I know, I know...Apodizing Filter

[John Phillips]

I prefer intermediate phase filter (SoX) over linear phase filter for long term listening. Sometimes with linear phase I get a sense that the sound is cold. Not all the time but now and then.

This has never, ever happened to me with intermediate phase.

On a quick A/B I hear no difference, so if this is a real thing it's subtle.

All comments in my own system. And I don't say a minimum-ish phase filter alone is sufficient for good sound, other things must be right too.

PS: See https://pinkfishmedia.net/forum/threads/digital-filtering.221996/ for pictures, minimum vs intermediate phase in SoX.

I agree about listening longer-term to get a good feel for what you like (or not). Whether or not break-in happens in equipment I am sure it does happen in the listener's ear-brain system and my experience is it needs a week or two.

I did run my up-sampler with different filter settings for reasonably long periods (a week or more) including trying out SoX's -p 0 (minimum phase), -p 25 (intermediate phase - your preference I think), -p 45 (Archimago's "goldilocks" setting - probably very similar to the AKM "low dispersion" filter), and -p 50 (linear phase - including some very tight bandwidth settings which probably use much of SoX's default 32,000-tap capability). I didn't go as far as trying the maximum phase setting as @adamdea did for commendable completeness.

Actually the last time I updated the streamer software I didn't enable up-sampling and left the DAC to handle all sample rates with its minimum phase filter, which I find perfectly satisfactory.

I'm sure you are right to suggest that how it works in practice may depend on the complete system. I guess it may also depend on the ears of the beholder as well. Mine are getting rather old. Old enough for me to be certain I don't hear things as I used to.

 

[tuga]

The key point of having a standard 'time symmetric' sinc reconstruction filter is that it outputs the analogue waveform *defined by the recording process* which includes the choice of ADC. Thus it is the 'correct' result. i.e. if the creators of the digital recording wanted to have no pre-ringing they could have produced a recording that provides this with a standard 'sinc' DAC reconstruction filter.

So in reality, if you have a 'slow' reconstruction filter it may time smear a recording that was originally made with an ADC input chain that avoided pre-ringing. So if you think you dislike time 'smear' it may make the results worse in such cases. It isn't a magic bullet.

My personal guess is that when people prefer one reconstruction to another they may well simply be using the choice as an HF 'tone control' to tweak the HF levels they hear. Possibly offsetting a departure elsewhere in the record -> replay chain.

But again, personally, I've never felt the difference were worth bothering about in the systems I've used. YMMV

You've mentioned time-smear, so perhaps not just as a way to roll-off the top but perhaps to smooth out some edginess?

Some types of distortion sound nice to some people, and I guess that they can also be used to compensate or tone-down particular shortcomings of a system.

 

[tuga]

I agree about listening longer-term to get a good feel for what you like (or not). Whether or not break-in happens in equipment I am sure it does happen in the listener's ear-brain system and my experience is it needs a week or two.

I did run my up-sampler with different filter settings for reasonably long periods (a week or more) including trying out SoX's -p 0 (minimum phase), -p 25 (intermediate phase - your preference I think), -p 45 (Archimago's "goldilocks" setting - probably very similar to the AKM "low dispersion" filter), and -p 50 (linear phase - including some very tight bandwidth settings which probably use much of SoX's default 32,000-tap capability). I didn't go as far as trying the maximum phase setting as @adamdea did for commendable completeness.

Actually the last time I updated the streamer software I didn't enable up-sampling and left the DAC to handle all sample rates with its minimum phase filter, which I find perfectly satisfactory.

I'm sure you are right to suggest that how it works in practice may depend on the complete system. I guess it may also depend on the ears of the beholder as well. Mine are getting rather old. Old enough for me to be certain I don't hear things as I used to.

I find this type of comparison an absolute bore but I did spend some time comparing minimum phase and linear phase and my preference for the latter is unequivocal.

The HQPlayer manual states that in the author's opinion minimum phase is "good for pop/rock/electronic music containing strong transients such as drums and percussion and where recording is made in a studio using multi-track equipment" whilst linear phase and asymmetric (shorter pre-echo and longer post-echo) are "good for jazz/blues, and other music containing transients recorded in real world acoustic environment"

HQPlayer offers a large selection of filters.
I don't dislike the Asymmetric FIR filter but in my opinion the sound is not as refined as that of one of HQP's more advanced filters.
I am able to identify a notable difference between the Chord-like million tap poly-sinc-xtr (linear phase polyphase sinc filter with extreme cut-off and attenuation) and the designer's preferred poly-sinc-ext2 (linear phase polyphase sinc filter with sharp cut-off and high stop-band attenuation for extended frequency response while completely cutting off by Nyquist frequency.).

