Jump to content

Building the ideal(ish) Music Server


Recommended Posts

Following on from our escapades with Mac Mini, Windows laptop, Rasperry and Banana Pi, I was still looking for something better to match the output from my Accuphase DP-700 feeding the DC-37 DAC via HS-Link ( Accuphase i2s style implementation). The Pi machines actually came close in terms of noise levels and tonality but there was still something missing in terms of resolution, imaging etc so I thought I would do some research.

b9120da10afd184f968ca19facf0325d.jpg

Theory

I first decided to hitch a ride with those that have done the research and then put their findings out in the market place for the world to examine. These are the commercial manufacturers of music servers and they all had the following common ground:

1. They all use modest CPU's ( e.g. Intel Atom) to minimise EMI/RFI. EMI generation is linked to power dissipation amongst other things

2. The "better" models used linear power supplies

3. Antipodes and others use Sotm or equivalent PCI/PCie USB output card. PCI/PCie cards have a degree of isolation from the main board and the potentially noisy mainboard USB bus. They also reclock the signal to be sent to the DAC and have on board filtration, but can be powered directly by an external clean linear power feed

4. SSD seems to be the pick for better SQ

5. Antipodes also claim there are different types of memory

6. They mostly use light weight, headless Linux O/S with no GUI, specifically tailored to the application

1268a933697b6ef3401b682bedf88e18.jpg

There are also those that claim the most powerful CPU is the way to go. But since I did not need extra capability for DSP , a GUI or a heavy operating system, I decided to first build a server based on the above formula.

Findings

Each of the following may be a subtle point of difference but combined makes a significant difference

1. Motherboard- I have tried a couple- of different makes - Jetway and intel using Intel atom 2800 CPU. Surprisingly, there were differences using different Linux players with each tending to sound better with different distro's and players- (subtle differences however)

2. The motherboards have hard wired CPU with 12V inputs which made things easy. An HD PLEX linear power supply outperformed a standard laptop linear PSU in terms of tonality , depth and perceived dynamics, (at much greater cost however). With Linear PSU, tonality was a few steps closer to the Accuphase player reference

3. Testing from motherboard USB vs SOTm output produced interesting results. The Sotm card improved things further in a number of ways. Im not sure if t was the lower noise, reclocking or a combination of those things.

4. On my system SSD sounded better although it was harder for me to pick differences when using a 2.5"spinning HDD with Rock music. I also tried SATA power filter and thought they took away more than they gave back.

5. Memory. This is one that I did not expect. Mark Jenkins did not specify which memory sounded better but since he also uses lower power boards I figured it could have something to do with single rank vs dual rank memory. I understand that single rank memory (1R) is configured to allow the CPU to access all the memory chips on the board at once ( even if spread over two sides) whereas with reguar or dual rank memory, only once side of the memory stick can be accessed at a time by the CPU.

I ordered 1Rx8 micron memory sticks and to my surprise they did sound better than the memory I bought from the local computer parts store. The difference was not as subtle as you would think.

Software

A friend who is also a Linux guru custom built a real time Linux kernel for me with a number of players, including MpD and Jriver. I also had downloaded configurations for Vortexbox and Daphile. They are all bit perfect players but can sound different nevertheless. I can say that the "best" one is dependant on individual taste as well as the the rest of the system. If you like things crisp and accurate and perhaps with a bit more edge than others, then Daphile may suit. If you like a more expansive, large scale sound with very good tonality, Vortexbox may suit and if you like fast, crisp clean sound with rock solid imaging, then Mpd could be the one.

Result

I am still running a couple of variations of software but I can say that the performance level is such that I can now happily listen to CA as a alternative to the CD player through the same DAC

Edited by Tasso
  • Like 19
Link to post
Share on other sites
  • Replies 3.2k
  • Created
  • Last Reply

Top Posters In This Topic

Top Posters In This Topic

Popular Posts

Following on from our escapades with Mac Mini, Windows laptop, Rasperry and Banana Pi, I was still looking for something better to match the output from my Accuphase DP-700 feeding the DC-37 DAC v

I’ve read this thread with much interest, and although I’m severely technically challenged when it comes to all computer and technology related matters, I embarked on a CA server project.  This wa

My new build server is progressing well today. An all in one unit as concept. All credit to @Doncentric with what it was a prelim-concept meeting, and within hours, he had 85% of its done! The remai

I  bought mine on ebay but this is the memory - I used 2 sticks 

 

http://www.amazon.com/Micron-2GB-1RX8-PC3-12800S-11-11-B2/dp/B00DU1BS7A

 

For motherboard, best to decider if you will use a Sotm or equivalent card first. If so you will need a board with a PCIe slot - My jetway board is PCI only and Sotm no longer supply PCI cards although they might crop up on Ebay   The 2800 Atom has a total power dissipation of 7.5 watts and I think you can choose a number of boards which have similar specs. A quad core  Celeron 1900 for example  looks interesting with only 8.3 watts TPD.

