Offline Upsampling

copy_of_a · February 18, 2015

Very strange - have noticed no problems on my mid-2009 MacBook Pro (2.53GHz Core 2 Duo, 8GB RAM).

Hi Jud.

I also did not encounter problems when batch converting files through AuI and, say, running Mail and Firefox.

But when batch converting many files with AuI (especially FLAC files that first need to be decoded) and at the same time running another processor intensive task (for instance batch converting photo RAW-files with a processor utilizing multi cores or batch converting another set of audio-files through Wavelab - reading and writing from/to different drives) you will run into problems. Never had the same issues when running, say, Wavelab, Capture One (photo processor) and Avid Media Composer (Video Editing Software) simultaneously processing large (!) amounts of large (!) files ... and they all utilize all cores and play very well with "Grand Central" (Mac's Multicore Doubler/Handler).

So on behalf of AuI maybe not really an issue for daily use but clearly an issue for professional use and therefore certainly a flawed design.

craighartley · February 18, 2015

Dear Copy_of_a,

New version with True Gapless Mode ready. Wait news tomorrow

This mode, in contrast usual separate file conversion, allow fully avoid ringing artefacts between songs.

Parallel processing (multiCPU) don't used for this mode due need serial processing "file-by-file".

Best regards,

Yuri

Yuri,

I'm a bit confused reading this thread. In which situation (which type of conversion) has AulConverter been inserting silences? I've just converted a large batch of classical albums which should be gapless, some pcm to dsf, some dsd 64 to 128, some dff to dsf, and some ISO to dsf. Which do I need to convert again with the new version?

Craig

copy_of_a · February 18, 2015

I've just converted a large batch of classical albums which should be gapless (...) Which do I need to convert again with the new version?

all of them

craighartley · February 18, 2015

all of them

Aaaarghhhh!

copy_of_a · February 18, 2015

Aaaarghhhh!

yep!

same here...

and re-processing on a single core will take ages (AuI is already pretty slow when using all cores).

hmh ...

Jud · February 18, 2015

Understand now c of a, thanks.

craighartley · February 18, 2015

yep!
same here...

and re-processing on a single core will take ages (AuI is already pretty slow when using all cores).

hmh ...

What's worse is that I've just redone many of them due to improved filtering of current version over previous....

rodrigaj · February 19, 2015

There is no way I am going to redo the D128 again on a single core. It simply takes too much time.

Perhaps, I am not sensitive enough, but the other night I listened to a redbook recording of Berlioz's Damnation of Faust after conversion to D128 and could discern no problems.

Jud · February 19, 2015

Yuri develops his program at a rapid pace, so I think a strategy of wholesale processing of large parts of your collection is bound to end in tears.

An alternative you may want to consider is to convert what you plan to listen to, and perhaps get rid of it later so it doesn't take up space. In other words think of the originals as permanent but conversions as temporary.

copy_of_a · February 19, 2015

Understand now c of a, thanks.

you're welcome! I should have been more detailed in my first post about this issue ... sorry!

Yuri, quite honestly I do like your SRC algorithm (though AuI's key feature IMO is its minimum phase filter the linear phase filter is where it shines when upsampling music that doesn't require "punch"/"directness" ... IMHO). Converting 44.1kHz redbook to DSD128 also sounds really good ... but still a bit denser than HQ Player upsamling 44.1 redbook to DSD128 on the fly (on my 2014 MacbookPro i5, 8GB RAM, slimmed down Mavericks OS X for music playback purposes only).

When you add the shortcomings of the software design, the self backed dithering that is essentially a 16bit-Dither even when you output 24bit files (noisefloor at around -110db ... but should max out at -140db with 24bit) - and while it is a stereo-dither it still somehow seems to "monofy" stereo-music when applying cross-fades when editing audio files -, when you also add the gapless issue and the fact that there is no option to tune any setting (beside the dither level and the output gain), when you also add the fact that it utilizes ffmpeg for input conversions (freeware) ... do you really think it's worth the steep price you are asking?

I am really happy to support innovative projects and would certainly donate quite some money for your software, because I like the general idea. But, honestly, IMHO in its current state it should be donation-ware ... or maybe something like 35-50$ since right now it's not up to the task and - while doing some really good things - it is actually Beta-status ... IMHO.

Sorry for talking so bluntly!

I don't feel like beeing ripped off - not at all!! But using the software I discover one issue after the other ... and I think at some point you should also refer to and respect some of the well establised standards (for instance regarding dither) in your software design rather than fiddling around with completely proprietary algorithms.

