Explorations concerning computer-based audio

[kaka]

 

Antipodes;97744 wrote:

You need to ditch that compressed stuff anyway, WAV and AIFF rule. FLAC and ALAC are for backups.

 

I've found a Windows player that thrashes Foobar - I'm running a trial version of J River Media Jukebox and am getting vastly improved results playing Flac 16/44.1 I saw it mentioned on the Empirical Audio forum and have been meaning to try it. The WASAPI memory player failed.

 

I tested some 16/44.1 wav vs Flac 44/161 and yes they were better in Foobar, but the 24/96 I upsampled (with Adobe Audition) were aiff and J River can't play them. I will upsample wav 24/96 and see how they go, and try some more 16/44.1 wav

[Gary.M1553552674]

Don't know how relevant this is... I run Linux on all machines here except my DVD player which is XP.

 

In my office when I working on PCBs I like to listen as well. I have a pair of Wharfedale Diamond 8.1s on stands and a small sub. So it has the potential to sound good.

 

Currently I've been running an M-Audio Mobile Pre (USB) with playback via Decibel Player (a simple Linux front end).

 

Linux has seen changes in recent times and PulseAudio, which is at the heart now, seems to have a very good resampler.

 

Sound quality certainly seems good at the moment...

 

Those who like to delve deeper might want to play. You can make your Windows box dual boot Linux very easily. Linux Mint 7 would be a good easy introductory install.

[Antipodes1553552706]

I did choose AIFF over WAV for tag reasons, but couldn't hear any difference between them. Yes, I can say that I do agree with Steve on sequencing the bit depth before the resampling. They are done separately in SampleManager and you can batch them together to happen one after the other.

 

I will fire up a Win7 OS and give J River a try.

 

Linux will be interesting when I get to it properly, but I suspect it needs a fair bit of experimentation.

[nixon76]

 

Antipodes;98080 wrote:

Yes, I can say that I do agree with Steve on sequencing the bit depth before the resampling.

 

This means 16bit->24bit and then 44.1Hz->96Hz (or vice-versa) ?

[Antipodes1553552706]

Yes, 16bit->24bit and then 44.1Hz->96Hz. The main problem I have with SampleManager is that some cache seems to be filling up during the batch runs and this restricts how many albums I can do at a time. I haven't bothered finding out more as it isn't a hardship. I also turn off analysing the files first, and turn off 'undo', and that means I can run longer batches. SampleManager will also reject some files if they are a bit dodgy. I mainly had problems with high bit rate FLACs from places like B&W. I will download them again as WAV files and see how that goes.

[Antipodes1553552706]

Hi kaka, I recently tried J River on a nice sounding little Netbook using ASIO and then WASAPI and it sounded pretty good, as you say better than Foobar, until I used it on battery power and then it sounded very nice indeed. In general I find all PC versions to not have the PRAT or propulsive and tight bass of the Macs. In terms of mid and treble tonality I can get a better result out of the best PC versions we have discussed here, but Amarra fixes the slight hardness and glaze with the Macs and leaps ahead with its smooth and vivid presentation. All in all, with Amarra, I think the Mac wins but at a cost - the equivalent powered PC is cheaper and Amarra is a significant added cost. Without Amarra, the mid and treble tonality may (or may not) be preferred on the PCs - it will depend on the implementation.

[kaka]

I was playing J River Media Centre 14 yesterday - it can be made to play from memory by the way, and in theory via WASAPI though I can't get WASAPI to work with the Offramp.

 

The sound was somewhat homogenised at 16/44.1 (all the various threads mixed together, without the right prominence) but when I used its internal DSP to take it out to 24/96 things separated out and became very good.

[Ernie1553552694]

Hard upsampling is obviously worth the disk space then?

[kaka]

That was on the fly upsampling - the Adobe product on the PC was clunky. I should have a TradeMe probook turn up tomorrow so I can try the the product Antipodes has been using.

 

Future shock will set in, 27 years of PC experience, zilch mac.

[Ernie1553552694]

Good move. You won't look back.

[nixon76]

 

kaka;98257 wrote:

 

Future shock will set in, 27 years of PC experience, zilch mac.

 

Inevitably some things will really (really) annoy you about OS-X. It's not perfect. If you can accept that, things will be much easier. Good luck!