 

[Jim Audiomisc]

You've mentioned time-smear, so perhaps not just as a way to roll-off the top but perhaps to smooth out some edginess?

Some types of distortion sound nice to some people, and I guess that they can also be used to compensate or tone-down particular shortcomings of a system.

The problem is that the frequency and time domain behaviours tend to be linked. Add in the fact that 'lazy' filters tend to generate HF anharmonics, and you end up with all three behaviours arriving as a package deal. Given that, it's almost impossible to know which - if any! - of them is producing any audible effect. But of course some may like the result on some material in some systems whilst others don't.

Just like using conventional tone controls, yer pays yer money and yer takes your choice.

I recall that when a friend of mine (who had more money than me at the time) bought his QUAD 33/303/ESL system the man who delivered it recommended always applying some HF filter cut when playing LPs. Mellow tone, etc...

 

[DimitryZ]

Anyone with signal processing background care to look at the AKM4497 low dispersion filter from the datasheet and compare it to the standard apodizing filter, presumably in MQA?

I find it hard to believe a typical recording engineer will pay attention to the filter used in an ADC, and I dont know if that is or was available. I suspect the default setting was used in most recordings.

We also have competing positions that DAC filter settings are pretty inaudible or that they can be used as "tone controls". In my limited experience, Sabre and AKM have clearly audible filters, while Wolfson chips not so much.

 

[darrenyeats]

Analogue filters also exhibit phase shifts and lead to post-ringing. Put another way, a physical filter e.g. - a passive electronic filter, or consider a tweeter - when receiving an input that causes it to ring, can't start its ringing earlier in time than said input. You can call this smear, but it's the natural order of things.

So either you use a NOS DAC (relying on the natural filtering of amps, tweeters, maybe air, but with the downside of possible distortion products in amps/tweeters leaking down into audible range), or put a filter in the DAC. If the latter, I argue that minimum-ish phase digital filter is more like a physical filter.

The downside is that phase shifts accumulate, so that if a sound engineer has already phase-shifted in the recording and you come along and phase-shift again ... This is part of the justification of MQA, that they know what was done upstream and account for it at playback. Or try to, it's conceivable some multi-track recordings may make it impossible. (That doesn't mean I like everything else about MQA.)

 

[Jim Audiomisc]

Analogue filters also exhibit phase shifts and lead to post-ringing. Put another way, a physical filter e.g. - a passive electronic filter, or consider a tweeter - when receiving an input that causes it to ring, can't start its ringing earlier in time than said input. You can call this smear, but it's the natural order of things.


The downside is that phase shifts accumulate, so that if a sound engineer has already phase-shifted in the recording and you come along and phase-shift again ... This is part of the justification of MQA, that they know what was done upstream and account for it at playback. Or try to, it's conceivable some multi-track recordings may make it impossible. (That doesn't mean I like everything else about MQA.)

Its unavoidably that response can't lead input. But that doesn't mean the response peak has to be immediate. So - depending on the design - any technolgical method can produce what seems like 'pre ringing' because the system had an overall propagation delay. As a general rule, in principle digital and analogue filters are interchangeable - just that in some cases doing so would be insanely difficult!

The principle of correcting for the ADC chain is a nice one in theory. But as per the above, if the material actually went though a variety of process stages that differ from one 'mic/source path' to another, this becomes difficult once you have them combined. Particularly for old recordings where such details are unknown.

And if you're going to do that, on vocal music, one of the biggest 'effects' tends to be by the microphone. Many have 'ring and fall' profiles well below 20kHz. And it varies with mic angle and distance from voice... Seems weird to correct relatively tiny ADC chain effects at higher frequencies yet ignore this.

All part of the fun of trying to nail down the 'orginal sound', eh?...

 

[Julf]

Bruno Putzeys had a pretty good comment:

These folk are trying to have their cake and eat it. Either aliasing doesn't matter because there is no signal in the transition band and then the precise shape of the transition band doesn't matter either (ie the ring tails have no conceivable manifestation) or the absence of ring tails is critical because there is signal in that region and then the aliasing will result in audible components that fly in the face of MQA's transparency claims.

Doesn't that just sound like the arguments DSD folks used to make? The requirement for 100kHz bandwidth was made based on the assumption that content above 20k had an audible impact whereas the supersonic noise was excused on the grounds that it wasn't audible. What gives?

 

[Joe P]

I'm not a technical person, but I'd be careful when fiddling with the apodizing filter settings. The last thing you'd want is for the stupid filter to lop off your foot or something!