 

but these are the boards I used so far.

 

http://www.mini-box.com.au/Intel%20DN2800MT%20Mini-ITX%20Motherboard.html

 

http://www.mini-box.com.au/Jetway%20JNF9C-2600%20Mini-ITX%20Motherboard.html

Link to post
Share on other sites

How can one establish an Antipodes clone business when all are transparently published?!? :D

Edited by Chanh
Link to post
Share on other sites


Looks good Tasso, i have been running a similar setup for a few years now.

I also use the hdplex lps and sotm pcie audio card.

Have you tried windows server in core mode without the gui?

Would be interested to hear your opinion :thumb:

cheers

Liam

Link to post
Share on other sites

I think I have been very lucky. I vaguely realised most of the points Tasso made in the opening post and my local computer store just happened to have an Intel Atom DN2800MT mounted  in a double height case . No fan and internal SSD. I don't have a separate USB card but  I'm using a powered USB hub (powered by the 5V output of a HDPLEX LPS). The same LPS powers the MB from its 12V output.

 

I'm running Windows Server 2012 and JRiver.

 

Interesting about the memory. I have  a 4GB DDR3 Kingston memory stick. I might get my tech to replace this with the 1Rx8 micron memory sticks.

 

A couple of guys who are anti computer audio have listened to my system and have said it was the best they heard from computer audio.

 

I'm yet to try JPLAY6 which according to folks who have done this takes SQ to a whole new level.

Link to post
Share on other sites

When people talk about a server, I always imagine a NAS where the music is stored, or a DLNA server which streams content into TV's, radios, and the likes. I'm guessing this is more a music dedicated HTPC build, or I have to get used to the jargon. I do have a dedicated server where I store and stream data, but it's hidden in the bowels of my office room.

 

I just received a 2010 Dell laptop, i5 with a few gigs RAM on it. I'll change the HDD into an SSD and install Xubuntu 14 with all the goodies, MPD, Kodi (XBMC), but I'll run it for the time being through the integrated audio outputs (please don't throw anything at me!). I'll upgrade later, say Xmas time to the Marantz HD-DAC using the USB output.

 

I'm always curious about the use of a linear PSU as opposed to the standard SMPS. I might be barking up the wrong tree here, but don't PC's have switched power supplies internally anyways? Somehow they must generate the 3.3V, 5V and 12V for distribution.

 

All in all, a dedicated PC in that case looks heaps better, I'll have to hide the laptop somewhere. Luckily there are ways to control it with the phones or remote keyboards.

 

Cheers,

 

Alberto

Link to post
Share on other sites


There are switch mode type "buck" voltage step downs on various points on the motherboard on these 12V boards. But linear PSU still makes an audible improvement - it wouldn't be to the same level as say on a preamp or phono stage but an improvement nevertheless. For ITX boards, Teradak make a comprehensive linear power supply which provides different voltages to each section of the board.

Link to post
Share on other sites

Looks good Tasso, i have been running a similar setup for a few years now.

I also use the hdplex lps and sotm pcie audio card.

Have you tried windows server in core mode without the gui?

Would be interested to hear your opinion :thumb:

cheers

Liam

At all times I was conscious of building a music server, not a PC. With that in mind, windows comes with a lot of baggage which is fine for a PC, but for a music server it needs fidalizer and a 2 PC setup to get the best out of it. These Linux distros have been built without the same " baggage" for audio only.

Having said that, I am mindful of how different software can sound and it is not as predictable as I first thought , so I would like to try the windows setup to see what the final result sounds like.

Link to post
Share on other sites

I think I have been very lucky. I vaguely realised most of the points Tasso made in the opening post and my local computer store just happened to have an Intel Atom DN2800MT mounted in a double height case . No fan and internal SSD. I don't have a separate USB card but I'm using a powered USB hub (powered by the 5V output of a HDPLEX LPS). The same LPS powers the MB from its 12V output.

I'm running Windows Server 2012 and JRiver.

Interesting about the memory. I have a 4GB DDR3 Kingston memory stick. I might get my tech to replace this with the 1Rx8 micron memory sticks.

