• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

When is a very slow or absent reconstruction filter not audible?

Multicore

Major Contributor
Joined
Dec 6, 2021
Messages
3,394
Likes
4,060
I was curious about a remark in @amirm 's video on the Bluesound Icon and posted it here but got no reply. So I restate it in its own thread and hope for your confirmation.

I understand the reconstruction filter to be fundamental in digital systems. Without it, or with very slow filter like Bluesound's, spurious (nonsense) signals will be present in the output of a DAC. So when Amir said "Is is audible, I can't tell you. Most likely probably not audible," I wondered, what are the conditions under which this will not be audible?

The spurious output signals will be above the Nyquist frequency which in the cases we're considering are ultrasonic, i.e. will have frequencies beyond human hearing. So it seemed to me that human hearing itself can provide the reconstruction filter so long as the system between the DAC and the hearing is linear in the ultrasonic range and not producing any nonlinear products in the audible frequency range.

Do I got that right?

We don't usually get detailed specs or measurements for the performance of equipment in the ultrasonic range. One can hope that it would be linear enough and maybe it usually is.
 
I was curious about a remark in @amirm 's video on the Bluesound Icon and posted it here but got no reply. So I restate it in its own thread and hope for your confirmation.

I understand the reconstruction filter to be fundamental in digital systems. Without it, or with very slow filter like Bluesound's, spurious (nonsense) signals will be present in the output of a DAC. So when Amir said "Is is audible, I can't tell you. Most likely probably not audible," I wondered, what are the conditions under which this will not be audible?

The spurious output signals will be above the Nyquist frequency which in the cases we're considering are ultrasonic, i.e. will have frequencies beyond human hearing. So it seemed to me that human hearing itself can provide the reconstruction filter so long as the system between the DAC and the hearing is linear in the ultrasonic range and not producing any nonlinear products in the audible frequency range.

Do I got that right?

We don't usually get detailed specs or measurements for the performance of equipment in the ultrasonic range. One can hope that it would be linear enough and maybe it usually is.
You are right - the artefacts are all above the audible range, so in theory you don't need to care. But that's only true if you are confident amplifiers and tweeters etc. will not be affected by the ultrasonics. If they are, you can get intermodulation back down IN the audio range. So that's why it must be filtered.
 
In case of no filter the HF droop that begins around 6khz is most certainly audible on 44.1khz sampling.
 
You are right - the artefacts are all above the audible range, so in theory you don't need to care. But that's only true if you are confident amplifiers and tweeters etc. will not be affected by the ultrasonics. If they are, you can get intermodulation back down IN the audio range. So that's why it must be filtered.
I wonder if we could come up with a practical demonstration of this using actual audio products.
 
In case of no filter the HF droop that begins around 6khz is most certainly audible on 44.1khz sampling.
In the case of the Bluesound Node Icon, they do implement compensation for the first order hold (aka sample and hold) early roll off, and its FR is flat to 20 kHz.

index.php
 
If it were an NOS DAC without filter it is down enough to hear if you have good hearing to the higher frequencies. So when is it not audible? When you are 90 years old? Such a DAC would be down more than 3 db at 20 khz and the droop starts around 5 khz beyond which you would hear it.

Of course there are various other filters people use even with oversampling DACs. Some have ripple which I don't know if any have enough to be audible. Some which are minimum phase have a bit of a droop one can hear in blind tests with quick switching though I rather have the opinion that with music you are unlikely to hear it in normal listening.

Here is the slow vs sharp filter in an SMSL which uses a different DAC than their other offerings. It isn't filter-less, but you can see it would be audible by younger listeners. Here is the DAC chip used. https://www.rohm.com/products/audio...audio-dacs/bd34301ekv-product#suppInfoArticle
1742053734697.png
 
I had a soundcard with no filtering once. I never noticed anything wrong and I was SHOCKED to see a stair-stepped waveform when I hooked-up an oscilloscope to do some experiments. (I don't remember what the experiment was about.)

But when I thought about it, I realized that the harmonics were above the audible range, plus the amplifier (in the powered speakers) might not go much above the audio range, and the speakers would be limited too, providing "mechanical" filtering.

Anti-aliasing filters on an ADC are more important... I'd say REQUIRED... I didn't check the input on that particular soundcard (and I had never even used the inputs). But even then, any ultrasonic harmonics (above the Nyquist limit) are likely weak so the aliases will also be weak and MAYBE not as bad as we might think.
 
Back
Top Bottom