Explorations concerning computer-based audio

[Antipodes1553552706]

I have decided to use SampleManager for the conversion. It has an excellent sounding resampling algorithm and the price is great at just USD79. I have made myself a second Hackintosh to cope with all this processing and still listen to music.

 

dBPoweramp is a bit odd. I used it to convert FLAC to AIFF, and iTunes coped with the resulting files and tags. But SampleManager reckoned the files were corrupt and wouldn't open them. I remember having a similar experience with dBPoweramp when converting to ALAC once (though not last time). The cure was to use iTunes to convert the AIFF files to AIFF. Then SampleManager was happy.

 

SampleManager lets you do many many things to your files, but the only things we are interested in are; repairing any dud files (I don't have any yet but storing in AIFF is more risky than in ALAC or FLAC); dithering to 24 bit; resampling to 96kHz.

 

You can select all your files and set off a batch to do any one of these or combination of them. For me I did the 24 bit change (use the MBIT+ option) in one hit, and then did the resampling to 96kHz (use the iZotope option with highest quality setting) in a second hit.

 

I would strongly recommend you keep your native files (16 bit 44.1kHz) as a backup. One day someone is going to tell you there is a much better piece of software for upsampling and you will need to go back to your native files to be able to get any benefit out of it.

 

If you are doing all this on a PC then try a free download trial of Adobe Audition 3. Its resampling sounds pretty good. But to own it costs more like USD300.

[Antipodes1553552706]

Stating the obvious - if you started with FLAC files, allow for needing six times the storage space for 24bit/96kHz AIFF...

[kaka]

 

Antipodes;97258 wrote:

....

 

SampleManager lets you do many many things to your files, but the only things we are interested in are; repairing any dud files (I don't have any yet but storing in AIFF is more risky than in ALAC or FLAC); dithering to 24 bit; resampling to 96kHz.

 

...

 

Did you compare the SampleManager 96 files via Foobar, with those using the Foobar SRC upsampler ?

 

Filesizes 6x bigger plus SSD isn't a wondrous recipe !

[coleus]

 

Antipodes;97258 wrote:

 

dBPoweramp is a bit odd. I used it to convert FLAC to AIFF, and iTunes coped with the resulting files and tags. But SampleManager reckoned the files were corrupt and wouldn't open them. I remember having a similar experience with dBPoweramp when converting to ALAC once (though not last time). The cure was to use iTunes to convert the AIFF files to AIFF. Then SampleManager was happy.

 

not sure if this is your issue or not , but if the flac has embedded art it can sometimes give an error when transcoding

[nixon76]

Interesting - over/upsample 16/44 tracks so that Amarra handles the whole thing. Never would have thought of that!

 

There are many over/upsampling dacs (often to 24/192) but as I understood it there was a move of late to go back to native sampling (Red Wine Audio etc). I'm guessing the user of the system you're building would have to be comfortable with this. Although it may simply be a case of letter the ears decide.....

 

Your findings (after experimentation and discussion with Steve Nugent) are great news! No more talking (from me) of bodging something up to get a lynx card attached to a mini or macbook. A dedicated C2D mini/aspara/offramp combo should be fine, except for the storage issue (it only has one 2.5" drive bay and no eSata interface - unless you want to start hacking again ).

 

I've a full 24/96 album in AIFF - it takes ~2.5GB. If we take a 256GB SSD and lop off 56GB for software/OS etc that leave room for ~80 albums. Whilst it reduces convenience/usability, maybe a favourites/more frequently listened to regime could be in order. i.e. keep your best 80 albums locally and listen to the rest off a NAS, moving things back and forth as the mood takes you, something a well written script could probably take care of.

[kaka]

 

nixon76;97325 wrote:

Interesting - over/upsample 16/44 tracks so that Amarra handles the whole thing. Never would have thought of that! ....

 

I saw your mention that Amarra drops back to itunes for 16/44.1

I have skimmed a clump of Amarra stuff without seeing that mentioned again -do you have a reference to it ?

 

I've seen claims that the Amarra advantage comes from their treatment of floating point, so it seemed to me that would still apply ???