[Gary.M1553552674]

Don't underestimate the Linux kernel. If you have a PC with Windows running, you can create a 12GB empty partition, download and burn the Linux Mint 7 iso, boot from the CD and install to that empty space have a dual boot install running alongside Windows in 30 minutes. Then play with that and learn too...

[kaka]

 

Gary.M;98279 wrote:

Don't underestimate the Linux kernel. If you have a PC with Windows running, you can create a 12GB empty partition, download and burn the Linux Mint 7 iso, boot from the CD and install to that empty space have a dual boot install running alongside Windows in 30 minutes. Then play with that and learn too...

 

Amarra Mini is the attraction, rather than the OS. The makers (Sonic Studio) have been deeply in the pro mastering studio arena for ages, and their products run on macs. I figure at the price I got the laptop for I can move it on with minimal pain if the Amarra trial downloads don't have me dancing in the aisles.

 

Have a skim through this if you haven't already - it certainly sounds like a product worth trying out.

http://www.computeraudiophile.com/content/Part-4-5-Symposium-Main-Listening-and-Equipment-Comparison-Session

[Antipodes1553552706]

I am busy collecting bits and building another computer specifically for running OS X at the moment, with carefully selected motherboard for the purpose, so haven't started down the Linux track yet. But it is on the list for sure. Still waiting for the iLock for the Amarra too. What are the best sounding players for Linux, anybody? I will do some Googling.

[Gary.M1553552674]

 

kaka;98281 wrote:

Have a skim through this if you haven't already - it certainly sounds like a product worth trying out.

 

http://www.computeraudiophile.com/content/Part-4-5-Symposium-Main-Listening-and-Equipment-Comparison-Session

 

An Interesting read. I'm not knocking the approach at all, I have an open mind on it. I'd just like to see you guys throw Linux in there and hear what your assessment is.

[Gary.M1553552674]

 

Antipodes;98282 wrote:

What are the best sounding players for Linux, anybody? I will do some Googling.

 

The Linux players are all front ends. The audio system is integrated into the kernel, and the decode is handled by one or two possible engines. The commonly used one is xine. Some links below...

 

Information and discussions...

 

http://www.diyaudio.com/wiki/index.php?page=LINUX+Audio

 

http://www.diyaudio.com/forums/showthread.php?s=&threadid=93315&perpage=25&pagenumber=1

 

http://www.diyaudio.com/forums/showthread.php?postid=1883595#post1883595

 

The long diyaudio linux thread starts in 2007 and a lot has changed since then so posts at the beginning are of historical interest only.

 

A couple of players as examples...

 

http://decibel.silent-blade.org/

 

http://www.atunes.org/

[Antipodes1553552706]

Thanks Gary. I will focus my googling on these.

 

The computer in my main system at home, the USB cable, Offramp 3 and AES cables are all almost burned in now. I am much more able to hear what Amarra is doing now and it sounds much more like an 'everything gets better', whereas my early impressions were clouded by new equipment.

 

It is similar to switching from our Katipo cable to our Reference cable. Detail is similar but tonal colour and coherence are far more real. The sound I associate with digital (dryness; 2D images; lack of connection between initial attack, body and decay; unnatural edge definition) all seem to go away. I say 'seem' because the reduction in these issues is maybe 90+%, but that assumes I know what 100% would sound like, which I can't as it is too difficult to know how much of an issue remains with the DAC or other parts of the system.

[SoundMark]

Antipodes,

 

I'm curious to know what USB cable and Offramp3 converter you're using.

 

I'm looking to purchase the Locus Polestar USB cable and the Offramp3-Superclock4 converter from Steve Nugent, and just wondered if you were using the same conponent/cable combo....

 

Cheers

Mark

[Antipodes1553552706]

Offramp3 with Superclock4, but with my own design USB cable. If you can stretch to the Axis (one level from the Polestar) I hear it is well worthwhile. I am intending to get an Axis in to compare.

[Antipodes1553552706]

Another quick recap of my (not definitive) conclusions.

1. Yes, you can beat conventional CD players and Transports with computer based audio. Don't look down your nose at computer based audio - it is theoretically and practically a superior technology to a CD Player, it is just that few are executing it fully for audio in simple products.

