Audirvana 3.1

[copy_of_a]

13 minutes ago, Musicophile said:

3.1 works flawlessly here so far.

same here.

Recent versions all have been running absolutley trouble free on my 2 systems.

Also AU Plugins load and run as supposed to.

 

[copy_of_a]

@damien78

 

since we have a new thread I'll copy my question from the former thread ... because I really would appreciate a reply.

 

Thanks!

 

_______

 

@damien78  and/or  @mansr
… or anyone who knows fur sure…

 

A short question about the internal processing chain when upsampling to DSD and using AU Plugins.

 

According to the A+ manual when upsampling to a higher PCM rate the processing chain is:
1.) file loading / decoding
2.) sample rate conversion
3.) AU Plugins (in top to down order)
(4.) loading processed PCM into RAM for playback)

For forced upsampling to DSD the processing chain must be different (since there are no AU Plugins that work at DSD sample rates starting at 2.8MHz).

 

Am I correct assuming that AU plugins are applied at the source sample rate when we upsample to DSD?
So for DSD is the processing chain looking like this? ->
1.) file loading / decoding
2.) AU Plugins (in top to down order)
3.) sample rate conversion (iZotope) & DSD modulation (SoX)

Many thanks in advance!

 

[copy_of_a]

2 hours ago, damien78 said:

That's the right sequences of operation in both cases

thanks a lot Damien!!

[copy_of_a]

1 hour ago, jos said:

I think Damien should remove this option from the preferences menu.

pretty bad suggestion if you think about users with older OS X versions (Mavericks, El Capitan...) or those who do replace the audio kext file.

 

If at all the setting maybe could be hidden in some kind of „expert settings“ section in the prefs to make life easier for new users...

[copy_of_a]

1 hour ago, damien78 said:

This is exactly what is done in 3.1.5: Direct Mode option is hidden when it is checked to be not available.

So users on Sierra or HighSierra who have patched the kext still see the option.

great solution!!

didn't recognize this change since I am on El Capitan...

[copy_of_a]

30 minutes ago, trybeingarun said:

Is there a way to get rid of sibilance. I am fine if resolution goes down a bit. I tried changing integer mode, it din't help much.

 

The following is my A+ config

 

Integer mode: 1

Exclusive access mode: true

Direct Mode: true

OS: Sierra 10.12.6

izotope params: 150/130k/1.02/200/0.86

Upsampling: Max sample rate upsampling

 

MC config: (haven't touched too many things)

Sample rate: upsampled to max

Exclusive access mode: true

Integer mode: true

 

Try this:

 

Integer Mode: 2

 

change in izotope parameters:

steepness: 31

cutoff:  0.96 (or 0.97)

Anti Aliasing: 100

 

 

[copy_of_a]

3 hours ago, trybeingarun said:

Thanks, will try. Currently going through your post explaining how the settings work  .

if you want to try my settings I've further adjusted them a little bit ...:

Steepness: 24
Filter: 512,000
Cutoff: 0.97
Anti-Aliasing: 100
Pre-Ringing: 0.68

... and -3db gain reduction to avoid intersample peaks (well, almost all... some few even peak above +3db)

 

[copy_of_a]

2 hours ago, trybeingarun said:

By -3db gain reduction you mean setting "Max. volume level" to -3dB, right? When I do that audirvana's volume goes down by 18 dB and I have to crank up my amp volume way too high (which is not a problem). Still, I din't imagine dBs reduced in "Max. volume level" will affect the total volume 6 times over!

yes, I was referring to the max. volume level (in software only mode, though) but I wasn‘t aware that the messy setting is still not fixed. Since I use an EQ in the Audio Units section I reduce gain there.

Don‘t worry... most DACs have no issues with intersample peaks. 

[copy_of_a]

On 14.10.2017 at 10:07 AM, trybeingarun said:

I've been running this settings since you quoted and it's pretty good :-) . Thanks! Now i'll hope the max volume bug is fixed, as sometimes my volume goes up by 6 dB which is really annoying.

you can use AU Plugins to reduce gain.

There are many free plugins, here are just 3:

 

Simple Gain Tool (for volume control only certainly the easiest to use):

https://www.bluecataudio.com/Products/Product_GainSuite/

Download  -> "Mac AU" (installer contains 3 plugins: "stereo", "dual", "mono". Actually you'd only need to install the "stereo" version)

 

EQ: http://www.voxengo.com/product/marvelgeq/

 

EQ: http://www.tokyodawn.net/tdr-vos-slickeq/

 

[copy_of_a]

6 hours ago, Musicophile said:

Some changes coming to A+, see the attached article (in French):

 

https://www.on-mag.fr/index.php/topaudio/actualites-news/16887-audirvana-plus-l-application-hifi-audiophile-de-reference-pour-mac-enfin-disponible-sur-pc

 

Changes to the UI, and a Windows version.

the redesigned interface looks really nice!