A couple of guys who are anti computer audio have listened to my system and have said it was the best they heard from computer audio.

I'm yet to try JPLAY6 which according to folks who have done this takes SQ to a whole new level.

Well done Peter. I think you would like what the SOTM Card does for the sound . HDplex has a 9V rail specifically for the SOTM PCIE card

Link to post
Share on other sites

@@Tasso, I notice you mention the Teradak power supplies, have you had experience with Teradak products and was a good experience? I am using a stripped down 2009 Mac Mini and the only linear PSU that I can find that outputs the required 18.5 volts AND comes with the appropriate cable to ensure the Isense wire is connected AND is significantly less than $1000 is the Teradak, but I cannot find anyone who has actually used one.

Link to post
Share on other sites


That said, I am sorely tempted by an Intel DN2820FYKH Celeron NUC with a 2GB SSD as a music only player ................... and it is 12 volts ............ much more choice for linear PSU ............. :)

Edited by Ancientflatulence
Link to post
Share on other sites

My experience with Teradak is limited but from those I know who have them, they seem quite satisfied. They seem to be well designed PSU's. I don't think is worth spending megabucks on PSU in this application so I wouldn't be inclined to spend more than the Teradak price for the Macmini .

No problem with NUC except that it probably doesn't have a PCIe slot. I wish I could tell you the SOTM card makes no difference , but it does add a level of refinement and naturalness to the sound.

Link to post
Share on other sites

The Atom and equivalent boards  don't need a lot. Total power consumed in full flight would be  20- 22W, possibly less with SSD.  

Link to post
Share on other sites


I have been using the SR50 shunt regulator in a number of my projects and it seems to do a good job.

http://www.ebay.com/itm/SR50-Variable-Shunt-Voltage-Regulator-5-50V-8A-DIY-Kit-for-Amplifier-/221233153412

post-131373-0-55132800-1432024517_thumb.

Cost is $30 and you can get all the other parts to build a cased unit from Jaycar.

I swapped out the supplied electrolytic caps for Nichicon Gold tune/ Muse as I had them in my parts bin.

Audio-Gd will supply you with a good quality R-core with 10v and 15v taps for $35 plus postage if you want .

  • Like 2
Link to post
Share on other sites

Thanks for the answer on the Teradak, @@Tasso, it now comes down to whether to continue with the Mac (and a new power supply) or have a go at building another machine using a Linux based player. I see the point with the NUC and the SOTM Card and I can also see that the compactness and "complete unit" concept of the NUC means that very few changes can be made to the hardware, whereas with a small mother board in a capacious case, modifications over time would be less of a challenge.

  • Like 1
Link to post
Share on other sites

whether to continue with the Mac (and a new power supply) or have a go at building another machine using a Linux based player

 

I said 'why not both!?' and ran Linux on the macmini...  but right now I am using Windows the most.

  • Like 1
Link to post
Share on other sites

I was determined to run the server headless - ie it will automatically boot into the player software without keyboard, screen etc and can be accessed by iPad. Some packages require their own HDD for convenient installation but those with software skills may find an easier way. I am not one of those so to compare how each sounded, I was swapping boot drives.

The Linux drive had several Linux players installed but Vbox and Daphile needed their own ( Daphile can be run from a USB stick to try it out )

a5655150b9aad8778b5c6fc4ff857a10.jpg

Sent from my iPhone using Tapatalk

Edited by Tasso
Link to post
Share on other sites

It is very hard to generalise about digital transport tuning .... as it is so inextricably linked to what the receiver does with respect to electrical isolation and/or jitter immunity ---- devices out there are designed very differently on these front.

  • Like 1
Link to post
Share on other sites

It is very hard to generalise about digital transport tuning .... as it is so inextricably linked to what the receiver does with respect to electrical isolation and/or jitter immunity ---- devices out there are designed very differently on these front.

 

Quite right Dave...what works for the goose will not work for the gander.

Link to post
Share on other sites
Guest myrantz

Quite right Dave...what works for the goose will not work for the gander.

Poor choice of words/idiom... Coz reading very quickly (as I always do) - you just called Tasso a goose.. :ph34r:  :ph34r:  :ph34r:  :ph34r:  :ph34r: 

Also... Shouldn't it be "What's good for the goose is good for the gander"? 

 

  • Like 1
Link to post
Share on other sites

It is very hard to generalise about digital transport tuning .... as it is so inextricably linked to what the receiver does with respect to electrical isolation and/or jitter immunity ---- devices out there are designed very differently on these front.