[nixon76]

I'll see if I can find that reference again. Ah here it is:

 

 

 

You use iTunes like you always have, select the songs / playlists and play. AMARRA follows

 

what you do in iTunes and takes over playback whenever a full resolution uncompressed AIFF

 

or WAV file ranging from 44.1-192 kHz sample rate (16-24 bits) is played. Amarra seamlessly

 

switches to iTunes for playing back low-resolution

compressed

files such as MP3 or AAC

 

type, then back again once a high resolution file is detected.

 

I've misread it - it flips control back to iTunes for compressed files.

 

My bad .

 

So no need to over/upsample 16/44 files.

[Antipodes1553552706]

I prefer trying these things for myself, so now I have my SSD installed I will, but from talking to some who have tried this out, the SSD is only necessary for the software, and the music can sit on a conventional drive and even USB is OK for attachng it.

 

It seems that there is also agreement that upsampling is a good thing if done well and that it is best done offline. This is to say to resample the files and then play them as is with no further resampling. This is believed by many I have corresponded with to be better than using an upsampling DAC or using SRC on the fly such as in Foobar or playing 'native'. I haven't tried all these permutations but from how I understand it, and from my experinces to date, it makes perfect sense. I have been playing with a few of the SampleManager permutations to see which is best and am now reprocessing all my files with what I think is the best solution.

 

I think the hardest thing for people in all of this is the need to dedicate a computer to just streaming digital audio. If you start using the computer for anything else then you start to notice background noise rising.

[Antipodes1553552706]

I should add that I have extensively auditioned SampleManager offline resampling against native and there is no doubt in my mind that provided you use good algorithms that resampling sounds better than native. Use poor ones and it will sound worse. When people refer to native they are mainly talking about removing upsampling in the realtime process. What I am saying is I agree that you should not do it on the fly, if it can be avoided. But I definitely favour offline upsampling.

[nixon76]

Is it okay to let us know what settings you arrive at with SoundManager?

 

Good to know about the SSD usage. In my mini scenario, a little 64GB SSD might be fine for the OS/software and be quite reasonable price wise. Then an attached firewire drive for music storage. This should present an even smaller load to the CPU than with USB attached storage.

[Antipodes1553552706]

Sounds ideal - I found a 32GB SSD more than enough. Will keep trying with a couple of other ideas about SampleManager and its competitors. More than happy to share. I don't see any business opportunity in this stuff. Its just interesting to talk about it and share the discoveries.

 

The business idea I am working on at the moment is a preamp (with DSP, xovers , six DACs etc, all remote controlled) to partner with a three-way open baffle speaker. Bring your own amps (six channels required). Since giving up my day job I have been able to make lots of progress with that, at the same time as doing this computer audio stuff. But that is all I will say here as I don't want to be accused of using this forum to promote my products, and I won't be sharing how I am doing that one. That and some USB cables.

[nixon76]

 

Antipodes;97364 wrote:

 

The business idea I am working on at the moment is a preamp (with DSP, xovers , six DACs etc, all remote controlled) to partner with a three-way open baffle speaker. Bring your own amps (six channels required). Since giving up my day job I have been able to make lots of progress with that, at the same time as doing this computer audio stuff. But that is all I will say here as I don't want to be accused of using this forum to promote my products, and I won't be sharing how I am doing that one. That and some USB cables.

 

A similar approach to Emerald Physics (http://www.emeraldphysics.com/Speakers/CS1.html). They use the DCX2496 with pre-loaded settings. Supposedly a Dolby Lake is excellent too but costs a multiple of the speakers :eek:. I understand from some of the forum postings* on the net that they are working on an all-in-one DAC/DSP/Xover/MusicServer thingy to skip unnecessary ADC/DAC conversions.

 

Best of luck! :cool:

 

* Or was it an email from Reference Audio when I enquired about it :confused:

[kaka]

 

Antipodes;97352 wrote:

I prefer trying these things for myself, so now I have my SSD installed I will, but from talking to some who have tried this out, the SSD is only necessary for the software, and the music can sit on a conventional drive and even USB is OK for attachng it.

 

Did you notice that in the US symposium commentary they claimed a loss of SQ when the NAS was used ? Its in section 4 of 5

[Antipodes1553552706]

I did read that but thought it might have been because of the network activity which introduces a lot of added services. In theory the music is read out of RAM and so as long as the reading into RAM is simple the disk latency need not be that great. But assumptions based on deductive logic are so often wrong in audio and experimentation with listening is the only way to know.