 

[copy_of_a]

1 hour ago, jos said:

I don’t like the dark color. It seems to be modern, but I really don’t like it!

 

I couldn't care less whether it's bright or dark.

I don't use my music computer for color critical work...

 

The design is clean and the interface easy to handle.

Settings (top right) as well as wired and wireless connections (bottom right) are accessible right from the main screen. So no need to click trough the prefs tabs ...

Current playlist as a drop down menu (Qobuz Player style)...

 

Looks pretty good!

[copy_of_a]

37 minutes ago, Charente said:

'Current playlist as a drop down......'

 

I hadn't noticed that ... that will be a boon, and means I don't lose my place in the main browse window. Good one !

we are familar with those icons from other softwares... so I think they should work as expected.

But, well, it's just a screenshot ... so obviously I also could be wrong...

 

[copy_of_a]

17 hours ago, RunHomeSlow said:

I have low static tics when a song finish or CD finishes, and often when starting a CD the first song at 5 seconds

(saw that often the 5 seconds mark), have a 'vinyl pop' kinda of, for no reason, just the first song of the CD...

All that while upsampling PCM to DSD128 on High Sierra and A+ 3.1.7. Nothing, if played with no upsampling.

Same here. Upsampling to DSD often cuts the begining of the 1st track. And I also have those pops after 5 seconds in conjunction with AU Plugins. No such pops with AU Plugins disabled, though.

No issues at all when upsampling to higher rate PCM (with or without AU Plugins).

 

Workaround: create a file with 6 seconds of silence.

Make a playlist of this single silence track.

Drag the album you want to listen to to the silence playlist.

Start playing the playlist with the first track (so with the 6 seconds silence track).

 

@damien78 and @mansr really have to fix this ...

 

 

 

[copy_of_a]

1 minute ago, mansr said:

Not my fault. Your CPU power is probably marginal, especially with the combined load of DSD and AU plugins. Try messing with the buffer sizes. Sometimes that helps.

Nope. Short or long buffers, different SRC settings... it's always the same.

My CPU has enough reserves ... while processing the cores run at 60% or so.

[copy_of_a]

35 minutes ago, mansr said:

Are you sure one core isn't fully loaded, creating a bottleneck?

yes, I am sure.

Besides... it's really only an issue with the very first track.

So it seems there's something not in "sync" when playback starts ("sync" is probably an inaccurate term...)

 

[copy_of_a]

7 hours ago, Nico said:

Preferences 3.1.8:

Why I have to check Direct Mode in order to have Integer Mode options 1 and 2? Is Integer Mode not independent from Direct Mode?

 

I think the naming of the modes is a bit confusing.

 

Without „Direct Mode“ enabled A+ utilizes Mac OSX‘s native Integer Mode (as is). So, yes, Integer Mode is independent from „Direct Mode“.

 

With „Direct Mode“ checked A+ utilizes its own Audio Driver. In this case (only in this case) you can chose between two different modes (1 or 2).

 

Maybe it‘s more clear if you think about the modes as if they were called „Direct Integer Mode 1“ and „Direct Integer Mode 2“... 

 

hope that helps...

 

[copy_of_a]

SoX implementation seems to work just fine now with version 3.2.3.

 

However, the filter quality (with regard to the suppression of mirrored aliasing above the nyquist limit when upsampling) is really weak.

Below screenshots of Fabfilter Pro Q2 (passed through) in AU plugin realtime mode while playing 44.1kHz music upsampeld to 176.4kHz showing the frequencies above 22.050Hz. With my iZozotpe settings mirrored aliasing stays below -110db while SoX shows mirrored aliasing up to -45db (in this particular piece of music I was playing to check SoX).

SoX settings match my iZotope setting with regard to nyquist cutoff and pre-ringing. Steepness in my iZotope settings is moderate (28) while SoX produces more ringing (as it uses a higher steepness filter ... which is why it actually should suppress images better than iZotope with my settings). So either the SoX filter is really weak or something isn't working as supposed to...

 

SoX:
Bandwidth: 95%
Pre-Ringing: 68%

Anti Aliasing: 100
(steepness = fixed)

 

iZotope:
Nyquist: 0.95
Pre-Ringing: 0.68

Anti Aliasing: 100
(Steepness: 28)

 

[copy_of_a]

3 hours ago, damien78 said:

I've just released 3.2.4 with fixes for crashes some of you encountered upsampling with SoX.

Thanks again for the reports that have been very helpful to reproduce, and fix the issue.

Thanks Damien!!

 

I'm still wondering whether or not the implementation of SoX is really correct...

 

Attached 2 screenshots comparing the frequencies of SoX and iZotope when upsampling 44.1kHz to 176.4kHz.

