Audirvana Plus 3 (official thread)

[Jud]

1 hour ago, astrostar59 said:

I am also wondering if A+ is used to upsample, but that resolution was not the exact same value as the end DAC up samples to, are we not having the situation where the (possible) degradation will happen twice. If however, we send data to the end DACs sample rate upper limit, then would that 'shut down' the resampling function of that DAC. Unless it was disabled by doing this, in my mind we have 2 stage resampling / processing of the data. 

 

And what you are saying, all this is to negate the effect of the digital filter?

 

Getting more into the internals of most DACs:

 

- If you send it a RedBook file, it does 3 rounds of doubling the sample rate, first to 88.2kHz, then 176.4kHz, then 352.8kHz

 

- If you send it a file based on a 48kHz rate, it doubles until it reaches 384kHz, so a 96kHz input would go through two rounds of doubling

 

- From there (still on the DAC chip) the bitstream goes to sigma-delta modulation, where the sample rate is increased to a minimum of 2.8MHz, very often in newer DACs higher than that (in ESS DAC chips, I think it may wind up higher than 40MHz before all is done)

 

So if you have a 44.1kHz file and you choose to upsample in A+ to the highest rate the DAC accepts or the nearest 2x rate (let's say 176.4 or 192kHz for a DAC that doesn't take DSD input), you've avoided a couple rounds of doubling.  But more important than that, you have the opportunity, for the initial and most critical stage of upsampling, to employ different and possibly "better" filtering (for some value of "better," whether it be particular measurements, your listening enjoyment, or both) in software than the filters programmed into the firmware in your DAC chip.

[audiocanyon]

Let me try the attachment again.  I don't see it in my post.

Sierra Direct Mode.pdf

[audiocanyon]

11 hours ago, TommyCat said:

Thanks to jhwalker discovered my dac does upsample to 211 kHz (Benchmark HGC dac2).  Question remains whether or not to upsample before sending to dac (my tendency would be not to; assume Benchmark has figured out to do that more efficiently than my guesswork inside of A+)

Since I have the Benchmark DAC2 also, I'm inspired from these various posts to do some A/B comparisons between the A+ upsampling and John Siau's.  Gotta love audio tweaking...

Sierra Direct Mode.pdf

[RunHomeSlow]

Here is my Direct Mode Link for Sierra... All you need is inside.

Again... READ CAREFULLY and do ALL the steps.

 

https://www.dropbox.com/s/ex3ksuj1calu7p6/AudirvanaSierraDirectMode.zip?dl=0

 

Since Apple will release 10.12.4 surely in a couple of days... already at beta8,

you should wait till it is released if you don't want to do it again...

because you will have to do it everytime you update you macOS.

[Jud]

24 minutes ago, audiocanyon said:

Since I have the Benchmark DAC2 also, I'm inspired from these various posts to do some A/B comparisons between the A+ upsampling and John Siau's.  Gotta love audio tweaking...

Sierra Direct Mode.pdf

 

It will be interesting to see whether you even think you hear a difference.

 

Edit: You may want to try upsampling to DSD128 and see what that sounds like to you, if the Benchmark will accept that; if not, you can try DSD64.

 

Further edit: DSD will initially be lower volume, I believe.  So for any comparison you would want volumes equalized.

[Innocent Bystander]

10 hours ago, bhobba said:

 

 

This is a big decision for me.   I will need to a new Mac.  If its coming will get a new Mac and Audirvana 3 straight away.   Already have a copy of Audirvana 1 but since then my Mac bit the dust.

 

 

If I was at that point now, and if it is to be dedicated to music, I'd be seriously considering Melco instead, which is not a lot more expensive, doesn't have the RF contamination that needs an isolator to prevent degradation of sound quality in susceptible DACs like Hugo, and avoids the very frustrating limitations of A+'s library. Otherwise a secondhand late 2012 Mac Mini, which I think was the last user-upgradeable one, though now starting to get a bit long in the tooth. 

[progfan2]

Hi Jud,

 

I running 2.6 have found using any DSD mode on my 2011 I5 Mini, stresses the CPU with heat running about 95C.  Would you know if A+ 3 using DSD mode has the same CPU demands.