I keep an eye on your software, of course! But for the time beeing I am a bit tired of Beta-testing ...

audiventory · February 19, 2015

Am I missing something? I thought you could always limit the number of processors (i.e., 1, 2, 3 or 4)

Hi Augustine,

1. Default: AuI can use up to all available cores (1 core per 1 converting file).

2. Manually: You can restrict maximal number of used cores.

3. True Gapless Mode convert file by file, i.e. 1 file simultaneously. Before converting next file need get information about end of previous file.

[ATTACH=CONFIG]17009[/ATTACH][ATTACH=CONFIG]17009[/ATTACH]

Hi Copy_of_a,

I saw you message very later (night) and answer shortly due wanted write details on fresh mind at tommorow

What we see in the picture as gaps (fade in and out) is work of smooth DC functionality.

It allow avoiding click/artefacts at begin and end. When source file contains DC possible clicks without fading.

In new version Smooth DC can turned ON/OFF via switch in Settings window.

The switch available in intermediate online beta PROduce-RD several days and allow get traditional non-accurate gapless mode conversion (turn OFF position).

New True Gapless Mode (AuI trueG) accurate convert all files from "Input files list" as single audio stream, without transient processes.

It allow avoid ringing artefacts and difference between first samples levels of source and converted file.

Now no sense download the beta due more functional version plan coming today.

I prepares now short article about new version. And plan publish it after uploading new versions today morning (by Moscow time).

Thank you for questions, friends. I’m glad what I can prepare useful functionality before I'm asked

Best regard,

Yuri

audiventory · February 19, 2015

Yuri,

I'm a bit confused reading this thread. In which situation (which type of conversion) has AulConverter been inserting silences? I've just converted a large batch of classical albums which should be gapless, some pcm to dsf, some dsd 64 to 128, some dff to dsf, and some ISO to dsf. Which do I need to convert again with the new version?

Craig

Hi Craig,

Some file, especially, ISO images, has DC offset. I burned clicks for some cases.

I got many reply what need fix the clicks.

It was released via shortest fade in/out as simplest and fastest way.

Current True Gapless Mode conversion is new feature, hard in realization.

1. These fades in/out (as called gaps) only appear for places with high loudness level on borders of neighboring tracks.

Overwhelming majority of tracks has zero level in begin and end, therefore DC Smothing (fade in/out) don't impact to it.

2. It have length 10 ms and listened for some cases only.

In next versions I plan add right fix DC offset also.

Algorithms imrove with time, of course, therefore better always keep sources.

If you listen gap between neighboring tracks - need reconvert.

If you don't listen (most cases) no need re-converting.

Best regards,

Yuri

audiventory · February 19, 2015

this might be a good idea, too. On my machine (2009 MacPro 8x2.26GHz Nehalem & 24GB RAM) AuI behaves really strange. When batch converting multiple files either AuI crashes or it causes other applications simultaneously running to crash.

AuI utilzes 100% of all cores (which is good!) but it doesn't play well together with other softewares also utilizing mulitcore processing. Normally softwares utilizing heavy mulitcore usage should balance each other (at least that's how my other softwares behave ... they "share" the multicores; they do not block all the cores for their own processing).

I suppose here as crash you mean non-responding user interface?

(crash, as abnormal termination, of one application don’t lead to crash other application)

Any app don’t control utilizing CPU power. It do operation system.

Some app allow choose priority, but how it will work in reality almost unknown - all control operation system.

I added to wish list priority settings.

If need reduce AuI’s CPU consuming need manually select maximal count used cores. It allow free resources for other tasks.

craighartley · February 19, 2015

Yuri develops his program at a rapid pace, so I think a strategy of wholesale processing of large parts of your collection is bound to end in tears.

An alternative you may want to consider is to convert what you plan to listen to, and perhaps get rid of it later so it doesn't take up space. In other words think of the originals as permanent but conversions as temporary.

An important part of the point for me is to provide versions to add to my library with consistent and adequate metadata of albums I have bought in dff and ISO. I wouldn't want to delete them after all the effort to convert and tag them.

While I realise that future improvements in Yuri's program may encourage me to redo them at some point and batch copy the metadata from my previous versions using Yate, the provision of true gapless is a basic necessity for including them in my library, or playing them at all, in the first place. So for the moment I am condemned to this Vale of Tears, or in other words, I don't see an alternative to investigating how long it will take to redo them with Yuri's new version.

craighartley · February 19, 2015

Hi Craig,

Some file, especially, ISO images, has DC offset. I burned clicks for some cases.

