DC Offset not showing properly

GoldWave general discussions and community help
Cyberdude
Posts: 5
Joined: Wed Oct 17, 2007 5:04 pm

DC Offset not showing properly

Post by Cyberdude »

I was looking at a test wav file from a voice recording on my MP3 player. Under Goldwave, there is a solid bar showing over the signal. When I zoom in on a "quiet" spot (no other signal) it remains a solid bar until I zoom close enough to show a single line, DC offset.

As can be seen by the following pictures, Goldwave (ver 5.25) is not properly showing the waveform. It should appear as in the Cool Edit picture.

Whole (full) waveform looks like this in Goldwave:
http://www.4shared.com/file/76697996/49 ... eform.html

Zoom in to quiet area showing DC offset as a solid bar:
http://www.4shared.com/file/87479241/3e ... splay.html


Zoom in same area one more zoom level to show proper DC offset:
http://www.4shared.com/file/87479266/af ... splay.html

Same (full) waveform viewed through Cool Edit properly showing signal with DC Offset.

http://www.4shared.com/file/87513727/91 ... _file.html
Last edited by Cyberdude on Wed Feb 18, 2009 3:28 pm, edited 2 times in total.
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

Just a guess. ... It looks like there might be something funny about the file format, and GoldWave is not reading it properly.

I assume it also sounds bad when played in GoldWave?

Can you tell us anything about the file format? This is a WAV file? (VORC001.WAV) Is it a "regular" PCM WAV file? How many bits?

Did you make an "analog" recording from your MP3 player? If so, did you record it with GoldWave? CoolEdit?

The raw data in an 8-bit WAV file uses unsigned integers, and it's supposed to be "offset". But, the software should re-center it and you should never "see" or hear the offset.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

While I can't view your files because you used a file sharing site and not a picture sharing site, I can only speculate.

I will say this. Goldwave draws it's overall wave envelope as solid blocks up till around a 1:1 view. If you have an offset, and zoom out, it will look like either the positive or negative half of thw waveform is gone. If you've got a quiet section..since the waveform silences with a bias, goldwave will draw the zoomed out waveform as a block on either size of "inf dbfs" it's normal. CEP/Adobe Audition draws it's waveforms much differently. I will probably edit this soon as I get home and have a chance to look at your screenshots.
Cyberdude
Posts: 5
Joined: Wed Oct 17, 2007 5:04 pm

Post by Cyberdude »

DougDbug wrote:Just a guess. ... It looks like there might be something funny about the file format, and GoldWave is not reading it properly.

I assume it also sounds bad when played in GoldWave?

Can you tell us anything about the file format? This is a WAV file? (VORC001.WAV) Is it a "regular" PCM WAV file? How many bits?

Did you make an "analog" recording from your MP3 player? If so, did you record it with GoldWave? CoolEdit?

The raw data in an 8-bit WAV file uses unsigned integers, and it's supposed to be "offset". But, the software should re-center it and you should never "see" or hear the offset.
I recorded this file on my Sansa Clip MP3 player (it records this way in any setting). I can't hear the DC offset. The info on the file (as seen in the Goldwave editor) is:
"Wave PCM signed 16 bit, 24,000 Hz, 384 kbps, mono"

The PROBLEM is that Goldwave does not show the DC component as a straight line as it should (until you zoom down quite a bit). CoolEdit displays the waveform, with offset, properly.

I consider this a bug in the editor code. If you would like to look at it yourself, here is the link:

http://www.4shared.com/file/76697919/11 ... RC001.html
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

I consider this a bug in the editor code. If you would like to look at it yourself, here is the link:
I think you're right! I think you've actually found a BUG! :shock: I just used GoldWave's Effect -> Offset to create an offset in a known good file. I got similar results.

I suggest that you report this directly to GoldWave via the Web Support Form. Chris is very conscientious, and I'm sure he'll fix this.
GoldWave Inc.
Site Admin
Posts: 4372
Joined: Wed Mar 10, 2004 6:43 pm
Location: St. John's, NL
Contact:

Post by GoldWave Inc. »

Actually that's not a bug. GoldWave has been drawing the waveform that way for a while now.

When zoomed out, GoldWave estimates the shape of the waveform and draw it as a solid bar graph showing peak maximums and minimums. When a dc offset is high enough that both the maximum and minimum peaks are well above (or below) zero, GoldWave still draws the bar graph from the zero level. A solid waveform is an indication that you are not seeing an accurate waveform (just peaks levels around that time).

