advertisement


96kHz/24 bit v 192kHz/24 bit from Mac Mini?

wilsonlaidlaw

wilsonlaidlaw

pfm Member
When I came back from my French house this year, I brought the new Benchmark DAC-2L I had bought there back with me. I am streaming from a 2011 Mac Mini via USB, mostly from ALAC files uploaded from CD's. I used to do this to my previous M-DAC (now on my bedroom system) via an M2Tech Hiface 2 re-clocker but given the DAC-2L's asynchronous Ultra-Lock 2 system, the Hiface should not be necessary. The DAC then connects via XLR cables to a Krell Vanguard Integrated amp and then to ATC SCM19 Mk.1 speakers plus an ATC powered C1 sub woofer. My set up in France is not wildly different except in a much bigger room, I am using Krell Evolution pre-amp and Krell Evo 2250 power amp to the larger ATC SCM40 Mk.2 speakers with a Monitor Audio Gold 15" powered sub. The Mk.2 ATC speakers are noticeably gentler and less painfully bright than the Mk.1 versions.

I normally have the Mac Mini set to output on USB 2.1 to the DAC-2L at 192kHz/24 bit. Since coming back to the UK, I notice, particularly for solo classical piano music, that it sounds better, smoother and less over-bright/harsh with the output set to 96kHz/24 bit than 192kHz/24bit. It has been suggested to me that the Mac Mini's audio chips, which I think are Intel on the main board, can really struggle with 192 kHz and will produce artefacts, which make the music sound odd and unpleasant. In the UK I am in a much smaller room, closer to the ATC Mk1 speakers, so I may just be noticing what was always there at 192kHz but the brighter speakers just make it more apparent, together with being a lot closer to the firing line.

Thoughts folks?
 
What playback software are you using on the Mini? try Audirvana+ this automaticaly sets the output to the same resolution that media was recorded at. There is no point in setting the Macs output to 24/192 or whatever if the material is recorded at 16/44.
 
My guess is that the OP is using iTunes and that may be the culprit. I suggest that he tries an evaluation copy of a 'proper' music player. I use Amarra and have no problems with Hi Res files. I can even mix resolutions as Amarra detects them on-the-fly. I have a 2010 and a 2012 Mac Mini and both are fine but I do use Firewire.

Cheers,

DV
 
I have tried Audirvana and another audio output "improver" (Soundflower with Audio Essentials) and did not like the results. It made the music far too warm and rich sounding for my personal taste. The other point is that my more recent tracks are all studio grade downloads at 96/24, so other than the option of auto setting the output with Audirvana or similar, I should not get any negative effects from up sampling the 44kHz/16 bit CD files to 96/24. I do seem to get adverse effects by up sampling to 192/24 and as I only have a couple of downloads at this bit rate, there is little to be gained. What I would like to know is if others have experienced this problem, which I am fairly sure, originates in the Mac Mini not in the DAC-2L.
 
The other point about players other than iTunes is that I really dislike their interface and their remote to my iPad worked poorly, whereas iOS iTunes Remote works beautifully.
 
It has been suggested to me that the Mac Mini's audio chips, which I think are Intel on the main board, can really struggle with 192 kHz and will produce artefacts
Click to expand...
But your dac is doing all the work isn't it?
 
wilsonlaidlaw said:
The other point is that my more recent tracks are all studio grade downloads at 96/24, so other than the option of auto setting the output with Audirvana or similar, I should not get any negative effects from up sampling the 44kHz/16 bit CD files to 96/24. I do seem to get adverse effects by up sampling to 192/24 and as I only have a couple of downloads at this bit rate, there is little to be gained. What I would like to know is if others have experienced this problem, which I am fairly sure, originates in the Mac Mini not in the DAC-2L.
Click to expand...

Actually upsampling from 44.1k to 96k can easily add flaws to the result unless the resampling is done well.

I can't really see the point of doing anything other than sending the data to the DAC *at its source rate* unless you want the interposed resampling to alter the sound in some specific way - or have a DAC that doesn't work correctly at some rates.

Personally I've always set up the sound system on my machines so the output 'follows the rate' and sends the LPCM to the DAC via USB at a rate that is always the same as the sample rate of the source material being played. If you have a high quality DAC I'd expect it to do a good job on this. Whereas I've seen and measured too many foul-ups in computer software to trust them without a lot of careful measurements and tests.

Jim
 
I have just looked at another Mac Player, Fidelia. The end results may be good but its library interface looks like something somebody designed on a Sinclair ZX81. It is close to unusable as far as I am concerned.
 
Try BitPerfect which works seamlessly with iTunes whilst improving the overall performance.
You can buy it for peanuts from the Apple Store.

R
 
One of the reasons I mentioned Amarra is that you can if you wish use the iTunes interface. However one of the reasons I chose it is that I can't stand iTunes!

Amarra is also a professional solution used by recording studios so if the music sounds bad with this player then the fault lies elsewhere in the OPs system.

