enter image description here

I've done a lot of audio editing in Sound Forge over the years. Occasionally, my wave map is upside down. It doesn't affect the play back, but as an editor I get a little freaked out, especially because I feel I've lost control of the file. Does anyone know why this is happening, and more importantly, how do I control it.

I've inserted an example of a file with both types of wave forms

  • 1
    how do you define normal vs upside down ? Both seem to present some dc offset. – audionuma Jan 14 '16 at 10:10
  • Thanks for your comment ... I need to study more on DC Offset in Sound Forge to understand how to deal with it – Michael Jan 15 '16 at 19:41
  • It is not simply because the phase is inverted? It looks like that.. – Michael Hansen Buur Mar 18 '16 at 7:12

That's due to something called aliasing. You see you can hear from 20Hz-20kHz so you need to sample sound at twice that frequency. This is why CDs use 44.1kHz. Now I know what you're thinking. Your sound is sampled at 44.1kHz or 48kHz, and you're right but hang on.

Whenever you display your audio on your screen you're effectively resampling a sampled waveform. Your program is deciding to display each second of audio as a certain width on the screen. This corresponds to a certain number of pixels per second. So if you have a 1600x900 pixel monitor displaying 8 seconds of audio on a full screen you have 1600/8=200 pixels per seccond. So you're now viewing a signal that has been resampled to 200Hz which causes all kinds of nasty stuff. If you were to zoom in on the signal you would probably notice they're the same up close. If your program doesn't support that try an oscilloscope program.

200Hz gives you 100Hz bandwidth which is fine for most audio editing functions. You're supposed to be able to see relative volume at a glance. You're getting an idea of the envelope and amplittude over time. You'll be able to see any clipping that's occuring. What you're not going to be able to do is get a detailed view of your 44.1kHz wave form after it has been resampled to 200Hz.

All sounds are made of sinusoidal waves added to each other. So say you have a 1kHz sound wave at cos(2*pi*1000*t). If you start your samples at t=0 vs your sample at t=1/2000 (we're using seconds for t and radians for cosine) you get signals that are upside down(180 degrees out of phase/ or inverted polarity in electrical terminology). It's the same signal and sounds exactly the same because that t=1/2000 is just a .5ms time shift. If you swap the polarity on a speaker you can't tell the difference as long as you swap the polarity on all speakers at the same time after all. At each pixel you're sampling the waveform. The pixels don't line up the same way every time so you get an effective shifting in time. It just so happens that this particular signal is inverted when you resample and shift it using this particular resample and shift. This is because cos(x+pi) = -cos(x) (radians again). Cosine is called an odd function for this reason. What that equation says is that if you shift a cosine wave in time by a certain constant, it is equal to itself only negative. Notice something else interesting

cos(x+pi+pi) = -cos(x+pi) = cos(x)

This means that if you shift a signal by the amount required to invert it and then do it again, the signal ends up equal to itself. Also pi radians equals 180 degrees. That's why it is called a 180 degree time shift.

| improve this answer | |
  • Thanks for this excellent answer ... I need to study it but I have a sneaking suspicion I will be smarter at the end of it. – Michael Mar 17 '16 at 21:35

That's an asymmetric waveform, mostly found in male vocal recordings.

It would be a good thing to remove it, someone else would say it's not mandatory.

I think you'll be able to find a lot of plugins that do it OR run the signal through an outboard with a transformer and you'll be ok.

| improve this answer | |
  • I'm having a little trouble understanding dc offset. My technical expertise doesn't get that deep, although I've been editing for over 15 years.I will study it a little more and possibly get back to you with some questions ... thanks for your input, I truly appreciate it – Michael Jan 15 '16 at 19:42
  • "Found in male vocal recordings" ... What? There is no gender bias to DC offsets. DC bias is down to electrics, not sound pitch or timbre. – Rory Alsop May 17 '16 at 8:29
  • potato potato for studios. Many times used in audio as DC offset , many times could be wrong anyways . – frcake May 17 '16 at 8:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.