Zoom in 1:1 to see an accurate waveform.

Chris
Cyberdude
Posts: 5
Joined: Wed Oct 17, 2007 5:04 pm

Post by Cyberdude »

GoldWave Inc. wrote:Actually that's not a bug. GoldWave has been drawing the waveform that way for a while now.

When zoomed out, GoldWave estimates the shape of the waveform and draw it as a solid bar graph showing peak maximums and minimums. When a dc offset is high enough that both the maximum and minimum peaks are well above (or below) zero, GoldWave still draws the bar graph from the zero level. A solid waveform is an indication that you are not seeing an accurate waveform (just peaks levels around that time).

Zoom in 1:1 to see an accurate waveform.

Chris
Bug or not, the program is NOT displaying the waveform accurately! When I zoom 1:1 as you suggest, I cannot see enough of the waveform to see what is going on - therefore it is difficult to tell it's a DC offset.

The way it displays in CoolEdit is the proper way to display the waveform. Even looking at the entire waveform, it is easy to tell what the problem is. I didn't discover that I had a problem until someone with CoolEdit was able to tell me (see the linked CoolEdit display of the same waveform). This would have been impossible to diagnose with Goldwave - the way it displays the offset.

I like Goldwave for most of my work, but this is a problem. Another is trying to make a recording of unknown length - such as a speaker presentation. I use Audio Record Wizard for that. Goldwave just is not the appropriate program to use in this case either.

Goldwave is good for cleaning up noisy recordings, but is not the easiest thing to use (it's what we call in the industry - "expert friendly"). This is just another example. :wink:
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

Another is trying to make a recording of unknown length - such as a speaker presentation.
There is an option to record "Unbounded". If that's the issue... It will record until you hit the stop button, or I suppose, until you run out of disk space.

If you record too big of a file, you'll run into the WAV file-size limitation when if try to save the recorded file in WAV format. But, you can get around that by saving as MP, FLAC, etc.

It probably is easier to use a special-purpose recording program (rather than an audio editor) for recording.
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

Cyberdude wrote:Bug or not, the program is NOT displaying the waveform accurately!
Your screen doesn't have enough pixels to display the waveform "accurately". Some decision has to be made about what information is "lost", and how the remaining information is displayed.
GoldWave wrote:GoldWave estimates the shape of the waveform and draw it as a solid bar graph showing peak maximums and minimums.
I was confused by how the offset was displayed too, but I think I understand what Chris is trying to do... I have another audio application and it appears to display every "nth sample" (when you're zoomed-out). This will show offset correctly, and will give you a general approximation of the audio envelope. This might be what CoolEdit is doing. The problem is, that you can completly miss a short-term spike or peak, and (at some zoom levels) the waveform can look completely different depending on which samples happen to be displayed.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

Most audio programs will only show you nth sample when zoomed out..you gotta remember, there's 44,100 samples a second...if you're looking at a 5 second section, that's 220500 samples...the only way you can look at a waveform on a sample by sample basis is 1:1, for starters, I mean, if you think of it as each sample taking up one pixel, then you're only able to show (in my case for example) 1280 pixels at once across the entire screen. The program IS showing it accurately. Hold on, I'll have screenshot links soon as I'm done.

As far as "hearing" a DC offset...you can't. The only way you hear a DC offset is when the file starts playing and you get an audible click...otherwise a DC offset is just the waveform either on the positive or negative side of the zero crossing....in the real world...this translates in to your speaker being locked one direction or the other...sure...it will "sound" normal in this state, because the speaker is still moving both directions, but rather than resting at it's mid-point, it's being pushed/pulled to the value of the offset, and I will tell you now, if you come back saying one SHOULD be able to hear a DC offset, I will continue to discredit you, because it's simply not true.

So, here's the screenshots I just took.

Image

that's your file in goldwave. you seem to say it looks wrong, however, knowing goldwave, it's entirely correct for how goldwave displays the wave.

Image

this is an empty file with a smple .5 offset - this is how Goldwave shows the offset just one zoom level off 1:1 - it's a solid block going from zero-crossing to .5 - this is just how goldwave draws it's waveorms.