I got many reply what need fix the clicks.

It was released via shortest fade in/out as simplest and fastest way.

Current True Gapless Mode conversion is new feature, hard in realization.

1. These fades in/out (as called gaps) only appear for places with high loudness level on borders of neighboring tracks.

Overwhelming majority of tracks has zero level in begin and end, therefore DC Smothing (fade in/out) don't impact to it.

2. It have length 10 ms and listened for some cases only.

In next versions I plan add right fix DC offset also.

Algorithms imrove with time, of course, therefore better always keep sources.

If you listen gap between neighboring tracks - need reconvert.

If you don't listen (most cases) no need re-converting.

Best regards,

Yuri

Thanks Yuri,

If I understand correctly, if I do want to reconvert a large batch of files it would be better for me to hold off until you release a version that fixes the DC offset?

Do you have an approximate timeline for that?

Best

Craig

audiventory · February 19, 2015

Hi Jud.

I also did not encounter problems when batch converting files through AuI and, say, running Mail and Firefox.

But when batch converting many files with AuI (especially FLAC files that first need to be decoded) and at the same time running another processor intensive task (for instance batch converting photo RAW-files with a processor utilizing multi cores or batch converting another set of audio-files through Wavelab - reading and writing from/to different drives) you will run into problems. Never had the same issues when running, say, Wavelab, Capture One (photo processor) and Avid Media Composer (Video Editing Software) simultaneously processing large (!) amounts of large (!) files ... and they all utilize all cores and play very well with "Grand Central" (Mac's Multicore Doubler/Handler).

So on behalf of AuI maybe not really an issue for daily use but clearly an issue for professional use and therefore certainly a flawed design.

Me seems, here need check what number of CPU cores use each of these application. If used not all cores by each application - general CPU loading will less then each app use all cores. In this case enough resources for all simultaneously runned applications.

Here possibly try set different priority for all tasks:

1. Set priority in OS's task manager,

2. Set priotrity in software settings (in AuI planned for future versions).

3. In settings AuI choose less cores than all.

audiventory · February 19, 2015

Thanks Yuri,

If I understand correctly, if I do want to reconvert a large batch of files it would be better for me to hold off until you release a version that fixes the DC offset?

Do you have an approximate timeline for that?

Best

Craig

Dear Craig,

Last months AuI's sound algorithms was improved many times. And, very possibly, will further improvements.

I think, don't need re-convert library for each changing.

For estimate time of releasing DC elimination me need learn more code of the software. I think release it nearest versions.

Performing this task (group processing output files) allow me create base for other same tasks (album loudness normalizing, as example).

Best regard,

Yuri

audiventory · February 19, 2015

An important part of the point for me is to provide versions to add to my library with consistent and adequate metadata of albums I have bought in dff and ISO. I wouldn't want to delete them after all the effort to convert and tag them.

While I realise that future improvements in Yuri's program may encourage me to redo them at some point and batch copy the metadata from my previous versions using Yate, the provision of true gapless is a basic necessity for including them in my library, or playing them at all, in the first place. So for the moment I am condemned to this Vale of Tears, or in other words, I don't see an alternative to investigating how long it will take to redo them with Yuri's new version.

Yes, in case metadata it's real trouble.

Same to it, if you buy new DAC with more high sample rate ability.

I will think how provide copying metadata, I added it to wish list.

Now I moves AuI to music manager role. Thus this task (including management of originals) must be consigered as music manager function.

Jud · February 19, 2015

An important part of the point for me is to provide versions to add to my library with consistent and adequate metadata of albums I have bought in dff and ISO. I wouldn't want to delete them after all the effort to convert and tag them.

While I realise that future improvements in Yuri's program may encourage me to redo them at some point and batch copy the metadata from my previous versions using Yate, the provision of true gapless is a basic necessity for including them in my library, or playing them at all, in the first place. So for the moment I am condemned to this Vale of Tears, or in other words, I don't see an alternative to investigating how long it will take to redo them with Yuri's new version.

Right, think I'm understanding better now (seems leaping to conclusions is something I'm doing fairly often in this thread).

So what you're primarily doing, at least at this point, is converting dff and ISO to dsf with proper metadata?

If that's the case, then yep, once there are no more clicks left when converting, it seems to me that the dff to dsf and ISO to dsf conversions would be lossless (as long as you don't change the sample rate), and you could treat the resulting dsf files as originals.

audiventory · February 19, 2015

Yuri develops his program at a rapid pace, so I think a strategy of wholesale processing of large parts of your collection is bound to end in tears.

