Why do we need sharp filters if we do oversampling

Isn’t one reason of using oversampling to make the filter easier and more gentle? So why do dac chips still offer sharp filters?

If the filter is selectable, sharp is offered mainly for preference’s sake. There IS a real benefit but its audible benefit is debatable. The real benefit is feeding less very high frequencies to your amp. Those very high frequencies could be problematic for the amp to reproduce and affect it’s performance in the audible range.

1 Like

Doesn’t oversampling put the images further away? Let’s say we oversample a song with the sample rate of 44.1khz to 176.4khz. Wouldn’t a slow filter remove the images completely since we have more room for the filter? Why would we need a filter in this situation at all since the images are at very high frequencies?

I understand aliasing for Analog to Digital converters since the real world isn’t band limited but songs are. But for DACs, it’s confusing.

Sorry I don’t understand - what images are you talking about?

This thread talks about it:
The second picture shows images. It’s like aliasing but for DACs (or seems like it).

I don’t know, maybe I totally got it wrong.

What I understood is that reconstruction filter helps DAC produce the most accurate signal. It’s essentially oversampling + low pass fitler.

Now I don’t understand when DAC can operate at like 384khz, why is the digital low pass filter still sharp and starts at 20khz? Why didn’t they push the filter to like 100khz or make it slower?

That all depends on what the DAC’s designer decided to do. They’re not all the same.

1 Like

I don’t understand what you mean.
Oversampling in simple terms just means looking at the data more often to correct for errors in the output signal.

That way the circuit can produce all sorts of nasty noise beyond 20kHz and it does not matter to the output signal as it is just filtered out.
Makes it easier to design.

1 Like

By oversampling, I meant interpolation, not zero-order hold.

Oversampling gives more room for the filter to be more gentle and by oversampling, the nasty stuff goes way further, right?

In a 348khz sampled file, there’s no nasty stuff below 96khz. Whatever that is there is part of the music. If we want to put a sharp filter, why wouldn’t we choose a frequency like 80khz?

With NOS DACs, I understand the sharp filter at 20khz since for a 44.1khz song, there’s only 22.05-20=2.05khz room to filter nasty stuff. But what about OS DACs?