A440 note into its partials, and compare the two fundamentals (A440) of the two instruments. They will be indistinguishable because they are both perfect sine waves with a freqency of A440. One may have a greater amplitude (loudness), but their shapes are both sine waves and are identical in shape.

Now play the two intruments together and record them, and break the sounds down into pure sine waves. The first will be the fundamental (A440 frequency), but there is only one wave of A440. How much do you attribute to each horn? Well, you really can't tell how much is oboe and sax.

Now try to separate the second overtone, etc. If the fundatmental is impossible to sort out, and the overtones compound the problem, how far will you get?

Considering that noise is almost certain to have some frequencies that are identical to some of the frequencies of the music, how easy is it to remove the sound without removing some of the music?

If you know "exactly" the frequency pattern of the noise, and the exact volumes of each it may be possible, but herein lies the problem - noise is seldom perfectly uniform in frequency and volume.

Once noise is recorded into music, it's virtually impossible to completely and perfectly remove. This is why soundcard makers strive to get SN ratios over 100 - the noise will not be noticeable, but it will be there.

Every component in your signal chain is vital, including cables, connections, soundcard, microphones, the whole works. One weak link, and the chain is broken.
But it is possible to "clean up" old wax cylinders and old recordings from the 1930 so that they sound, while not modern, definitely 1000% better than they used to. So obviously there are technicians, and equipment that are able to remove a lot of noise from these old recordings. I have heard these recordings in the original and heard the cleaned up versions and yes you can distinguish the different instruments and MOST of the noise is gone.

So Yes it is possible to clean up sound recordings -- what they use I don't know about I suspect that it is worlds away from Audition and Sound Forge and Gold Wave and costs many thousands of dollars.

My point is that it is possible -- Dewdude thinkgs it is not - but obviously it is as it has been done and is being done all the time with very, very old recordings. And part of cleaning up old recordings is distinguish the sound of one instrument in an orchestra from another.

Will it sound like a modern recording in todays sound studios - NO - but it will be very, very good in relation to what it was and vast amounts of noise have been removed.
Coriolanus
Posts: 181
Joined: Tue Apr 29, 2008 3:50 pm

Post by Coriolanus »

DewDude420 wrote:look... the point is....and let me make this as big as i can so you can see it clearly:


and i love how you say your hexeditor is "professional", as far as i knew a hex editor could only do so many things and most of the "professional" fetures had been covered by free open-source offerings. obviously if you're trying to elevate your software to a level it shouldn't then i'm not even sure WHY i'm arguing with you at this point..you won't get it, you'll NEVER get it.
A free hex editor bears the same relationship to a Professional hex editor that Audacity does to Audition. But if you have only worked with Free Hex editors when you would not know that. :evil:
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

But it is possible to "clean up" old wax cylinders and old recordings from the 1930 so that they sound, while not modern, definitely 1000% better than they used to. So obviously there are technicians, and equipment that are able to remove a lot of noise from these old recordings. I have heard these recordings in the original and heard the cleaned up versions and yes you can distinguish the different instruments and MOST of the noise is gone
i hate to tell you...but the majority of that probably wasn't digital post processing. there are things you can do during playback of old formats to reduce the amount of noise.

the other thing you have to consider is the type of noise on a wax cylinder vs the type of noise in your files. Yours isn't noise..it's more like distortion, distortion that IF SOMEONE HADN'T OF DONE NR TO BEGIN WITH, it MIGHT be able to be removed.

the programs the pros use? here's a small partial list i can remember: Pro-Tools, Sound Forge, Audition, DARTPro, DC7, other custom created software, Antres plugins, izotope plugins, Sony Suite plugins....the list goes on. I've got MUCH of the same stuff the pros use. So don't go thinking the pros have some magical stuff...beucase they don't. They just know how to use it, and even they would say your files are unfixable.


Piano Nick is right, he's EXACTLY right, I've been trying to say something similar to that for a while. you CANNOT seperate elements of a waveform, yes, it's digital, but it's STILL Based off an analog format. it's much in the same way you can't just click on a person in a photo and remove them, sure, you CAN, but what about the photo you took them out of....there's no background....there's no information from where they were standing on what's behind them.

audio is the same way...the programs don't know how to remove things...there's no way looking at the hex you'll magically go "oh, look, a saxophone", all you're going to see are a bunch of samples that are going to make up a squiggly line that when turned into analog represents the movement of a speaker cone to produce sound..ok...so really, i cannot even begin to imagine someone who only has a degree in programming will be able to manipulate those samples because you know how to make that speaker vibrate the right way...you don't. sure, you might know a speaker vibrating 1000 times a sec makes a 1khz tone, but what about if you throw some noise into it? then you've got your speaker moving 1000 times a second and THEN noise that's being slightly modulated like that.

you said yourself you don't know how to read a waveform...that right there disqualifies you from being able to do anything with raw sample data in hex...plain and simple. you HAVE to know one to know the other, seriously...looking at a connect the dots without the dots connected doesn't really help even if you know WHAT it is.

are you getting it, is it getting ANY clearer why your idea of using hex is just so stupid it's GUARNTEED to be complete utter fail?

do you have any idea why i've thus far discounted ANYTHING you've put into the forum as serious?

do you get WHY i'm being such a prick?

you apparently DO NOT understand the basics of audio, and untill you do, you're not going to be able to do anything....WORKING IN HEX WILL TELL YOU NOTHING.

Seriously...anyone who DOESN'T have to sit here and yell like that but still continues to do it MUST know something, yet you think you know more. You don't. I'm sure everyone is agreeing with me that your idea is completely bonkers. I've seen proper Guiness pours with a better head on 'em.


the point is...give it up. it'd be easier for you to find another source (or better yet, go out and actually purchase the CDs)
mh
Posts: 133
Joined: Thu Aug 10, 2006 6:20 pm

Post by mh »

Bottom line - you might think there's a pattern, but the pattern isn't there. If you have a portion of a file that you know has a silent background, cool, you can analyse that and say "that's my noise", BUT the nature of noise is that it's a random fluctuation. So your noise "pattern" is only relevant for the portion of the file containing it. It might come close for other parts of the file, or it might be way way off.

All you have in a wave is the sum of signal plus noise, and this is just a number. What two numbers when added together will give you a result of 638? You can pick an infinite amount of numbers. In a 16 bit wave file there's over 65000 combinations. How do you know which of those two numbers is the signal and which is the noise? And that's only relevant for that particular sample. If you then pick another part of the file, say it's 17452, how do you know the 2 numbers you chose to get 638 are any way relevant?

Like I said, noise is random. It's been proven to be random, and trying to pretend it's anything else is a waste of your time. Any pattern you think you might see is an illusion.

There are some small things you can do. If you know that your noise is mostly focussed in a certain frequency range, you can apply some EQ to accentuate frequencies where it is not. The end result might sound good, it might not. It's entirely up to you if it's tolerable.

Generally, if the noise is not too bad, the best thing is to just leave it. Better minds than any of us have never been able to come up with an infallible noise reduction process.
The Great Watbol!
Posts: 34
Joined: Mon Sep 22, 2008 5:07 am
Location: In the back of your mind(s)

missing files.

Post by The Great Watbol! »

*edit: the files in question (so everyone knows) are here:
http://jay.is-lost.org/22.mp3
http://jay.is-lost.org/23.mp3
.
.
.
I tried to click on them, they're gone.
What a drag.
------------------------------
¿¿¿-:·}~ «-{Watbol}-» ©2000® ☆*★*☆
(•¿•)
Post Reply