2. Probably the simplest way to get good or great sound is to use almost any computer and a Transporter or Modwright Transporter. Unpack compressed files (eg FLAC, ALAC) in the Computer, not the Transporter. The Linn solutions aren't bad either. At lower cost the Sonos and Squeezebox are good value for money. More seem to be coming, such as from PS Audio.

3. Going all out for the best sound and with a bit more choice over the DAC you use, a Mac/iTunes/Amarra solution is excellent provided you use something like the Offramp or Lynx L22 or EMU/PaceCar to feed your DAC. The Amarra site will tell you what devices work with it. The Empirical site will tell you what combinations work well. Don't think that feeding your DAC with the Mac/iTunes is going to get great sound - it won't. And you need to use good cables - including USB if you are using one. A Macbook Pro has enough connections and performance to do the job reasonably well, and sounds very good when on battery power.

[kaka]

 

kaka;98247 wrote:

I was playing J River Media Centre 14 yesterday - it can be made to play from memory by the way, and in theory via WASAPI though I can't get WASAPI to work with the Offramp.

 

 

 

The sound was somewhat homogenised at 16/44.1 (all the various threads mixed together, without the right prominence) but when I used its internal DSP to take it out to 24/96 things separated out and became very good.

 

Now that I have got the WASAPI side of JRiver Media Centre 14 going (under Player - Playback Options - Output Mode Settings there is a box ticked on "present 24 bit data in a 32 bit package.." - get rid of it) I am most impressed with the sound. It offers detail and dymnamics that Foobar simply doesn't, and is the best computer based sound I've heard so far.

 

And thats on the fly upsampling to 24/96 from flacs, so there would be more to come with pre-upsampled wavs and other optimisations.

[kaka]

 

Antipodes;98091 wrote:

Yes, 16bit->24bit and then 44.1Hz->96Hz. The main problem I have with SampleManager is that some cache seems to be filling up during the batch runs and this restricts how many albums I can do at a time.....

 

If the bit length is pushed out from 16 to 24 without resampling, am I right in thinking that all that happens is padding with 8 extra bits ?

 

If that is right, then could DBPpoweramp do that job while doing the conversion to AIFF, removing a step from Sample Manager ?

 

I've got to the stage of having a networked mac with Amarra, and 83 pages of Sample Manager manual printed out, and I am wondering about the quickest route through that process.

[Antipodes1553552706]

You can pad it with zeros or dither it to 24 bits, which is not the same thing.

 

Dithering is seen as beneficial in the recording process and almost universally used. The idea is to randomise the quantisation error, to reduce or eliminate quantisation error that is correlated with the signal. So in dithering to 24 bits, random (sort of) noise is added to the signal to enable it to be quantised to 24 bits without just padding it with zeros. In theory, once this is done once, then doing it successive times will not be beneficial as it can only cause bit errors.

 

However, kit like the Meridian 518 was almost universally lauded for the improvements it could bring dithering a signal from 16 bits to 24 bits, and it offered a range of noise shapes (so not random after all, but not correlated with the music). Some, including me, found that the 518 softened the sound too much, and this was in my view due to conscious choices made by the designer.

 

In theory, problems occur when you introduce successive stages of dithering where, for example, you offline dither 16 bits to 24 bits then play that with software that dithers the signal to enable digital volume control. In the first stage you might be reducing banding with a trade-off for some bit errors. In a later stage, if the banding has been eliminated you are just introducing bit errors.

 

If you are going to do multiple stages of digital manipulation of amplitude/wordlength (such as volume, DSP, xover etc) then it is arguably best to do it in a single conversion where the maths of each step are combined into a single conversion, with all the intermediate maths executed at say 48 bit depth or more.

 

But as with some of the previous discussion in this post, this isn't always true, as more real time manipulation makes it more difficult to control jitter. So, as with everything, there are no hard and fast rules. Try it and trust your ears.

 

With computer audio there are also some additional problems to overcome. With 16 bit audio in Vista and OS X, then Microsoft's or Apple's engineers redither your files to 24 bit by default. I have had arguments with my contacts at Microsoft on some of these issues, but it is hard to argue with someone that believes 192 kb/s MP3s are indistinguishable from higher resolution standards. So the choice can come down to doing it offline with a method you like the sound of or trust it to some computer engineers who don't have much appreciation for good sound.

