New Filter request(s)

[charliemb]

I found two good examples of where newAI filter using 0 (min phase) and TPDF as its variables sounds much better than the new Custom.  I post these two titles so that you may try them for yourself if you happen to have this music in your collection.

The first is Traffic Far from Home.  This is in the class of popular or rock music.  Here, using Custom, Steve Windwood's vocals sound as if only midrange is present.  Assuming your system might reproduce what I'm hearing, it sounds almost as if he is singing through one of those hand-held "megaphones" or institutional outdoor speakers used for public announcements.  For sure the feeling that something is missing.  Now switch to AI.  With AI it sounds like a professional recording, proper, with an entire range of high frequencies blended in with the midrange.  It sounds like a real person singing.

Next is Antal Dorati and the London Symphony Orchestra, as released by Mercury Living Presence Léo Delibes Sylvia & Cappélia, Disc 2.  This is classical music and there are all sorts of different stringed instruments.  The focus here is on the delicate sound of the strings.  With Custom, it sounds good;  but a little bit like a CD, which it is, and so has a little bit of that digital hardness.   Now switch to AI.  With AI it sounds wonderful, like real strings.  Forget CD sound, it sounds like HD material or SACD.  There's just a beauty to the strings that is not there with Custom.  Folks,  it sounds like a live concert, very smooth, very analog, and with a very nicely balanced high end.  Further,  with this recording, there's no buzzing / zooming or distortion as I'd normally expect with the extra ringing.   

Both were played at volume levels that sounded best for each:  -3 db for Custom,  and 0db for AI.  (And for Traffic I think it was -4.5db for Custom)

Because it takes so long to load AI at 8x, I tend to use Custom.  So making the above realizations can take some time.

[PeterSt]

Hi Charlie,

With these two clear examples I am surely going to try.
Next up would be to incorporate "AI" in the Custom Filter (settings), although I can't tell at this moment how difficult that would be (read : when ever I'll have the probably long time needed for that).

Thanks once again ...
Peter

[charliemb]

In contrast, this album The Very Best of the Manhattan Transfer plays by far the best with Arc Prediction upsampling at the highest sampling rate I have available, 352.8.
It plays hyper cleanly with arc prediction, very analog, and the vocals sound lush without the help of any added ringing/reconstruction. 

With Custom (176 @ 384) it sounds very good, but if you focus on the vocals there is clearly some buzzing going on from unneeded reconstruction.  So this recording has plenty of post processing on the vocals to get a certain sound.

With the newAI (set to 0, TPDF) this album plays with clearly audible distortion in the vocals (only).  There is too much ringing / reconstruction going on.

----------------->  Not tied specifically to this album:

I've noticed with these three upsampling methods that the vocals climb in the mix with added ringing (more added harmonics) and fall in the mix with less.  In one extreme we have newAI with the most ringing and where I perceive the vocals as louder.  In the other extreme we have arcPrediction with no ringing where I perceive the vocals a little lower in the mix.  In the middle we have Custom.

What's interesting to observe between these methods is that, because the ear is most sensitive in the vocal range, I tend to adjust the volume based on the vocal's loudness or perceived loudness.  The minimum step on my dac for volume is 1db.  So in practice, after I adjust the volume to compensate for the vocals, it is the band / instruments that vary in the mix.   The end result is that with Arc Prediction I get the most "band" in the mix, with newAI I get the least amount of "band," and with Custom I get somewhere in between.

This is all very subtle.   Even though subtle, it can make a big difference in perception and overall experience.  With more band from arc prediction can come the  perception that arc prediction brings the best sound stage and imaging (which appears to consistently be true).  At the other extreme we have the newAI.  Custom comes in very close to Arc Prediction for perceived sound stage, usually.

[PeterSt]

Awesome description ...
!

[charliemb]

Hi Peter

Whenever it comes, I'm happily anticipating a 384KHz Custom Filter.  And even if it is beta, I'll gladly try*any one*.   This is because the 192 is not right at 384, with too much buzzing, and with ArcP being too much on the clean / dry side most of the time.   

Much has happened here with my DAC; I've had to return it due to that same initialization / synch problem, and so the fixed DAC works right  but I've had to break it in a second time and so it has been a moving target.   

I was previously running with the slow roll-off filter and now that has changed as I'm now back to the brickwall on the DAC.   It sounds better right now.  And believe it or not the best sounding filter within xx is back to the new "AI" with minimum phase settings and TPDF dither.  This could be because it is a preprocess or because it is at least tuned to 384 and maybe I just can't hear the buzzing because it is too high in frequency.  Along with this, Miska's asymFIR filter at 384 is the best sounding while xxHighEnd is still the most impactful by far.

But I highly suspect that any custom 384 filter in XXHE will be, both  the best sounding and most impactful. 

Bring it ON!!!!

[PeterSt]

Hey Charlie,

I looks I just finished a very long "project". The filter stuff has always been postponed until the end, which should be about now. So first I wil make a 384 filter, especially for you and because you have been so patient. You will receive it as soon as I have finished it. Only you.

Peter

[charliemb]

Thanks!

[PeterSt]

So ...

I think I just spent 3 days of tuning filters for 768, 384 and 192 (not for 96). All tested with 16/44.1.
And that was even by automation (let the program change the filter parameters automatically and meanwhile watch the analyser).

All with the least of ringing and all without changing the phase anywhere.

And the change in SQ ?
hahaha ho ho ho
Not a downside that I can hear (all is as accurate) - only upsides.

Peter

[Robert]

I guess we can look forward to a new upgraded Xxhighend soon!!!

How will you find time to make the Clairixa's?

Robert

[charliemb]

Awesome.  Looking forward to hearing your filters.

[charliemb]

Awesome.  Looking forward to hearing your filters.

To anyone subscribed to this topic, there is a new 2.01 version that now supports these filters.

In one word, they are UNBELIEVEABLE!!!!!   

[Edit: Only as of May 31, 2015 are my settings accurate in my signature lines below,  in case anyone is interested in my settings.  They are important.]

[christoffe]

Awesome.  Looking forward to hearing your filters.

To anyone subscribed to this topic, there is a new 2.01 version that now supports these filters.

In one word, they are UNBELIEVEABLE!!!!!   

 

Hi Peter,

Is it possible to know your settings within 2.01. 

Joachim

[PeterSt]

Joachim, you are right - I must update my sig.
I will do that soon.

Regards,
Peter