[Antipodes1553552706]

I had an interesting afternoon yesterday comparing my Hackintosh with a Macbook Pro and a Linux box. The Hackintosh came out on top. The comparison with the Macbook Pro was interesting. The sound was very similar but the Hackintosh had blacker backgrounds, more fine detail and better depth. I guess this implies that despite the quality of the Mac audio drivers there is benefit in the 'no moving parts' approach. The Linux comparison was maybe unfair as it was just a Linux box playing files and hadn't been optimised for audio playback in any way. I suspect getting player, drivers etc sorted for a great Linux player might be a significant endeavour. The sound was allright but not close to being competitive. I think I will stay with the Hackintosh for the moment. I went out and bought a copy of Leopard last week to ease my conscience.

[nixon76]

 

kaka;97386 wrote:

Did you notice that in the US symposium commentary they claimed a loss of SQ when the NAS was used ? Its in section 4 of 5

 

NAS and locally attached USB/Firewire/eSata are different things/protocols. If we're talking Amarra compatibility and a NAS, then it needs to be one that can provide a 'mounted volume' verses a shared network drive. Not all NAS's offer this (on a Mac).

 

Steve N is going for the eSata interface (as per his TuneBank product)

 

 

 

The Tune Bank interfaces to your Laptop using an included SATA II Card or a SATA II PCI-E 1 card (PCI or PCI-Extended) for Desktops. It is important not to use USB for this storage interface.

 

This (use of eSata) requires a Mac with a Express/34 slot - as mentioned in my earlier posts, this means older style (15/17") MacBook Pros, or a new MacBook Pro 17", or a big tower of course.

 

But if it came down to Firewire or USB the Firewire should present a lower load CPU wise.

[Antipodes1553552706]

To recap, all in one place. Remember, these are only my findings and I am not claiming that the process was scientific enough to be definitive.

 

1, Computer based audio, when streaming a digital signal from lossless files, can exceed the performance of expensive CD Transports.

2. The most promising method is to use an asynchronous feed so that the data can be buffered and clocked outside of the computer, and to my knowledge the options to do that are fairly limited today - Squeezebox and Sonos use ethernet, and there are asynchronous USB solutions from Empirical Audio and Wavelength. it looks like there will be growth in this area with a couple of other products in the pipeline. However, in my experimentation I found that using the Empirical Audio Offramp (not an asynchronous solution) gave me better sound than using the Squeezebox SB3 or Transporter as transport (that is, to provide an SPDIF or AES/EBU digital feed to my DAC). Sometimes implementation is everything.

3. If the feed from the computer is not asynchronous then the cable used to deliver the bits from the computer is important. A USB cable used to do this needs to be well designed and made. Most USB cables are made for use in block mode transporting of bits, and that is not a very demanding task. To deliver bits in realtime together with the clock information from the computer is a very demanding task. Some devices you connect to the computer will cope with a little jitter better than others.

4. It is possible to use a PCI card for conversion to SPDIF or AES/EBU, but there are only a couple of cards worth looking at for that, they are expensive, and it does require a lot of attention to reducing electrical noise in the PC - which creates compromises. For example, reducing electrical noise in the PC can mean under clocking the PC and minimising RAM, but more speed and more RAM can be very beneficial for audio streaming.

5. It appears to me that the Mac audio stack and drivers are superior to Windows. You can get an excellent result out of a dedicated PC using Windows, but you have to attend to a lot of detail in order to do that. Particularly, turning off a lot of services and tasks that will distract the PC from the task. XP is relatively easy to make sound good, Vista is harder. Windows 7 appears to have a better audio stack than either of them, but like Vista is still burdened by a large number of counter productive services. For all of them there is value in using things like ASIO or WASAPI to bypass some of the windows audio stack.

6. Out of the box, the Mac with iTunes is almost as good as the best I could get out of a PC. Doing some of the tweaks to the Mac as I had done with PCs took the Mac ahead of the PCs but to achieve that I had to build my own machine using PC parts and use a hacked version of Leopard. There may be other ways of getting good results by working at the Unix level but I have not tried that yet.

