Building A Basic Music Server: A Beginner's Guide/Commentary - Part 1
Decision 7: Which Type of Drive?
Hard drives have their own complexity, but the choices offered may make a difference to you. Let’s break this down.
Capacity: first, you need to decide how big a hard drive you need. With external storage, this isn’t a life or death decision because you can always add more drives later (actually that’s true with internal storage too, but you still have to consider whether you want music data and the OS on the same drive).
We’ll get into music storage formats later, but for now you should know that these formats affect how much capacity you need. To be conservative, if we assume everything will be stored as uncompessed (e.g. .wav) files and if we assume the average CD has 400MB of data on it, you need about 500GB per 1000 CDs (or 10,000 songs). That amount will give you a little extra capacity. 1000 CDs is a lot, but not uncommon for avid music collectors.
I was surprised when I counted the CDs in the lab and found the count at 1600. Not all of those are keepers, but then again I’d like some headroom. Since 1TB is not much more expensive than 500GB, we’ll use that as a base capacity for the system we’re building.
Interface: you’ll need to determine which interface to use for the hard drive. On a Mac Mini (or iMac) this means choosing between Firewire and USB. USB is the slightly less expensive way to go. Firewire is slightly higher performance (Firewire 800 has higher transfer rates than USB 2.0 and this is rendered even more relevant in practice because Firewire is not CPU dependent as USB is).
You will want to consider at this point which DAC interface you plan to use. Firewire is also a better DAC port to use for the reasons stated above, but there are far fewer Firewire DACs than USB DACs. There is debate about this, but my feeling at this point is that I would ideally (cost no object) get the DAC interface right and live with the lesser drive interface (one could put DAC and drive on the same interface too, but that seems risky).
That said, I plan to use the Firewire port for the hard drive on our project server. Why? The key reason is that we will be testing a lot of USB DACs. I also think that a focus on USB parallels the current focus of users on USB DACs.
Backup Scheme: there are a lot of choices here. First, given the effort to rip a big library, combined with the cost of losing downloads, most users will want a back-up scheme. In other words, you could have the backup scheme called “none” but I can’t recommend it.
One basic backup approach is to have a RAID drive as your external drive. This will be RAID level 1, which means you will have two drives in one cabinet and each drive will automatically be written to identically mirror the other. If one drive fails, you have the second drive. To me that is comforting, and RAID raises the price of 1Tb of capacity by only about $100 so the cost isn’t horrendous (but it’s your money, so don’t let me spend it). Note: remember that you’ll have to buy a 2TB RAID drive to get 1TB of effective capacity.
Drives can fail, so you might want to backup your music to a NAS (Network Attached Storage) drive. This is just another external disc with an interface to your network router rather than directly attached to your PC/Mac. Even RAID drives can fail, so NAS backup can be a good idea either way.
Quietness: if your external storage is going to be in a listening area, you will want a drive that is quiet. This can be hard to assess, but there are fan-less drives. You could also address this by putting the server in an enclosure.
Speed: Presumably lower seek times and faster transfer rates are advantageous. Since the transfer rate of a typical hard drive exceeds the data rate needed for music by a factor of 50-100, at this stage I’m willing to assume it is a minor factor. Seek time could matter and we may get to roughly test this if we can use an SSD (solid-state drive) in another project. Again, I’m not sure if the small seek time differences among traditional hard drives matter, nor do I know if quoted specs are a reliable indicator of actual speed.
In the end, given these criteria, I chose a LaCie 2Big Quadra 2TB RAID drive. This gives us 1TB effective capacity. It has Firewire 800 ports (as well as USB, eSATA and Firewire 400). The LaCie has hot-swappable drives which seems convenient. And it is partially passively cooled which I hope makes it quiet enough to have in the listening space.
The LaCie costs $295 (street), so to this point I’ve spent $1143 of my allotment (we could add $100 if you need a monitor and subtract the $58 I spent on needlessly fancy hard drive and processor, for a total of $1185 all-in).
All content, design, and layout are Copyright © 1999 - 2011 NextScreen. All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Comments
Man, pretty good so far. I have used macs for years but am not a techie. I have high end stereo gear and am aiming towards a mac based server. I cannot understand much on Audiogon, Audioasylum, Computer audio, Weiss homepage, etc. but this is much better, more basic, step by step.
Couple of ??'s:
So, OS on Mini, use Mini superdrive to rip, I want to use bit for bit, non compressed (Flac? Wav? ....?). When I rip cd's, where do I direct them to? The Mini's HD? the external drive? I thought the back up was the mirrored Lacie 2 TB; what is this NAS thing and where does it fit in, how is it attached?
