Audirvana Studio

DACs and Digital Software
21

More disturbingly, despite claims to the contrary, it appears that Audirvana Studio is modifying files in my library - adding a “MUSICBRAINZ_TRACKID” tag to them.

Well, some of them … as I see no pattern as to which files it updates and which it doesn’t, but that right there terminates my use of, and interest in, the whole thing.

While relatively benign, in theory, this also causes every track (assuming it’ll eventually fiddle with all of them) to have to be backed up again, which with the size of my library is a non-trivial concern. Especially for the cloud backup stage.

So … that’s that little experiment done … I don’t need software fucking with my files without warning. Especially not to add random bits of data for which I have no bleedin’ use. There’s absolutely no reason that ID can’t live in the applications database.

8 Likes
22

Following this thread, intriguing :slight_smile:

23

I just ran a diff against my local backup to see how many files it had touched. In the 20+ hours it had been running, it had hit just over 5,000 files.

At that rate, it was going to take another 4,000 hours to finish “analyzing” my library.

That’s 166+ days.

7 Likes
24

So I guess you’re not going to sit and stare at the monitor until it finishes…

2 Likes
25

Interesting - as the trial period is only 30 days! :laughing:

3 Likes
26

Interesting find. This might be the trigger, why iTunes starts messing with the star ratings.

27

Then again, who does a trial with 40 yottabytes of music? :wink:

28

Well, if you want to find out, if the new application can handle your big data base. They all work fine with just a handful of files.

2 Likes
29

I’m just glad I caught it before a) it touched all my files and b) the cloud backup had started.

Right now I’m restoring the library from my local backup, so those files won’t show as changed when the cloud backup runs.

The ability to automatically fill-in/show metadata not present in a file from a properly curated source is a useful feature. It works well in Roon. But Roon doesn’t take half a year to process my library, nor does it fiddle with my files.

So … nice potential feature, poorly implemented (or not well thought through, at least). And since it cannot, at present, be disabled, that means no Audirvana Studio for me even if everything else was perfect.

5 Likes
30

I also wonder what Audirvana Studio is doing since my local audio files already have MusicBrainz and Beatunes tags embeded in the files. I have a library of about 12,000 songs and Audirvana Studio started the analysis at 15:30 local time today. I’ll report back on roughly how long it takes to analyze my relatively small library.

2 Likes
31

Most of the complaints I hear about Roon relate to the price - either that it is a subscription, or that the lifetime license is $699.

But for that price you can run as many clients, users and end-points, all playing different tracks, to different devices, as you want.

Want to use Audirvana Studio for two people simultaneously? Well, now it is going to cost you more than Roon would. Never mind that Studio doesn’t have anywhere near the same level of functionality, capability or polish.

And if it’s just you, but you want to run it on your main computer and a laptop? You have to fuck about logging out of your account on one machine (from that machine) so that you can log in on your other. Forget, and you’re buggered.

That’s all ignoring all the other issues with it, including many not fixed from 3.5. Obvious stuff here, like menu options taking you nowhere, buttons that don’t do anything, settings that won’t stick/save.

And then there’s the big one … it adding tags to your files. Which seems relatively benign (it isn’t modifying existing tags as far as I can tell, just adding an ID into MusicBrainz) until you realize it’s going to force all your files to get backed up again. Not necessarily a big deal if you have a small library and do local backups. But a real problem if you’re using cloud-based backup, especially with a large library and/or limited bandwidth.

It’s a real shame, as I have used Audirvana for a long time on my laptop when traveling. The $79 price was perfectly reasonable, and it was light enough to fit well in that scenario.

Now, for $2 more a month, I might as well put a new Roon license on that laptop and just skip Audirvana entirely. And I get way more capability for that $24/year.

If it wasn’t for the potential of Apple’s move with lossless/hi-res music potentially holing TIDAL and Qobuz below the waterline, I’d do another lifetime sub. But I’ll let that one play out a while before I do.

…

From my perspective, it is quite a feat to turn a long-term customer away from your product to the extent that they move to a “more expensive” competitor … and in doing so force them to realize that it’s actually cheaper given what the respective tools require/do … in less than 24 hours.

11 Likes
32

Welp, looks like I’m going to look into Roon, then.

By the way, any Windows 10 users here? Is the Kernel mode a big improvement?

1 Like
33

Out of purely morbid curiosity*, and because it was a slow day, I tried out the just-updated version 1.2 of Audirvana Studio.

One of the things the update adds is an off-by-default toggle to control whether Audirvana actually updates your files with the MUSICBRAINZ_TRACKID. That does, indeed, stop Audirvana from updating your files. Though if you choose to enable Replay Gain, it will still put those tags (eight, in total, I think) in all your files, so bear in mind the toggle is JUST for the Music Brainz stuff.

So that’s one step foward.

It still goes through it’s “Analysis” step which, while faster still hasn’t actually completed. And despite an explicit statement that you can quit the app and restart and it will pick up where it left off, it seems to start from scratch every time. And this step still chews up CPU usage and incurs a lot of I/O.

DLNA/UPnP replay seems to have gotten worse. It is not correctly finding/identifying the DLNA/UPnP capable devices on my network. Sometimes it sees them, sometimes it doesn’t, and when it does, sometimes it is showing incorrect names.

Some of the UX issues have been resolved (menus/buttons going nowhere), but most of the ones I ran in to have not. For example, plug-in settings still don’t save correctly, filtering/sorting doesn’t work right.

And then scrolling through Albums is suddenly a VERY staccato/jerky affair. 1.0 was pretty much the same a Audirvana 3.5 in that regard, now it’s markedly worse.