 

Thanks,

Mark

[stefano_mbp]

@damien There are two bugs related with SysOptimizer:

- if I check Disable Spotlight option in the SysOptimizer panel, I play something in Audirvana (3.0.3) and then I quit Audirvana, no smart folders in OS Mail will work anymore, I must restart my MBP to let them work again. If I uncheck the option all is ok.

- this morning I updated iTunes to 12.6 and it started to crash after using Audirvana (3.0.3). I could solve the problem unchecking Optimize System for Audio Playback in the SysOptimizer panel and restarting my MBP

 

[Innocent Bystander]

1 hour ago, 27Globes said:

Trying out Audirvana 3.0... but find Library Manager confusing.. trying figure out how to filter out library by genre. Looked over user manual... still don't get it.  Can someone layout the procedure... would appreciate.  Does not seem as accessible as it was when integrated with iTunes.  Thanks.

 

I never used A+'s iTunes integration, partly because @Damien78 claimed higher quality sound not using it, and partly because I wasn't impressed by iTunes previously on an iPhone. However, the library manager is A+'s weak point. A real pain if you have less than perfect and consistent metadata (and sometimes still does unexpected things when metadata is correct and consistent, such as doubled up tracks of alternate resolutions if you have more than one version of the same album in different resolutions, and some tracks simply not appearing when others do). It is worst with classical, with its variability in assignment of genre whether on CD or downloads, e.g. different recordings of the same piece of music variously being assigned 'symphony', 'orchestral' or simply 'classical' (likewise when it comes to assigning artist). There are quite a few people waiting in hope for Damien to implement a facility for browsing by file file/folder structure/location to solve these problems, including me!

[Jud]

25 minutes ago, progfan2 said:

Hi Jud,

 

I running 2.6 have found using any DSD mode on my 2011 I5 Mini, stresses the CPU with heat running about 95C.  Would you know if A+ 3 using DSD mode has the same CPU demands.

 

Thanks,

Mark

 

I'm unaware of anything that would change the CPU usage, but Damien and mansr would obviously know better than I do.

[RunHomeSlow]

I think he meant Upsampling to DSD stress the CPU....

[mansr]

31 minutes ago, Jud said:

 

I'm unaware of anything that would change the CPU usage, but Damien and mansr would obviously know better than I do.

 

I haven't done anything to reduce CPU usage. Perhaps I should.

[RunHomeSlow]

4 minutes ago, mansr said:

 

I haven't done anything to reduce CPU usage. Perhaps I should.

 

YES! if you can that would be awesome...

My CPU works at 250% for half the time, then it rest for half the time at 0% when the song is loaded...

But if you can reduce the loading time CPU usage less then 250% to 100% that you be perfect  :-)

 

That is for only Upsampling PCM to DSD... playing DSD at DSD source doesn't require CPU more then normal usage.

[astrostar59]

1 hour ago, Innocent Bystander said:

If I was at that point now, and if it is to be dedicated to music, I'd be seriously considering Melco instead, which is not a lot more expensive, doesn't have the RF contamination that needs an isolator to prevent degradation of sound quality in susceptible DACs like Hugo, and avoids the very frustrating limitations of A+'s library. Otherwise a secondhand late 2012 Mac Mini, which I think was the last user-upgradeable one, though now starting to get a bit long in the tooth. 

I disagree. A Mac Mini is £500 approx. And add the Uptone Audio MMK at £130 and a Linear PS at around £150 and you then have a very user friendly, updatable and compact solution. I get the music server pre-built ideas, but I much prefer to be able to tweak and update a computer.

[Jud]

19 minutes ago, mansr said:

 

I haven't done anything to reduce CPU usage. Perhaps I should.

 

It depends.  If reducing usage = greater efficiency, wonderful.  If it = slowing down the conversion process noticeably, obviously not as wonderful.

 

I'm guessing your code was already pretty efficient, so what would you contemplate changing?  Taking advantage of more modern CPU capabilities where they exist...?

[progfan2]

Hi RHS,

 

Yes, I'm having the same issue with up sampling to DSD. Though I don't seem to go back to 100%. Can you tell me what you are using for system preferences?