Joe

 

[tuga]

Analogue filters also exhibit phase shifts and lead to post-ringing. Put another way, a physical filter e.g. - a passive electronic filter, or consider a tweeter - when receiving an input that causes it to ring, can't start its ringing earlier in time than said input. You can call this smear, but it's the natural order of things.

So either you use a NOS DAC (relying on the natural filtering of amps, tweeters, maybe air, but with the downside of possible distortion products in amps/tweeters leaking down into audible range), or put a filter in the DAC. If the latter, I argue that minimum-ish phase digital filter is more like a physical filter.

The downside is that phase shifts accumulate, so that if a sound engineer has already phase-shifted in the recording and you come along and phase-shift again ... This is part of the justification of MQA, that they know what was done upstream and account for it at playback. Or try to, it's conceivable some multi-track recordings may make it impossible. (That doesn't mean I like everything else about MQA.)

This would make sense with real-stereo recordings but how can they know what was upstream in a multi-track album?
Do studios keep a record of which ADC models and filters where used for each track?

 

[DimitryZ]

This would make sense with real-stereo recordings but how can they know what was upstream in a multi-track album?
Do studios keep a record of which ADC models and filters where used for each track?

My guess is that MQA probably have a few recording profiles developed based on the era, general equipment used at the time and maybe some historical trends. Maybe something like "60s Jazz, Verve" and "70s Rock, UK"

 

[tuga]

My guess is that MQA probably have a few recording profiles developed based on the era, general equipment used at the time and maybe some historical trends. Maybe something like "60s Jazz, Verve" and "70s Rock, UK"

60s jazz and 70s was recorded to tape. A lot of it has been digitialised a few times. What about the music that was recorded digitally? And then mixed in analogue, and then digitialised again? I think it's good in theory but won't work in practice.

I see no need whatsoever for the existence of MQA.

 

[DimitryZ]

60s jazz and 70s was recorded to tape. A lot of it has been digitialised a few times. What about the music that was recorded digitally? And then mixed in analogue, and then digitialised again? I think it's good in theory but won't work in practice.

I see no need whatsoever for the existence of MQA.

So your choice is clearly NOT to listen to it.

 

[tuga]

So your choice is clearly NOT to listen to it.

I clearly see no reason to listen to MQA considering that it's lossy and technically flawed and I don't have the necessary hardware (not sure how much I can "unfold" with HQP) and that I don't stream from a remote library.

 

[DimitryZ]

So as long as your statement above really reads

"I see no need whatsoever for the existence of MQA for myself" ----

We are all good.

 

[Shadders]

I have an MQA DAC - M2TECh Young III, and an non MQA modern DAC - Topping D70. On the same material available on Tidal as MQA and on Qobuz as hires, these both sound VERY similar if the low dispersion filter is enabled on rhe D70.

Hi,
If the two tracks sound "VERY" similar, then obviously MQA processing is near non-existent, and it is the filter that causes the sound to be "VERY" similar. The two filters will have slightly different coefficients, if you have matched them appropriately.

As you have referred to in later posts, the AKM DAC has different filters, as do other DAC manufacturers. If two DAC's subjective sound is VERY similar due to the filters used - then you have your answer ?

Regards,
Shadders.

 

[TheDecameron]

Are you all living in another dimension or something.

I’m looking for a filter that will push all financial ringing to the right in the time domain, ie after I’ve shuffled off the time domain.

 

[darrenyeats]

But that doesn't mean the response peak has to be immediate.

I accept the technical correction!

I think it remains fair comment that it's natural for systems to post-ring. Tweeters do for example. I'll have to take your word for it that "many" microphones don't ... but I think usually they post-ring.

 

[DimitryZ]

Hi,
If the two tracks sound "VERY" similar, then obviously MQA processing is near non-existent, and it is the filter that causes the sound to be "VERY" similar. The two filters will have slightly different coefficients, if you have matched them appropriately.

As you have referred to in later posts, the AKM DAC has different filters, as do other DAC manufacturers. If two DAC's subjective sound is VERY similar due to the filters used - then you have your answer ?

Regards,
Shadders.

Perhaps. There has long been speculation that the basis for MQA "sound" is the apodizing filter.

 

[Jim Audiomisc]

60s jazz and 70s was recorded to tape. A lot of it has been digitialised a few times. What about the music that was recorded digitally? And then mixed in analogue, and then digitialised again? I think it's good in theory but won't work in practice.

Add in "Dolby A" on old tapes, and consider what *that* does to the signal timing. 8-]