Role of Drobos? (more simplistic RAID for us simpletons?)
So when the music plays back into the dac via firewire, it is playing from the Lacie drive (in your example)?
Do you need a keyboard/monitor at all or can everything regarding the ripping be controlled from another Mac (say, a Macbook using Apple remote desktop; and if so, how?)
Some sites state with the Mini being used as a dedicated server, it should be configured and certain inherent programs (bluetooth and others) be trashed before beginning to rip cd's. Correct? & if so, what should be gotten rid of?
How should one prepare cd's prior to ripping? (professional cleaners? demagnatize?
I am sure you will get to Amarra later; lots of ???'s there.
Lots of scattered queries; they reflect my ignorance of the whole process but I encourage you to continue as this has been enlightening thus far.
I plan to go into configuration of software in the next phase, so I hope you can wait.
For now, let me clarify my plan (I've also edited the main text in hopes of making this clearer). The primary copy of each music file resides on the external hard disk in this setup. So, I'm playing files from the external drive, which is sitting inches away from the Mac Mini and is connected via Firewire. The operating system is on the local hard disk. My reasoning is to have the primary music files on a separate box that can be easily transferred to another computer.
I chose to have the external drive be RAID. That means the first line of backup is on the external drive sitting next to the Mini. RAID simply doesn't raise the price enough in my mind to make it rational to skip it. But as I said, don't let me spend your money.
NAS is network attached storage. It is just a hard drive with an interface that allows it to send files over a network. In this case, the NAS drive sits next to our network router and is plugged in there. It is connected to the Mini via Ethernet. If you're following, the RAID is primary backup, the NAS is a second backup. Who you callin' paranoid? Actually, the NAS is entirely optional in my view, but I like the ability to access the NAS from other computers.
You could run music files from the NAS, but I don't like the idea of possible lags in that system. 100 Base T has a theoretical throughput lower than Firewire 800, but it also has more resource competition (other stuff on the network, and routers and such to go through). Apogee and others with more experience than I have caution against network drives as primary storage. Others who've tried this say it works fine and your network may be Gigabit Ethernet, but latency is a concern more than data rate and, well, the external disc is pretty cheap.
AVGuide webmaster and general drudge
Very well done. I'm very interested in building a home music server, but many of the articles are just hard to understand and to get concrete, actionable advice from. This article may be the best I've read so far. Keep it coming.
I've been using a NAS for sometime hooked up to my wireless router means I can access my music from any device in the house. Freenas is ummm FREE and works well. I've also been looking at Windows Home Server but is not as resilient as the options on Freenas. I suggest you look at this for building a server it won't cost $2K! more like $300 ! as you make it out of practically any old PC and has software raid so I have 3 x 1TB making 2.2TB of storage.
Keep the blog coming and send me a mail when the next installment is ready.
I would like to say that the silly iPhone tuned me into a MAC head. AppleTV can easily stream FLAC or WAV (and the dreaded MP3) to your sound system if it has wireless (or wired) capability.
I never touch CD's anymore and I let my controller's DAC do the converting.
AppleTV becomes even more flexible when you add Fire Core's ATV Flash to it!
I have been an audiophile for over 40 years and I am so grateful that I can stream music, see album art and select what ever I want to hear from AppleTV.
Cheers!
Larry
Hi Larry
I dont know if u ever read this .
Please advise about how can apple tv play FLAC as apple does not support FLAC .
Secondly I don't understand this part of your comment( my controller's DAC do the converting).
Pls advise
Cheers,
Mohammad
Going back to the option of using my existing PC, I am using iTunes to manage my files and use the digital audio output to feed this to my Anthem D2v processor digital audio input. The problem is that the music quality decreases substantially. Would using a ASUS Sonar Essence STX card correct the sound, or am I a long way off?
Thanks,
John.
Good article so far. I have a Mac mini as my music server. Files are stored on an external drive (I went with 1TB Glyph - a costlier option but I've had 3 cheap externals die within one year - 2 WDs and 1 GTech - Glyphs are used by pro musicians who want to ensure their precious tracks are safe). Mac outputs via Toslink to a Bryston DAC (awesome if you can afford it). Only issue with the mini thus far is its inability to handle 88.2 kHz - perhaps due to deliberate software limitations..this is not the case with an iMac - and anything over 96 kHz. The later is much of a problem since there's little content with that high a sample rate and I suspect the difference is imperceptable. The 88.2 limit is a bit frustrating since 88.2 sources wind up being upsampled by iTunes to 96. Other hassle is having to manually set sample rate and bit depth via the midi tool everytime they change. Also requires an iTunes restart. At any rate, glad to see to letting people know there is a much better, less costly way to go digital than buying overpriced dedicated music servers.