 

Even if you can get 16 bit digital cleanly out of your computer, many DACs will redither to 24 bit and resample to 96kHz before converting anyway. They may do it better than your computer does, or may not, but the reason why computers have a definite advantage is that DACs suffer from doing it in real time. The whole push for upsampling several years ago was not about doing anything new (as the DAC chips did it already), it was about adding a separate stage that upsampled before feeding it to the DAC chip. The DCS stack has a whole separate and expensive box for the upsampling step.

 

Personally, I found selecting the dithering from 16 bit to 24 bit that I liked the sound of, to be best done offline and leave the DAC to do as little as possible. Once I found what I liked, I found dithering from 16 bit to 24 bit with the right algorithm to be as beneficial as resampling from 44.1 to 96 with an equally nice algorithm.

 

I don't see any point in padding the signal with zeros, except that it is probably what your DAC does, so eliminating that from the real time steps is a good thing. If you are going to dither to 24 bits then I recommend you should carefully audition the available methods before deciding to convert your whole audio collection. And keep the 16 bit original files in case you find a better way to process the files later.

 

There are no hard and fast rules here. You can't just apply the engineering theories as if they are all executed perfectly in the real world. If they were then not much of this whole thread would be relevant. I am convinced that the human ear/brain is highly susceptible to tiny phase distortions and that some are more easily processed by the ear/brain than others. I believe these are the issues that make the difference these days. The basics of low noise, dynamic range, tonal balance are sorted in stuff you can get from Harvey Norman. To go further you enter a world where the engineering models I was trained in are only rough instruments. You have to try permutations out for yourself and see what works, even if the theory says they shouldn't be beneficial. Only your ear/brain can tell you what enables you to connect with the music and what doesn't. So, regardless of all the science and deductive logic, you just have to try some things - and in particular try the things you think should sound worse, not just the things you think should sound better.

 

The Antipodes Audio cables are a very good case in point. The design includes things that shouldn't sound good, but they do, and you don't find them in other cables, most probably because the fact that they shouldn't work deters people from trying them.

 

In my case, offline dithering to 24 bit is worth it, and the iZotope MBIT+ works for me. To quote "iZotope's MBIT+™ dithering algorithm features psychoacoustic noise shaping, based on research with actual musical material". Others may have a different opinion, probably do, but I find that the differences before and after are all upside.

[kaka]

Thanks for that, and the hint on using iTunes to convert AIFF to AIFF.

 

Well I have established the first couple of things - the digital volume control in Amarra throws away dynamics so if I go that route it would be Amarra Mini, and I'm getting a softer less dynamic sound through Amarra from the Sample Managed 24/96 AIFF than the base 16/44s. More listening to do.

 

My network link to the mac is only working towards the mac, so it will take a little time to get the Sample Managed Wav 24/96s to where the JRiver player is. I will be able to compare flac and wav at 16/44, both formats on-the-fly upsampled to 24/96 and the Sample Manager pre-upsampled 24/96 on JRiver.

 

I have always wondered whether the Altmann is slightly laid back, but I don't have another dac to compare it with. I'm not finding the harsher, more edgy 16/44 sound objectionable, in fact it may be preferable.

[Antipodes1553552706]

I can understand that with the Altmann. I find with the more conventional DACs I am using that with 16/44 the edge definition is unnatural. With 24/96 notes start and stop (decay) more naturally, without any diminution of dynamics - quite the reverse, the fine gradations of dynamics are rendered better. The Altmann gives a fuller, richer and softer sound than the DACs I am using, which tend to be a bit faster and have more detail resolution. Two paths to the same end, I suppose.

 

My only slight reticence about the Amarra Mini is I wonder if as the software develops the Mini will be kept up with the pace. Otherwise I agree there is no particular need for the added features in the full Amarra, and around a grand difference in price.

 

One of the things to consider is a bit of an analog of the LCD native resolution thing. As my DACs will convert anything less than 24/96 to 24/96 then I am getting the benefit of doing that off-line and relieving the DACs of that real-time task. If you set the Altmann to 16/44 then that removes a reason to process the files.