An alternative you may want to consider is to convert what you plan to listen to, and perhaps get rid of it later so it doesn't take up space. In other words think of the originals as permanent but conversions as temporary.

Hi Jud,

I suppose right releasing of music manager conception allowed perform any reconverting tasks in background mode with auto keeping of matadata.

Best regards,

Yuri

craighartley · February 19, 2015

[quote name=Jud;399861

So what you're primarily doing' date=' at least at this point, is converting dff and ISO to dsf with proper metadata?

If that's the case, then yep, once there are no more clicks left when converting, it seems to me that the dff to dsf and ISO to dsf conversions would be lossless (as long as you don't change the sample rate), and you could treat the resulting dsf files as originals.[/quote]

Yes that was my thinking. I'll also be experimenting with pcm to dsf conversion to see if there are improvements that I can hear, but I see that that is more dependent on developments with filtering and on which DAC I have at the time.

There is another consideration, which is cost in terms of purchases and time. For instance I can buy Pentatone dsd recordings as ISO files from the Pentatone site cheaper than I can buy them as dsf files from the primemusic site. That's fine if I have a reliable and one-off way of turning ISO into dsf with edited metadata. But if the process is too much of a hassle I'd rather pay the extra for the dsf files in the first place.

craighartley · February 19, 2015

Right, think I'm understanding better now (seems leaping to conclusions is something I'm doing fairly often in this thread).

So what you're primarily doing, at least at this point, is converting dff and ISO to dsf with proper metadata?

If that's the case, then yep, once there are no more clicks left when converting, it seems to me that the dff to dsf and ISO to dsf conversions would be lossless (as long as you don't change the sample rate), and you could treat the resulting dsf files as originals.

[quote name=Jud;399861

So what you're primarily doing' date=' at least at this point, is converting dff and ISO to dsf with proper metadata?

If that's the case, then yep, once there are no more clicks left when converting, it seems to me that the dff to dsf and ISO to dsf conversions would be lossless (as long as you don't change the sample rate), and you could treat the resulting dsf files as originals.[/quote]

SORRY, posted again with quote intact:

Yes that was my thinking. I'll also be experimenting with pcm to dsf conversion to see if there are improvements that I can hear, but I see that that is more dependent on developments with filtering and on which DAC I have at the time.

There is another consideration, which is cost in terms of purchases and time. For instance I can buy Pentatone dsd recordings as ISO files from the Pentatone site cheaper than I can buy them as dsf files from the primemusic site. That's fine if I have a reliable and one-off way of turning ISO into dsf with edited metadata. But if the process is too much of a hassle I'd rather pay the extra for the dsf files in the first place.

audiventory · February 19, 2015

SORRY, posted again with quote intact:

Yes that was my thinking. I'll also be experimenting with pcm to dsf conversion to see if there are improvements that I can hear, but I see that that is more dependent on developments with filtering and on which DAC I have at the time.

There is another consideration, which is cost in terms of purchases and time. For instance I can buy Pentatone dsd recordings as ISO files from the Pentatone site cheaper than I can buy them as dsf files from the primemusic site. That's fine if I have a reliable and one-off way of turning ISO into dsf with edited metadata. But if the process is too much of a hassle I'd rather pay the extra for the dsf files in the first place.

Craig,

Usually, when converted ISO to DSF, metadata from ISO copied to output DSF.

As I understand, converted DSF (from Pentatone's ISO) don't contains metadata, don't it?

craighartley · February 19, 2015

Yes. But their metadata is inconsistent and does not usually match the consistent way that I have edited the metadata in the rest of my library.

There is so much inconsistency between download vendors in what metadata they provide and whether it is correct, that it is rare that I don't have to correct it.

craighartley · February 19, 2015

Craig,

Usually, when converted ISO to DSF, metadata from ISO copied to output DSF.

As I understand, converted DSF (from Pentatone's ISO) don't contains metadata, don't it?

I mean Yes, the Pentatone ISO files do contain metadata that is carried through to the dsf conversions, but their metadata is inconsistent and does not usually match the consistent way that I have edited the metadata in the rest of my library.

There is so much inconsistency between download vendors in what metadata they provide and whether it is correct, that it is rare that I don't have to correct it.

Offline Upsampling

Recommended Posts

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Create an account or sign in to comment

Create an account

Sign in