Nice article. I'd like to emphasize and expand on your first point - timing of the digital data (i.e., the issue of jitter, principally "pre-echo"). I read somewhere in the last six months a great explanation of this critical issue. Here goes my summary: our neuro-physical perceptions and processing of sound is quite accustomed to "echo" - that is, a form of trailing edge to a particular sound. What is quite unnatural to us is an "echo" that precedes the particular sound and it is this element of jitter (basically, the mis-timing of the data containing the sound we intend to hear) that has been the primary bugaboo of audiophiles who claim that digital music (CD's principally) is "hard" or "edgy" sounding when compared with analogue (principally vinyl). Take away the "pre-echo" and, with adequate bit/sample rates (all other things being equal), you restore the natural sound characteristics that we are accustomed to hearing in nature. So, it becomes critical to determine those components of a hi-def digital sound system that can contribute to jitter and evaluate what they do to eliminate it - especially "pre-echo". In this regard, I'd appreciate it if you would devote one of your upcoming installments to this topic and how your systems are addressing it.
Secondly, I think that solid state storage devices are important and becoming a much more relevant topic as the cost associated with these devices is coming down quickly and their capacities are increasing. They are quiet, stable and (seems to me) a perfect medium for music server storage. I'd appreciate your thoughts on this topic as well.
Thanks again for a very nice article and I hope that this is the beginning of regular, ongoing committment by the major audiophile pubs to cover this most significant form of music playback - IMHO, THE future for our hobby.
Kind regards and happy listening!
I think a separate piece on SSD may be warranted. Right now, SSD is a little pricey but that will change. There is a secondary issue, which I cannot quantify, which is that SSDs are limited to the number of read/write cycles they can do before data errors occur. No idea if this is a non-issue (read errors will occur, but just before the sun goes dark) or something to actually be concerned about.
BTW, this article and others does indeed represent an ongoing commitment to cover computer audio. Playback as a brand now focuses on computer audio, headphones/earphones, and desktop audio.
AVGuide webmaster and general drudge
When will your part-2 be published? Couldn't wait ;o(.
I have used an old 2nd hand Dell laptop from eBay and two external drives (one strictly for backup) for storage as a dedicated 2 channel (stereo) music server for 4 or 5 years. The music from the computer's usb 1.1! port goes to the usb input on a Pioneer Elite receiver, which converts it to PCM and re-clocks it, so jitter should be as good as the Pioneer's clock, which I would argue is essentially as good as that of most fancy usb DAC's. All files are stored as lossless. Total cost (for the computer and hard drives) was about $500, and the quality is - to my reasonably educated professional audio ears - at least as good as directly from the CD's from which I ripped the music, and I have complete data redundancy. I don't see how spending any more could possibly increase the quality significantly given the limitations of redbook CD's.
After trying virtually all the software music players, I finally settled on iTunes operated by wi-fi remote from an iPod touch which has an elegant user interface, album art, and instantaneous control without the kluge and inconvenience of a big sound reflecting touchscreen, wires, or computer keyboard near my listening position. I will be most surprised if a more convenient, better sound quality or more data secure server can be had regardless of cost, although some people will no doubt prefer the big and fancy graphics that can be had with some commercial servers.
I do listen to a fair amount of DSD SA-CD and DVD-Audio direct from optical discs, and the best of those do sound better, especially with multi-channel sound, but that material isn't really a viable option for storage and playback from a simple music server. The recent availability of downloadable 96 kHz 24 bit audio 2 channel high resolution audio files may finally force me to hook up an old firewire professional sound card that's been hanging around, as the usb 1.1 bus can't handle those, but I don't see any point, other than bragging rights, to upgrading anything else .
Very pleased to see this article and I will follow the series with interest. My main concern will be the critical audio path (yes I know that all audiophiles are critical, but I am referring to the 'critical path' whereby the audio quality is determined.)
I am quite interested in Hewlett Packard's MediaSmart series of home media servers running Microsoft's Home Media Server software. This is a readily available, mostly pre-configured system which could be a good basis for the server function, though it doesn't have a Firewire port for a DAC, but does have USB Ports.
I note that PS Audio is working on an interface bridge for their PerfectWave DAC which should only require an ethernet connection.