My decision to build the server was not for any cost saving considerations but to investigate and understand the machinations of CA beyond the the theories. I also have a couple of other hardware and software configurations I am testing at the same time. I have only reported on one.

Server streamer performance differences are often pinned to the way DACs handle the incoming signal and while this is relevant, my objective is to focus on what, if anything at all,a better server can bring. I would also be very keen to discover through actual trials, whether any specific DAC or Chipset is 100% immune to electrical noise.

The only possible way to test any hypothesis Is through experimentation which is the process I have embarked upon. Work done by John Swensen and others do prove that galvanic isolation can be breached but beyond that, it's time to test theories.

I am keen to find out more about the impact of my approach ( and that of 100% of streamer manufacturers) on different devices. So if anyone would like to be part of the experiments with their gear please let me know ( would need to be local for obvious reasons) .

Also, I would be grateful if anyone could point out what they consider to be state of the art DACs or chipsets that are immune to the issues commonly associated with CA. I will endeavor to experiment with those as well.

Link to post
Share on other sites

I have been using the SR50 shunt regulator in a number of my projects and it seems to do a good job.

Cost is $30 and you can get all the other parts to build a cased unit from Jaycar.

I swapped out the supplied electrolytic caps for Nichicon Gold tune/ Muse as I had them in my parts bin.

Audio-Gd will supply you with a good quality R-core with 10v and 15v taps for $35 plus postage if you want .

 

Thanks for the heads up Gordon. It looks like the unit will fit into  an appropriate case as well which will tidy things up. Possibly need 2 if also supplying power direct to a USB card.  Do you have any transformer in mind? I would  tend to go for extra capacity so 100-120 VA?

Link to post
Share on other sites

Isn't Shunt limited by its current demand? What have you been using it for, Gordon? Can this Shunt keeps up consistently at 2A current?

Edited by Chanh
Link to post
Share on other sites

Specs say it can do 8A which seems a bit optimistic. I've got one in a dac powering FIFO 's , works well, but very easy load. In critical circuits, shunts are much better than series regs, IMO.

The metric analog kits are well made.

Link to post
Share on other sites

Hi Mick,

Specs can say whatever it likes, especially from a sale man's point of view! Worse still when one is from Chinese source via eBay channeling. Only practical tests can truly determine its real capability. There is a distinctively differ between peak current (in nano seconds) and working current (consistently capable), and I haven't yet came across a Shunt that is consistently supplying higher than 1A in such compactness!

Btw, your FIFO current demand is minimal and in the region of milli-Ampere, a factor of 1000 less. In power a computer board, current demanding is much much higher in the region of 2000 times the magnitude of your FIFO if not exceeding more. :)

While Tasso is working on his world best CA server, I've also been researching in depth with World most affordable and light CA server. This due to our differences in pay-rate! :D

I look forward for a gtg so we could shoot-out his vs mine in a foreign setup. ;) Hope to post some pics and details soon!

  • Like 1
Link to post
Share on other sites

I have been experimenting with more components:

Single sided memory - this also outperforms the regular memory in my setup(s). These sticks are 4GB each

8a5828c975ad9eb6fabf63ec3bcebca2.jpg

A different power supply - this one by PPA studios

be1cea17928662184c4c72b66b6d2663.jpg

And a USB Pcie card from Paul Pang of PPA. Paul makes a big deal about using an TXCO clock to re-clock the signal to the DAC.

5512a025bdf989021609e3d99632ae10.jpg

I have to say that the PPA usb card is pretty special. In my setup it outperformed the SOTM card through revealing more detail and imaging. Not night and day differences however

Sent from my iPhone using Tapatalk

  • Like 4
Link to post
Share on other sites

Hi Mick,

Specs can say whatever it likes, especially from a sale man's point of view! Worse still when one is from Chinese source via eBay channeling. Only practical tests can truly determine its real capability. There is a distinctively differ between peak current (in nano seconds) and working current (consistently capable), and I haven't yet came across a Shunt that is consistently supplying higher than 1A in such compactness!

Btw, your FIFO current demand is minimal and in the region of milli-Ampere, a factor of 1000 less. In power a computer board, current demanding is much much higher in the region of 2000 times the magnitude of your FIFO if not exceeding more. :)

While Tasso is working on his world best CA server, I've also been researching in depth with World most affordable and light CA server. This due to our differences in pay-rate! :D

I look forward for a gtg so we could shoot-out his vs mine in a foreign setup. ;) Hope to post some pics and details soon!

The real current out will depend on final output transistors capability. I'll find datasheet and report back. It's quite heavily heatsinked so I expect it could power 2A easily. And despite your misgivings it is actually a capable design, I've had a good look at circuit, and implementation and it's fine .