This is a long way from being software that’s in a state I would pay for, much less do so on a subscription basis. Given how many obvious, and frequently broadly reported, bugs were still present in A+ 3.5 I would not hold my breath on this stuff getting fixed quickly.

Too bad …

*It really is morbid curiosity; my initial Audirvana Studio experience was so bad, coupled with the basic issues still present in A+ 3.5 after 3+ years, killed any interest I have in actually using Studio - to the point I simply bought another Roon subscription to dedicate to my laptop.

10 Likes
34

Is Audirvana the ‘new coke’ release of audio software :joy: ? Or is there a better example?

2 Likes
35

Wow. Thank you @Torq for all those insights.

I had similar issues with my local files and UX wise, it’s a pur mess. Tiny fonts and icons. Hardly legible on my 28" 4k screen and functionality hidden in places where I would not have expected to find it.

All in all this is another great example of most likely NOT testing with real users (except us) and probably not having employed an (experienced) UX designer.

Quote of project leader to the developers: " Just make it dark themed and copy some of the Screenshots (Roon) I sent you and we’re good."
“Who knows better than us, right?” :stuck_out_tongue_winking_eye:

On a positive note, I could not objectively discern audible differences between Audirvana Studio and Roon.

4 Likes
36

I’ve done digital captures of the output data from both Roon and Audirvana and they’re identical*, whether used remotely via any of several transports/end-points or when directly feeding USB and S/PDIF outputs on the computer they’re running on.

The bit-level/sample data is identical, provided neither has any DSP or upsampling nor other processing applied. Which is what you’d expect.

This is using protocols and transports that do not have embedded clock data, so there’s no opportunity for the source to affect sample jitter/timing** and measured/captured performance is identical.

So that really leaves the potential for sound quality differences to be purely in the realm of “Software “A” uses less CPU/is coded more closely to the metal than B***, so causes less electrical noise”. Which simple doesn’t apply if you’re using either an optical connection, or an air-gapped end-point using a protocol with no embedded sample clock.

Though for giggles, I’ve done high-resolution analog captures at the DAC, both directly and remotely connected, and found no reproducible differences**** in its outputs there either.

The mechanism by which they might sound different, without applying optional processing, remains a mystery to me, outside poorly regulated/borderline hardware.

Which is fortunate, since I hear no difference between Audirvana, Roon, JRiver etc. either. The only software I ever was convinced made an audible difference was Amarra … and I’m fairly sure that was twiddling bits (I never got around to do the digital captures to find out for sure there).

But that’s me …

*I did note that Audirvana Studio failed the RME Bit-Perfect test at 24-bit depth, but I haven’t circled back to see if that was a transient issue or an actual problem … on the basis that software was unusable as-is anyway and thus it didn’t really matter.
**Not that most user-land software has the ability to affect such things anyway; that’s at best buried in the driver and more commonly in low-level non-addressable hardware.
***Modern CPUs, by which I mean anything from about the last 25-30 years, doesn’t directly execute the code you give it, not even if you write in assembly language. Instead its gets translated into different instruction forms and optimized/re-written by the CPU. There’s a good chance it won’t even execute in the order your laboriously hand-optimized code says it will. It might execute two different paths and just throw away the one it decides is incorrect. There’s not even a guarantee that the same sequence of instructions will be translated to the same low-level operations twice in a row. So it’s all-but impossible to optimize this manually … which means without special tools you can’t even be sure what the CPU is really doing behind the scenes. The only real guarantees are the memory and register states for a given instruction, post retirement.
****Just ultra-low level random variations which are entirely attributable to measurement-pass variation.

11 Likes
37

You’re taking your stuff seriously, @Torq!

:wink::+1:

5 Likes
38

I haven’t studied the wire protocols so I wanted to ask if your measurement setup is apples-to-apples with a typical pc using the standard usb out.

Seems like there’s room for things to get mucked up after the streaming app has delivered the data to the o/s and before it exits the usb interface at the dac end.

Maybe I’m misunderstanding the process, just trying to ask if real-world issues could cause “not bit perfect” even though the streaming service itself is bit perfect.

1 Like
39

The packet capture (for USB) was done from the USB output of the source devices (a Windows laptop, a MacBook Pro and a couple of Roon and DLNA end-points), via a spec-compliant USB 2.0 cable.

So it’s after all the application, OS an driver pieces - all of which were set to do no extra processing/resampling.

Only if the OS isn’t setup correctly, its USB driver isn’t working properly or the USB output/card/PHY is broken in some way.

Once the data reaches the USB PHY all that’s left to screw things up is either at the USB cable or the USB interface on the DAC itself. The player application has no possible influence over those, so they’re not relevant to comparing whether the players are outputting the same data or not.

UAC 2/USB Audio is not error-corrected, so if a bit got flipped after the physical interface on the USB output it cannot be corrected (there is no ECC code, nor a retry/resend mechanism). The protocol does include a checksum which allows you to verify that the data sent is what received, but there’s no way to “correct” it if it’s wrong.

In practice, with spec-compliant USB cables, such bit-level errors are extremely uncommon. I’ve left tests of this nature running for days at a time, with $10 Amazon USB cables at the limit of the USB length specification, and not seen any bit-level errors.

It’s worth noting I repeated the tests using S/PDIF and various network protocols and the results were the same. The sample data was bit-perfect as long as no extra processing (resampling, DPS, stupid OS mixer settings) was active.

8 Likes
40

Anyone have a good equalizer VST3 to recommend for use with Audirvana? Ideally something like PEACE/EQ APO where I can customize the frequencies and q-values