Well, a DC shift that is calculated for every sample (or updated every sample), is a filter.

I would recommend the high-pass route.

Hello, since i played with de-emphasis i stumbled upon a New Age cd from 1987 of Steve Roach and others. On this sampler every song has it own massive! dc-offset.
I´d like to prepare the whole image in one go.
I know a dcshift as sox can do doesn´t do it best cause it shifts the whole file by the same value so the songs still will have different DC. Besides that you can´t specify that shift for the channels seperately.
One way recommended by many is to add a highpass at low frequency. This works well but i don´t like the idea of using another filter.
Now i stumbled upon a free plugin called DC Killer! from Tobybear. His plugin has different approaches to remove over the time varying DC.
The first and to my understanding simplest method of his plugin already works perfect but i don´t really understand what i do with it
In his manual he describes this method as "comparison and substraction of incoming signal with last sample"
Can anyone of the Pros tell me in simple words what this can be? I understand it as a kind of DCshift that is calculated for every sample. No idea if this is good, bad or even ingenious.

Any input welcome!

Edit: The plugin is Donationware, sorry no freeware as mentioned above.

In Adobe Audition or CoolEdit you 'amplify' 0 dB but check the 'DC bias adjust' to take it out.

G²

I know a dcshift as sox can do doesn't do it best cause it shifts the whole file by the same value so the songs still will have different DC. Besides that you can´t specify that shift for the channels seperately.

Can anyone of the Pros tell me in simple words what this can be? I understand it as a kind of DCshift that is calculated for every sample. No idea if this is good, bad or even ingenious.

Besides that i don´t use these programs i wonder what approach it uses to remove DC

Besides that i don´t use these programs i wonder what approach it uses to remove DC

?Select Absolute and then specify the final DC percentage in the L and R boxes. This option lets you cancel out DC that’s not constant throughout a waveform. An extreme low?cut filter achieves this result. Keep in mind that the actual amount adjusted varies with each sample. For example, if you have a significant DC change in one area of the wave, at that boundary where the DC changes, the Absolute option makes all parts the same. (However, there will be a dip or peak right at the boundary point.) To introduce a DC bias by skewing the entire selected waveform above or below the center line, enter a positive or negative percentage. For example, a setting of 50% moves the entire waveform up halfway, and one of ?50% moves it down halfway.

?Select Differential and click Find Zero Now to analyze the entire selected area to get the DC offset, and adjust every sample by the inverse of that exact amount. The correct L and R percentages are entered automatically.

You can't calculate the "DC level" for each sample...

My knowledge of DSP is limited, but I think a low-pass filter is easier to understand than a high-pass filter...  A low-pass filter is essentially a moving-average...  Do you inow how a moving average works?  (For example, take the average of samples 1 through 5, then take the average of samples 2- 6, then samples 3-7, etc.).  You can see how a moving average would "smooth-out" (filter) any fast (high frequency) changes.

I wouldn't try to correct for the DC offset problem in your digital file(s), because the analog stages your signal passes through during playback will remove the DC offset (continuously) by virtue of having coupling capacitors.  I don't think you'd enhance the sound quality of your recording by using any of the techniques mentioned in this thread.  I believe if anything, you might degrade the sound quality, although probably to an imperceptible degree.  This would make for an interesting listening test, if you were to post lossless clips from before and after your DC offset correction.

If not, I might try to do something like this by adding a DC offset to a recording that has no DC offset to begin with, and then go from there with the software you found useful.

I'm glad you were able to find a solution to the DC offset problem.

I was wondering if we could set up a listening test for anyone on HA who's interested in participating.  If you're able to post some samples, that would be great. 

If not, I might try to do something like this by adding a DC offset to a recording that has no DC offset to begin with, and then go from there with the software you found useful.

If you're talking about audio files, the best approach is to go through the whole file, sum all samples and divide by the total number of samples and then subtract that (with dither, and watch for clipping) from each sample.

If you're talking about audio files, the best approach is to go through the whole file, sum all samples and divide by the total number of samples and then subtract that (with dither, and watch for clipping) from each sample.

And yes when using dcshift on the seperated files there are clicks that must be removed between files.

That will not work unless the offset is constant throughout the file. If the offset is not constant, you can get a "click" at the beginning/end which is exactly what happened to Wombat (and to me):

No, do not use dither. Just round to the nearest integer and subtract that value from each sample.