• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

Merging Technologies founder talk - digital filtering and other things about listening

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
Interesting talk from Merging Technologies founder.

Especially the DSP part.

I'm not an expert but am I correct that he has a personal preference for slow roll-off linear phase filters, over fast roll-off?

For better time domain performance?

Or have I not understood clearly? Please correct me.

 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
The moment someone says "time domain performance" is when you know can stop paying attention. Nothing meaningful ever follows.

Hilarious. Can you please expand on why.
 

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
The moment someone says "time domain performance" is when you know can stop paying attention. Nothing meaningful ever follows.

Agreed. That said, Merging's Anubis is like a multi-channel friendly ADI-2, with as many features (including, importantly, advanced bass management) and expandable I/O. That's pretty progressive and not very expensive for what it is. I would love to see that measured.
 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
(including, importantly, advanced bass management)

How would that work with something like Dirac?

If it doesn't feature measurements, what bass management is possible? Or you can important files for the bass management?
 

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
How would that work with something like Dirac?

If it doesn't feature measurements, what bass management is possible? Or you can important files for the bass management?

I mean bass management loosely as in adjustable low-pass frequency (20-200Hz) and slope (up to 36dB/octave) for individual subwoofers.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Hilarious. Can you please expand on why.
Because if time domain is the goose, then frequency domain is the gander. What's good for one is good for the other. The phrase "time domain performance" is only ever seen when someone is defending shoddy reconstruction filters by pointing to a "pretty" impulse response without any of that ghastly "ringing."
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Agreed. That said, Merging's Anubis is like a multi-channel friendly ADI-2, with as many features (including, importantly, advanced bass management) and expandable I/O. That's pretty progressive and not very expensive for what it is. I would love to see that measured.
Merging make some nice-looking equipment, but apparently their marketing spiel left honesty behind.
 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
What's good for one is good for the other.

But I think there are arguments on what is good in the first place, as per my queries in opening post.

Rob Watts believes in linear phase ultra fast roll-off:

1589211943793.png


And as per my opening post, Claude sounds like he prefers linear phase slow roll-off (but I'm happy to be corrected).

And he proceeds to talk about time domain performance compared with fast roll-off.

So based on this then both FR and time domain response will differ in approaches.
 

tmtomh

Major Contributor
Forum Donor
Joined
Aug 14, 2018
Messages
2,636
Likes
7,491
I believe the issue is that the time-domain degradation aka ringing only happens near or at the Nyquist limit of the sample rate. So the phase impact - and the alleged "smearing" or otherwise audible degradation - happens at frequencies we can't hear.

Conversely. slow-rolloff filters that reduce this ringing, do so at the cost of producing aliasing, and if the filter rolloff is too slow, then aliasing can be created/allowed at high frequencies that are in the audible range. While you'd have to have a really slow, awful filter to produce sufficient aliasing at low enough frequencies where anyone over the age of about 20 could ever hear the effect directly, there's also the issue of intermodulation distortion created by the interaction of those aliased frequencies with both ultrasonic frequencies and other audible-range frequencies in the source.

I am not nearly as knowledgeable as @mansr and others here, so this could be an oversimplified or incomplete explanation.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Regarding the time domain versus frequency domain evaluation of performance, let me show an example.

First, take a look at this plot showing one period of a sine wave. Can you tell just by looking that it has over 2% harmonic distortion?
1589213383901.png


Now look at the spectrum instead. The third harmonic is impossible to miss.
1589213480652.png


And that is why we like to look at things in the frequency domain.
 
Top Bottom