6. I have not yet tried it, but the reports are that adding Amarra software to iTunes on a standard Macbook Pro is outstanding. The downside of course is that Amarra software costs USD1500.

7. Upsampling can be very beneficial. First, the conversion from digital to analog inherently makes a decision of some kind about the waveform that links the samples. If that waveform is just a straight line between the sample points then this does not sound as good as some smoothing of the signal. Applying some form of shape to the curve between the sample points is upsampling, as it is creating new data between the sample points. The real questions are about what the best algorithms are, and where in the audio chain to do it. There are a few good algorithms out there and some poor ones, and some sound worse than no upsampling (straight lines between sample points), and none of them can be perfect. But it is clear to me that offline upsampling is best. Doing the upsampling in the realtime audio chain is not a good idea as it creates unnecessary processing while attempting to perform the main task of streaming the bits. Better to do it once for all of your files and store the upsampled file. This will increase storage requirements. Similarly, WAV and AIFF will sound better than FLAC and Apple Lossless as playing the former requires less processing during streaming. The one caveat to this is that if you are using asynchronous transfer from your computer then the computer can do unpacking of compressed files and upsampling on the fly without affecting the quality of the audio streaming. I found that I liked SampleManager to batch process the upsampling to 24 bit 96kHz on the Mac.

[nixon76]

Great stuff - thanks!

 

 

Antipodes;97419 wrote:

there are asynchronous USB solutions from Empirical Audio .......... the Empirical Audio Offramp (not an asynchronous solution)

 

Is it that Steve hasn't released an async model?

[Antipodes1553552706]

The Pace Car and his Spoiler DAC work asynchronous, and the PC or Mac that you use needs modification (I think compliant USB) and you need an interface than can be slaved to an external clock, so it is more complicated. Steve's site explains it. It is a little frustrating to think you have to send your PC or Mac to the US and maybe buy an EMU box and send that to the US for modification too before using the Pace Car. The Offramp is a very clever device such that with a fairly good USB cable and attention to the PC or Mac, the performance is comparable to his asynchronous options, and I have preferred to go down that route because it is easier.

 

I should add that with the Mac, the downsides to the sound are a little forwardness and glare that really benefit from upsampling and expanding the word length using SampleManager. The PC has different sound characteristics, being a little woollier sounding but gentler on the ear, and so may sound better with different DACs or different resampling than work well with the Mac.

[Antipodes1553552706]

One little interesting thing - or at least interesting to me - but illustrates my point about pre-processing. I have a quite a few files that were initially encoded at 320kb/s mp3. There was clearly a gap between these and the FLAC files, and in my system you didn't really want to sit down in front of the stereo to listen to them properly. As part of the experimentation I converted these to AIFF and then processed them to 24 bit and 96kHz. They sound hugely better and are enjoyable to listen to, though not so good with more busy music, as one might expect. The improvements are in smoothness, naturalness and spaciousness. The improvement was much greater than the improvement to the FLAC files.

[Antipodes1553552706]

Nixon76, is there any thinking on using PCI eSata versus eSata off the Mobo Sata?

[nixon76]

If we're talking about using a MacBook Pro, then PCI eSata (via the Express/34 slot) is the only way to get 'extra' storage (other than the normal FireWire/USB/NAS route), beyond the 2.5" internal HDD.

 

As for the benefits of mobo e/sata vs PCI eSata in terms of CPU load (or other areas etc) I don't know - it's beyond my knowledge.

 

If we hark back to my wild eyed idea of using the ExpressCard/34->PCIe->Lynx, would this allow mitigation of electrical noise issues, as it's isolated (except for the tether) from the laptops innards? Perhaps with a wrapping of ERS paper to boot?

[kaka]

On a slightly different tack, is much known about the audio pluses expected from the mac Snow Leopard OS that is coming ?

[Ernie1553552694]

Hard to say until it's out there in the wilderness...

[Antipodes1553552706]

 

nixon76;97444 wrote:

If we hark back to my wild eyed idea of using the ExpressCard/34->PCIe->Lynx, would this allow mitigation of electrical noise issues, as it's isolated (except for the tether) from the laptops innards? Perhaps with a wrapping of ERS paper to boot?

 

I would have thought so, any process you take out of the box seems to be a good thing.