Cheers,

DV
 
wilsonlaidlaw said:
Not the up-rezzing, from 44/16 to 96/24 which is being done by the audio chips on the main board.
Click to expand...

Do you know that? I'd be interested in a link which tells me that. My understanding is that CoreAudio uses a sample rate converter plugin when the sample rates that the application is supplying and the settings for the hardware differ, and that this converter can run in various modes giving different quality settings. I'd have thought the possibilities for getting this wrong are quite high, as i can't immediately see how you configure how this behaves on the machine i'm typing this on.

Assuming the hardware supports sample rate conversion, I imagine that would appear to the application as AU being configured with one sample rate (so buffers of audio are supplied at this given rate) whilst the output is running at a different rate. Basically if the player sees 96/24, then it must supply that data.
 
If you have an ALAC file, which is stored at 44kHz/16 bit but which is exiting the USB socket as say 96kHz/24 bit (I can see this from the LED display on the front of the Benchmark DAC-2L) where else could it be being done except in the chips on the motherboard. There does not appear to be a separate audio card in the Mac Mini. If you look at CPU usage, it does not seem to increase, if you up-rez from 44/16 to 96/24, which indicates to me, that the up sampling is not being done by the CPU but must be being done by other chips on the motherboard.
 
I have downloaded Amarra and this is the first player I have actually quite liked. I will need to play around with the settings a bit as it sounds as if it is adding a bit of reverberation, which I don't need.
 
wilsonlaidlaw said:
If you have an ALAC file, which is stored at 44kHz/16 bit but which is exiting the USB socket as say 96kHz/24 bit (I can see this from the LED display on the front of the Benchmark DAC-2L) where else could it be being done except in the chips on the motherboard. There does not appear to be a separate audio card in the Mac Mini. If you look at CPU usage, it does not seem to increase, if you up-rez from 44/16 to 96/24, which indicates to me, that the up sampling is not being done by the CPU but must be being done by other chips on the motherboard.
Click to expand...

The audio interface on the machine is set to a single sample rate and bit depth. When the application starts up, and attaches to CoreAudio (the audio playing framework in OSX) it is presented with various outputs that it can attach to - for example, it can select the 'default' settings, which will return the sample rate and bit depth that you have selected for the machine (the settings that the audio hardware is running). It can also select different settings (for example, the bit depth and sample rate of the audio it is wanting to play).

If it selected a different output setting, then CoreAudio inserts a sample rate converter to process the buffer the application writes to, to convert it to the settings that the hardware needs. This is done in software using the Audio Converter Services (and you can directly use these to do conversions yourself if you so require).

Have a look here:

https://developer.apple.com/reference/audiotoolbox/1653485-audio_converter_services?language=objc

You can specify all sorts of conversion options, such as these:

https://developer.apple.com/referen...ample_rate_conversion_complexit?language=objc

Notice there is a reference to the runtime complexity, the CPU time spent running these algorithms. If I were to guess, i'd say the three modes are a straight linear interpolation (which is pretty shockingly poor, so maybe the name is bad, and it's really a spline interpolator), standard will be some sort of sinc interpolator, as will 'studio quality', the difference being the length of the window applied to the sinc function.

Let's estimate the load this would cause on the machine.

To give you an idea, with sinc interpolation, you can get very good conversion between sample rates using a window of around 15 samples either side of the sample, so basically using 30 input samples to generate one output sample. For each input, you'll need to do a multiply add, assuming a lookup table is free to load into the processor cache. So, for each output sample (say at 96Khz) the machine will need to run 30 multiply adds per channel, so let's round it up to 100 floating point operations, 100 thousand times a second, so that's 10 million per second.

A modern CPU core runs at 3Ghz, and can perform 4 floating point multiply/adds per clock cycle. That means 12 billion floating point operations per core per second is the procesisng power of the processor, and we need 10 million, so that's roughly 0.1% CPU load added by a sample rate converter running the above.

Doubling the output sample rate will double the load. Even if i'm out by a factor of 10 (possibly but unlikely) we'll be in the 1-2% load for doing this in software on one core, and a modern machine typically has 4.

I'm not saying it's not done in hardware, i'm saying if it is done in software, you won't notice on modern machines, and the quality could be low due to the default setting being used rather than the 'mastering' quality converter.
 
I'm using a 2012 Mac Mini with 8GB of RAM running iTunes & Audirvana +. iTunes runs at 44.1k & Audirvana at the native bit rate up to MHz for DSD files into a M-DAC+ via USB.
no issues but my least favourite part of of iTunes is the interface, especially with classical stuff.
 
busb said:
...but my least favourite part of of iTunes is the interface, especially with classical stuff.
Click to expand...

I much prefer it to the very crude native Audirvana interface. However the native Audirvana gives access to its remote app, and Qobuz which is why I use it.
 
You must log in or register to reply here.


advertisement


Back
Top