digital sound filter

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
so far i understand that a dac should have a "filter" , that turns a "crunchy" wave form of strait lines, into a smooth sinus with curved lines.
i think this is a very important circuit in digital sound.
but, if the speed of the digital system is high enough, the human ear wouldn't notice the little "crunchiness", and the digital sound will be as good as analog live sound.
at least for humans.
why then is this not used?
is this used?
 
This is what we normally refer to as interpolate. When we have a number of samples and we approximate the values between the known bits. Some people might call it a smooth function or the and approximation between sample, however the function is interpolate.
 
Most modern DACs do interpolate... it's actually cheaper to make a delta sigma ("1 bit") DAC than a non-interpolating R-2R type DAC. The latter type requires laser trimming, functional testing, etc... to verify it meets acceptable performance numbers. Delta sigma techniques do require a whole lot more gates to implement, but it's digital logic which can be made small, and there's a lot less analog hardware onboard that can get messed up during fabrication.

And the reconstruction filter required for a delta-sigma DAC is very basic; usually it's just a first order lowpass, while a non-interpolating DAC requires a very high order analog filter, which invariably requires very tight tolerance components with weird values.
 
DACs usually output picewise constant voltages or currents. These "square" waves include high-frequency information that you would want to filter out with an analog filter. With this filter you want to keep your audio signal intact while filtering out artefacts above fs/2.

In for exampel the CD format, fs is 44.1kHz, and your audio signal reaches very close to fs/2. This calls for a fairly advanced analog filter. However, if you manage to raise fs, the filter has to fall much less steeply between the upper audio frequencies as fs/2.

Sigma-Delta modulators are much cheaper to produce than are multibit converters. They require less analog circuitry and less matching. Using only a few levels (2 with one bit), you would expect signal/noise ratio to be really bad. However, feedback systems in sigma-deltas transfer quantization noise to higher frequencies where they are far from audio frequencies and thus fairly easy to filter out.

So in short: use digital oversampling to get a cheaper analog filter.

--
Børge
 
DAC really do not do interpolation, the digital filter does that. Any time you have an over sampling filter it will do some interpolation. Of course, the Delta Sigma DAC does have internal filters.

Interpolation; in terms I can explain draws or stretches the waveform, in concept a straight line between samples. Therefore, if we over sample a sample rate at 44.1K x 8 we effectively have 352.8K samples. So, the over sampling process stuff bits or sample between original samples, If the frequency is changing the sampling process will follow the change, however will add more sample between the original samples. The new samples are stuffed between the originals and follow the change indirection of the waveform creating a smoother transient. Hopefully.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.