Skip to main content

Hi friends,

what are anti-alias filters & DC offset. how important they are?

thnx

d-gautam

Comments

anonymous Thu, 08/19/2004 - 17:55

Anti-alias filters are devices built into any normal CD player that plays back regular CDs and CD-Rs containing audio recorded at 16-bit/44.1KHz. It cuts off the high frequencies above 20 KHz that are present in the audio.
The average human ear can detect frequencies between 20 Hz to around 21 KHz. People with exceptional hearing (which usually includes younger folks) can hear frequencies from !9 Hz to around 23 KHz.

The 16-bit/44.1 KHz format cannot reproduce freqs above 20 KHz, so the anti- aliasing filter keeps the frequencies in a range that the format can handle.

All this really means to you, is that if you want to be able to record and play back freqs above 20KHz, you have to record them in another format, such as 24-bit/96KHz. This format needs no anti-aliasing filter because of its increased sampling capacity.

DC offset can happen if your electrical power and your equipment don't play nicely with each other. While I don't understand the exact reasons this happens, I do know the problem that it causes. It decreases the headroom of your recording setup. I'm sure you have probably seen the waveforms that your music "makes" when doing waveform editing. (you know, the wierd squiggly lines that show your peaks.) If you have a problem with DC offset, you can see it in the waveform. The little thin line running horizontally straight through the center of your waveform (the "nothing point), will not actually be running down the exact center of your positive and negative waveforms. (positive is the spikes above the line, negative are the spikes below it.) The spikes on one side will therefore be longer than the ones on the other. So you lose headroom and cannot get as hot a signal as you should normally be able to without causing clipping and distortion.

I didn't mean to write a book here, but these are complicated subjects that require just as complicated answers.

Hope this helps.

Thomas W. Bethel Fri, 08/20/2004 - 03:58

I think we are talking about two different types of DC offset. The one you are referring to is the one that is present in amplifiers but the one most of us deal with on a daily basis is the DC offset in sound files.

DC offset is a source of even-harmonic, essentialy second-harmonic
distortion. That is, the positive-going and negative-going signals are not balanced around the zero center axis.

The effect of DC offset is negligible until one works on the the file since the ear is not as sensitive to even harmonics as to odd. Operations such as equalization and dyamic-range adjustment can magnify the distortion even though these processes may not alter or change the offset itself. Some plugins seem to generate more DC offset than others.

If you remember the old oscilloscope experiments you may have done in physic's class, you might remember a sine wave was half in the positive and half in the negative side of the center zero voltage line. The average position then, over a wavecycle, was zero if you added them together. This is analagous to 'alternating current' (A.C.) and is the sort of voltage one would see in power generation and the electricity you get from the wall socket. Battery electricity, on the other hand, is a flat voltage - called 'direct current' (D.C.), which would measure as a straight line offset from the center line on an oscilloscope (0 Hz).

Now, in digital sound-files, we can measure the component at 0Hz by measuring what the average sample value is, and how much this has 'drifted' from zero. Most sound files that have been recorded through analog equipment should have little, if any, DC offset. However, sometimes (due to digital processes) we may end up with a sound that does have a considerable DC offset.

Most software today can get rid of DC offset with a plugin. It should be done at the beginning of the work on the file and checked at the end.

Hope this helps.

-TOM-

anonymous Fri, 08/20/2004 - 04:44

I think we are actually talking about the same thing Tom. When I spoke about the waveform, I was talking about viewing it on an oscilloscope, just that it is built into the program or a digital multitrack device, instead of using a traditional hardware version.

When I spoke of the waveforms I was indicating that the positive and negative parts of the waveform are not balanced around the center axis.

You just managed to say it better than I did.

anonymous Fri, 08/20/2004 - 12:18

Anti-alias filters are in the A to D converters, they have to be used before the signal can be digitized. If a frequency above Nyquist gets into the system it will be interpreted as a lower, mathematically related frequency. In other words, frequencies above the range of human hearing could turn into something lower and therefor audible, obviously not good. This is the same reason car wheels and airplane props in a movie will sometimes look like they are going backwards or sitting still, you are seeing alias frequencies. Since very steep analog filters can cause all kinds of problems oversampling is usually used these days to allow the cutoff frequency to be much higher.

anonymous Sat, 08/21/2004 - 13:08

The reason that the anti-alias filter is need in the A-D converter is because a wave has both a positive and negative in its cycle. In other words, a sine wave of 1 Hz would take 1 second to complete its positive upswing and negative downswing. In order for that to be stored digitally, a sample of both the positive and negative must be taken to accurately recreate the 1 Hz sine wave. So the sampling rate to needed to recreate that 1 Hz sine wave would have to be at least 2 Hz in order to capture the positive and negative portions of the 1 Hz cycle. The anti-alias filter is needed ahead of the sampling process in order to filter out anything above the 1 Hz cycle because it would be misrepresented when played back because of the 2 Hz sampling rate that was used to capture the peak and valley of the 1 Hz cycle. Obviously, we don't use a 2 Hz sampling rate as I used it only as an example.

I hope that I explained this clearly.

-Erik