I look forward to the next installment!!!
I use HP's Media Smart Server as a music (mostly high quality mp3s encoded with LAME) and movie (uncompressed DVD rips) server. I have 3TBs of storage in it, and it allows you to pick which folders you'd like to back up across the drives. Plus, it does automatic back ups for the PCs on your network (up to some limit). I've been streaming to my laptop which I sometimes hook into my A/V receiver, but I'm planning to build a dedicated home theater PC soon. So far it has worked well for me. The only streaming glitches seem to come from my laptop's wireless; if I connect to my network via wire, it works very well. On the client side, I'm using iTunes and MyMovies (MyMovies also has corresponding software for the server side).
From reading several of these comments, I can see that we need some sort of soundcard/DAC series, including interface issues (S/PDIF, Toslink, USB etc). Thanks for the input, which I will relay to Chris.
AVGuide webmaster and general drudge
"The key reason is that we will be testing a lot of USB DACs. I also think that a focus on USB parallels the current focus of users on USB DACs."
I'll add a vote for more focus on different interfaces. I know I'm not the only one that already owns good quality DACs integrated into an expensive preamp & wondering if toslink from a Mac to it will put me at a sonic disadvantage compared to a USB or Firewire DAC.
Great article... eagerly awaiting the next installment!
I am hoping to share some of my insight in to digital transmission and music servers. I've worked in data networking for 25 years, and have been building music servers for 10 (including having an MP3 player in my car for 10 years)!
Timing and timing accuracy is clearly an area that confuses most audiophiles. There are three areas where timing precision is critical, during the A to D phase, during transmission and during the D to A phase. The former, you can only hope the recording company got right. The latter, is the area where most audiophiles need to apply thought - or should I say, the companies designing the products we use, need to apply thought. The middle one, is interesting - but it shouldn't be for computer based audio. Why, because we aren't using real time interaction and synchronous transmission. I would disagree with Gadgetman's comments that throughput is less important than latency. For real time interactive communications, latency is critical - because we need to interact. For reproduced music, we are not interacting with the source (CD player, disk, PC), we are listening to the music (interacting with it as an art form - but not from a technical pespective), and if it starts 1 second after we press play or 10 seconds - that's latency - who cares? Now if the music moves from 1 second latency to 1.5 second latency and then to .5 of second latency randomly during the same track - we do care, we care a lot - but that's jitter, not latency - more on fixing jitter later.
So transmission over Ethernet or worse, Wi-Fi. Ethernet runs about 40-50% efficiency on a congested network - which most home users will never have. That means it's only about 50 Mbps for a 100 Mbps Ethernet network. However CD quality, uncompressed music is about 2 Mbps - thus we have plenty of overhead. The problem with Ethernet is that its shared data transmission subsystem and cares nothing for jitter. To fix jitter, and it is EASY to fix, is to buffer the signal - buffering basically means "I read the data in when it arrives and store it. I process the data when I need it" the two are ostensibly unrelated, and buffering creates that buffer (sic). As long as my buffer is never empty, the precision of the timing of the data stream is now subject to the clock I use to read the data out of the buffer. So a NAS is fabulous (and cheap) music server (library) - but it is not a replay system (reader). The replay system needs to understand enough of the communications infrastructure to put enough buffering in place to deal with the jitter. If you've used Youtube you've seen the buffering in action (the slow start) and probably failing due to the massive variability in throughput and jitter of the Internet. And this is key - its the music replay application that needs to buffer the music, and it's independent of the source, so whether you use internal HDDs, external HDDs or NAS, or whatever, it simply doesn't matter as long as it's throughput is much faster than the replay speed, or your buffer is big enough (and you wait for it to fill) - and thus the application can grab the data when it needs it. What is more important is that the application can accurately time the D to A stage. If you REALLY cared about that, you'd use a RTOS (real-time operating system), which would guarantee that the replay application could get the access to the CPU when it needs it.
SSDs are a waste of money for a music server - the speed is irrelevant, the seek time is irrelevant, as the music application is the critical piece. And as many users have discovered, the application can mess up so many other more important things - like transcoding between bit rates without you knowing, poor quality digital domain volume controls etc.
So far so simple - where this all falls over again is when we discover that PCs (and I assume MACs as well) are awful at D to A - and instead we want an external DAC. And we have to consider the subject of clocking all over again. But now for different reasons.
Last comment, WAV is disk space wasted, use FLAC, or WMA or Apple's lossless compression - just be aware of where you can't use those compression algorithms due to the proprietary nature of two of them.
When's the next installment?