The screenshots show the interface of an EQ plugin in realtime mode while playing music.

SoX shows extremely weak filter quality producing a remarkable amount of mirrored aliasing (~45db).

I can't imagine this is how the SoX resampler is designed originally...?

 

 

SoX settings:

Bandwidth: 95% Nyquist
Anti-Aliasing: 100
Pre-Ringing: 68%

 

 

iZotope Settings:
Steepness: 28
Cutoff: 0.95 Nyquist
Anti-Aliasing: 100
Pre-Ringing: 0.68

 

 

[copy_of_a]

On 12/31/2018 at 9:17 AM, buonassi said:

the way an impulse looks doesn't mean it will provide the best sound quality.  And that there is more to great sound than just the 'pre ringing'. 

I agree! The „extreme“ setting in conjunction with FFProQ2 was aimed at offline upsampling!
As posted somewhere I’ve discarded the quoted workflow and settings myself for a long time.
I’ve also eventually found the strong filtering through FFProQ2 made the sound somewhat dull at the top end and it lost spatiality. iZotopeSRC alone is better suited… and sounds more open. So today I would not consider the respective settings really useful.
However, I still think the 22 (steepness) - 0.95 (cutoff) - 100 (AA) setting - either in minimum or linear phase mode - is pretty good if you prefer (relatively) low ringing. With regard to the transition it’s remotely similar to „poly sinc short“ in HQPlayer but somewhat steeper (and theoretically somewhat more prone to produce aliasing/imaging). With regard to the transition HQPs poly sinc short translated to iZotope would be something like steepness: 16 - cutoff 0.93 - AA: 100.

[copy_of_a]

10 hours ago, Jud said:

I'm thinking steeper = less aliasing/imaging, more ringing.

Basically, yes, that’s correct, of course.

However, it also depends on the cutoff shift. If you set an extremely steep filter but leave the cutoff shift at „1“ (=100%) you may produce aliasing/imaging as the filter turns around the nyquist frequency (see the 4th screenshot below with steepness: 150, cutoff: 1). The cutoff shift seems to be the most overlooked setting here on this forum in the iZotope and SRC threads and this is why it took me a while to understand how it works (although it’s actually pretty self-explaining). Generally speaking I’d say a cutoff shift greater than 0.97 is always prone to produce aliasing/imaging … unless you go really, really steep. It’s for a reason that iZotopes „No Aliasing“ presets set the cutoff to 0.96 (in conjunction with appropriate steepness).

 

Regarding my latest post above I’ve confused my settings. Steepness: 22, cutoff: 0.95 and steepness: 16, cutoff: 0.93 hit the nyquist frequency at the same level (only 1db difference)… so regarding aliasing/imaging they are almost the same. For reference some screenshots below showing the transitions of these 2 settings as well as the filter slope of HQPs poly sinc short.

 

 

10 hours ago, Uni said:

In short, you recommend iZotope setting as steepness: 16 - cutoff 0.93 - AA: 100?  I use "power of 2 oversampling only", which goes to 384k for my DAC.  There's no need to set any EQ cutoff at 18k/19.5k.  Did I get you right for all these?

In HQPlayer I like the poly sinc short setting a lot (which - with regard to the roll off - is similar to iZotope steepness: 16, cutoff: 0.93). In Audirvana with iZotope I actually favour my steepness: 22, cutoff: 0.95 setting.

You can chose the one you like better - and, yes, for both the settings there’s no need to further highcut through an EQ.

Basically you can also use the settings for any rates… but personally I always favour „power of 2 oversampling“ in my setup.

 

 

 

[copy_of_a]

12 hours ago, buonassi said:

my current settings are 16, 1.0 cutoff, 0.6 phase.  I don't care about super HF aliasing.  More concerned about preserving phase up to 16K.  I'm pretty sure these settings don't cause phase shift until the top-most octave.  That's based on my interpretation of graphs over at http://src.infinitewave.ca, but I could be wrong. 

attached the phase of your 16 / 1.0 / 0.6 setting displayed in Room EQ Wizard (orange curve).

You may find 22 / 1.0 / 0.72 (green curve) even better with regard to your goal to preserve linear phase up to 16k (and the somewhat higher amount of ringing at steepness 22 compared to steepness 16 is certainly inaudible).

Then again, if you are happy with your settings, why bother?

 

[copy_of_a]

2 hours ago, macuniverse said:

Can you provide some info/link on how we can do this using REW please? Thanks

I load a 44.1kHz Dirac Impulse into iZotope RX (4), apply the respective SRC settings, upsample to 176.4kHz (without "Anti Aliasing", though, since this is a feature of Audirvana and is not present in the original software), export to 32bit float wav. Then load the processed file into REW.

So... it requires a software that provides iZotope SRC and export function in the first place...