Not all Chinese stuff is crap, they also make some good gear, I actually have one of these regs, not just surmising how it is or isn't.

There's a point of diminishing returns with power supplies, and it's easy to get carried away.

Having said that, bigger and stronger is better than enough.

  • Like 1
Link to post
Share on other sites

......

All good Mick! Great reading that we are all open minded and provide each point of view on same subject matter. Can I encourage you try powering a audioPC from it? If it works, I will get a couple for my upcoming projects!

Look forward reading your findings!

Link to post
Share on other sites

Isn't Shunt limited by its current demand? What have you been using it for, Gordon? Can this Shunt keeps up consistently at 2A current?

Hi Chanh , current use is to power a mini-dsp 2x4 board ,so no great current demand. I have used it in other projects with higher current requirement without any problems.
  • Like 1
Link to post
Share on other sites

Thanks for the heads up Gordon. It looks like the unit will fit into  an appropriate case as well which will tidy things up. Possibly need 2 if also supplying power direct to a USB card.  Do you have any transformer in mind? I would  tend to go for extra capacity so 100-120 VA?

Hi Tasso , something like this: http://www.tortech.com.au/toroidal-transformers/toroidal-transformer/toroidal-transformer-120va-240v/2*12v-5.0amp Edited by Ozcall
  • Like 1
Link to post
Share on other sites

All good Mick! Great reading that we are all open minded and provide each point of view on same subject matter. Can I encourage you try powering a audioPC from it? If it works, I will get a couple for my upcoming projects!

Look forward reading your findings!

I'm not competent enough with computers to do that (yet), but tell me what voltage you will use and I'll tell you what current it will supply.

The beauty of shunt regs is that they sink or release just the right amount of current to keep the voltage and current of the powered device exactly as it should be. But for some ears, shunt regs are "too good" ,  probably because they are so fast.

Link to post
Share on other sites

True competent shunt regulator(s) do what you specified as above. If you look closely at my DIY DDDAC, there are 47 TentLab Shunts, individually doing its duty. :) At 50mA designated, they outputting close to 70*C.

The issue with Shunt on high current demand is it runs extremely hot for inadequate heatsink. Shunt must always actively ready for any surge in current. Therefore, when the current demand is low, those energy must be dissipating as heat. For a high current demand at 12 to 19VDC (typical a PC voltage/power requirements) a shunt is still feasible but costly unless we've mistaken it from a Super-Regulator design? Also note, shunt is only as good with low impedance network, e.g as closest as possible. It will be much less optimal or perhaps deceive the purpose when place it a distance away from where it's powering. No matter how fast a shunt is, if it must travel a distance to where it needs be...!?!

Anyway, this is how I understood it and I could be wrong. Feel free to correct me here if my understanding is imprecise?!:)

Link to post
Share on other sites

True competent shunt regulator(s) do what you specified as above. If you look closely at my DIY DDDAC, there are 47 TentLab Shunts, individually doing its duty. :) At 50mA designated, they outputting close to 70*C.

The issue with Shunt on high current demand is it runs extremely hot for inadequate heatsink. Shunt must always actively ready for any surge in current. Therefore, when the current demand is low, those energy must be dissipating as heat. For a high current demand at 12 to 19VDC (typical a PC voltage/power requirements) a shunt is still feasible but costly unless we've mistaken it from a Super-Regulator design? Also note, shunt is only as good with low impedance network, e.g as closest as possible. It will be much less optimal or perhaps deceive the purpose when place it a distance away from where it's powering. No matter how fast a shunt is, if it must travel a distance to where it needs be...!?!

Anyway, this is how I understood it and I could be wrong. Feel free to correct me here if my understanding is imprecise?! :)

 

At 12-20V 2A will be an idle for this reg, it will get warm, but not too hot. As I said it has good heatsinking. I'll connect up a 3A load on the weekend ad let you know how it goes. Whats max continuous current you need? 

  • Like 1
Link to post
Share on other sites

After testing for high current ability and noise levels I wouldn't recommend this shunt reg for computer use.

Thanks for the infor Mick. Great reading a practical experiment was conducted. I often found myself engage into an argument unintentionally when people just speculate with mindset "I know it all" without a practical experience(s) on a subject matter rather relying on non-scholar texts published on the net. :)

Anyway, I hope to catchup with you soon! ;)

Edited by Chanh
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recently Browsing   0 members

    No registered users viewing this page.




×
×
  • Create New...