Image

here's an offset at 1:1 level - just a bar at .5

the way goldwave is showing the waveform in both of those files is normal, it's not a bug, it's not a problem, it's simply HOW goldwave interpolates data to show.

Image

this is your file in audition. now, if you look at this and then goldwave, it's natural to think that goldwave is displaying the information wrong...however, it's not. This is just differences between audition and goldwave. audition/cep for example, scans the waveform while loading, so it actually has all kinds of peak/valley/envelope information in memory where as goldwave tends to scan things in realtime as needed.

hope this solves your questions.

i will tell you, your offset is difficult to correct. goldwave is good with a steady constant offset, but yours is not, it starts out at zero and gradually climbs. while it will fix the majority of the offset, the beginning where the offset climbs...it will in fact STILL be an offset starting below zero crossing. i've never figured out an easy way to fix a variable offset like this and hope maybe chris does in the future.

for the record, there is NO way of automatically correcting an offset in audition, you have to manually calculate the offset value and punch it in.
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

As far as "hearing" a DC offset...you can't.
Of course, you are right. And, that's why I asked... It doesn't look like offset... It looks very distorted, and I was thinking something else was wrong.
i've never figured out an easy way to fix a variable offset like this and hope maybe chris does in the future.

for the record, there is NO way of automatically correcting an offset in audition, you have to manually calculate the offset value and punch it in.
Maybe high-pass filter? I had a "funny" waveform once. It looked like offset, but it was actually just asymmetrical. A constant offset correction didn't work, but a 20Hz high-pass filter took care of it.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

You're right.

This offset can be corrected by doing a high pass. My offset sample file shows a response up to 200hz.

Image

I'm also wondering, after examining the waveform...if the recorder he's using doesn't have some kind of sub 16-bit ADC chip...because the offset seems to gradually increase from the beginning and isn't exactly smooth, but tends to be a bit more stairsteppy/jagged, which would slightly indicate this kind of suituation.

Looking at the full spectral for the first second is even more interesting

Image

There's an odd bit of frequency up around 11khz...and this pulsing in the file. (Wait...22khz sampling rate, cheap adc...normal quantization noise - the pulsing as i've heard is strictly from the fact the wave is stair-stepping up.)
Last edited by DewDude420 on Thu Feb 26, 2009 12:25 pm, edited 1 time in total.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

I just thought of a reason Goldwaves envelope display is a better choice than how audition does it.

If you're a beginner and look at the wave in audition without playing close attention, there's a chance you won't think there's an offset problem, because you see both halves of the offset wave, however, the way Goldwave does it, it's quite clear that there's a problem.

Its actually quite accurate for that purpose...waveform accuracy isn't a requirement till 1:1, just about any engineer knows a full zoom out is only a rough draft. Goldwave is expert friendly...depending on the level of expertness you are at. It's a little thing called "learn how your audio software works rather than assuming it works the same as everything else"

Also remember....cep/audition = $350 DAW with a huge team of developers and Goldwave has just one.
Cyberdude
Posts: 5
Joined: Wed Oct 17, 2007 5:04 pm

Post by Cyberdude »

I just thought of a reason Goldwaves envelope display is a better choice than how audition does it.

If you're a beginner and look at the wave in audition without playing close attention, there's a chance you won't think there's an offset problem, because you see both halves of the offset wave, however, the way Goldwave does it, it's quite clear that there's a problem.
Couldn't disagree with you more. It's much easier to see an offset when it is displayed as an offset, rather than as a solid block. The solid mass implies transitions from zero to the peak of the waveform rather than a steady-state, fixed value (as shown in the Cool Edit picture). From the display, even on zooming, I was expecting to see some high frequency component. This is what is implied. One zoom click down from the magical 1:1 ratio still shows as a solid block.

About the only point I will concede is the cost factor. That said, it is more of an economical than technical choice.
DewDude420
Posts: 1171
Joined: Fri Mar 11, 2005 11:15 pm
Location: Washington DC Metro Area
Contact:

Post by DewDude420 »

this leads me back to what I said.

learn *HOW* your audio program works rather than assuming how it works

i learned goldwave did that YEARS ago. goldwave is an entirely different beast...you have to remember that it was around before a LOT of DAW software even existed.

oh...and whatever your voice recorder is...from a professional standpoint...is garbage. you shouldn't even HAVE an offset to begin with. it obviously is a subpar 8 or 12-bit chip...it uses offset to mask that.
Post Reply