 

[Speed Racer]

5 hours ago, Jud said:

 

 

This does not have to do with "greater efficiency" in the way that you are thinking of it.  The filtering on a DAC chip isn't difficult for the chip and is trivial for a computer.

 

What upsampling in software is about is, first, the ability to run *better* (to some extent, more sophisticated) filtering on a computer than can be programmed into a DAC chip.  Compared to the CPU in a computer, a DAC chip has more limited resources to run sophisticated filtering.  FPGAs have more resources for this purpose than DAC chips (though still not quite what a computer CPU has), and if you look at DACs that use FPGAs for this purpose you will find many of them among the most expensive DACs available (though Chord has notably brought DACs with FPGAs well down in price point).

 

Now Chord has a very specific philosophy about what kind of filtering it does in its DACs, which brings up the second point about software upsampling: It enables you to have *choice* in the type of filtering you do or the parameters you set for the filtering.  So you don't have to be bound to the particular choices made by the designer at Chord or the designer of your DAC or the designer of the firmware-programmed filtering in the chip inside your DAC.  Some people don't care to fuss with changing, which is very understandable.  Of course with Audirvana Plus if you don't want to play around with different filter settings, you have the choice not to upsample, or if you do upsample, to leave all the filter parameters at their default settings.

 

Upsampling DACs do their upsampling in real time. DACs don't have limited resources to do this. They have all the resources they need to do this. While a computer may have more horsepower than a DAC, it does not do upsampling in real time.

[RunHomeSlow]

6 minutes ago, progfan2 said:

Hi RHS,

 

Yes, I'm having the same issue with up sampling to DSD. Though I don't seem to go back to 100%. Can you tell me what you are using for system preferences?

 

 

 

Hello, i was using before DSD upsampling with PCM upsampling... 8, 1.300.000, 1.02, 200, .72

but then came A+ with DSD upsampling... With those, i was always at 255% or more, CPU never resting...

with low low steepness i guessed.

 

Then i changed with the settings of Damien in the manual of v3: 45, 500 000, 1.00, 200, .73 (page 37)

was still good to me, then yesterday Watercourse give his... so i'm there today as in pictures.

 

I too found really difficult for myself to play with those, so i rely on people choices and test them

[mansr]

1 hour ago, Jud said:

 

It depends.  If reducing usage = greater efficiency, wonderful.  If it = slowing down the conversion process noticeably, obviously not as wonderful.

 

I'm guessing your code was already pretty efficient, so what would you contemplate changing?  Taking advantage of more modern CPU capabilities where they exist...?

 

There's definitely room for more optimisations. Bear in mind, however, that much of the time is spent in the iZotope resampler, and that's none of my concern.

[mansr]

9 minutes ago, Speed Racer said:

While a computer may have more horsepower than a DAC, it does not do upsampling in real time.

 

What makes you say this?

[Speed Racer]

2 minutes ago, mansr said:

 

What makes you say this?

 

https://en.wikipedia.org/wiki/Real-time_computing#Real-time_in_digital_signal_processing

[RunHomeSlow]

when i open activity monitor, in the graph at lower window... that is approx for 5 mins of music... i'm at 0% for half the time now with Watercourse and Damien settings... if mansr can to it better optimizing filters to be like regular DSD playing it would be great... for CPU  :-)

[mansr]

11 minutes ago, Speed Racer said:

 

https://en.wikipedia.org/wiki/Real-time_computing#Real-time_in_digital_signal_processing

 

I fail to see the relevance of that.

[jhwalker]

6 minutes ago, mansr said:

 

I fail to see the relevance of that.

 

Exactly - there is no such thing as "real time" processing (maybe when we get to quantum computing). 

 

Processing takes time, and it's code that does it - whether the code is in a chip (hardware) somewhere, or loaded in software, or . . . as long as the processing takes place ahead of playback needs, the difference between "real time" (misnomer anyway) and "a very short time" is inconsequential.  And computers indisputably have more flexibility for updated filters, etc., than can be encoded in a chip.

[Speed Racer]

10 minutes ago, mansr said:

 

I fail to see the relevance of that.

 

I can't help that you don't understand the difference between real-time and high-performance systems.