A DIY High Quality Hifi Music Server

[Tam Nguyen]

Inspired by others’ DIY music servers and freaked out by over-priced ones in the Hifi market with “just enough†design from engineering perspective but focusing more on cosmetics, I decided to make one myself and here it is.

 

A. MATERIALS
1/ Motherboard: Intel motherboard with Atom chip. I selected a motherboard with no 12 Volt DC input (DC input is ATX one) to keep the motherboard noise level down. Motherboards with 12 Volt DC in have switch mode regulators in them and these are noisy.

2/ RAM: 4GBs of 1R types

3/ SSD: OS installed in 60GB Kingston SSD; music is stored in Samsung 850 EVO 1TB Sata III SSD

4/ CD Drive: I don’t have the need of ripping CDs from this music server for now so I don’t install one.

5/ Power Supply: I build a very low noise, low ripple linear PSU. All the electronics parts used are high quality ones ordered from Element14 and Digikey.
- AC Main Entry is TE Connectivity Corcom top quality one with EMI filter
- I used high quality, low ESR Panasonic capacitors with total over 70000 uF of filtering at various stages to keep output DC very smooth
- High precision tantalum capacitor, Vishay foil resistor and Bourns potentiometers to keep the output DC voltage very stable
- Oversize 300 VA toroidal transformer
- (Over-engineering) EMI and common mode noise filter at PSU output with top Murata EMI and common mode noise filter down to 1 Mhz.
- 12V DC for USB PCI card is filtered further again and EMI and common mode noise is further filtered with top Murata part to make sure this power is very clean.

The whole music server consumes at about 30-40 Watts maximum but the PSU’s capacity is of over 150 Watts with oversize transformer and lots of capacitance energy storage. The big heat sinks of the case are there to cool down rectifier Diode Bridge and regulators but they never get hot to be cooled anyway. RAM can be a bit warm if running for many hours but the temperature of the whole motherboard and Ram is about 37 Celsius degrees.
For me, power cables do very little and have very little positive effects. They might have some good improvements for some particular system combinations by luck or by very poor PSU implementations.

Linear PSU quality is mostly driven by cost among hifi gear makers.
Good PSUs need no expensive power cable and this PSU is one. My CRO has minimum 1mV/DIV and with that, I cannot find any noise in the band from 10 Hz up of the DC output. More precise specs can be measured with better CRO.

 

6/ USB: I use Sotm USB 2.0 PCI card. USB 2.0 is enough for music so I don’t need to use 3.0 for my music server to achieve simplicity in electronics paths. Besides, the motherboard I chose does not come with PCie slot.
May Park from Sotm is very nice to talk to. Sotm USB PCI has low noise linear regulators which regulate +12V down to +8V then +5V to feed USB power line and +5V to +3.3V to feed USB host controller chip (Sotm uses NEC) and its low jitter 30MHz clock.

My DAC is Wyred4Sound DSDSe which has Async USB input so it benefits from these low noise linear regulators on SOTM PCI but it is nice to have low jitter from source as well.
I have modified my SOTM card by filtering USB output with top of the line USB common mode noise, EMI and ESD (Electrostatic Discharge) filter designed specifically for USB 2.0 and also replace some capacitors by Orange Drop ones.

7/ Case: Streacom FC10 Alpha Fanless case from Life FX. Andy of Life FX is nice to deal with. This case is beautiful, solid built and lots of volume for air circulation

B. OS:

Vortexbox (Fedora-based) with minimal services run for music and online music streaming only.

Running the latest LMS (Logitech Media Server) 7.9.0 with Squeezelite player on top of Vortexbox 2.2. This sounds better than Vortexbox Player.

 

C. OVERALL DESIGN:

Focusing on
- Low noise, low ripple, clean EMI and powerful PSU
- EMI control at many stages
- Good and shortest grounding paths
- Eliminate common mode noise and EMI as much as possible; eliminate ESD to the DAC
- Power line filtered at various stages; USB signal line filtered is done via top industry USB filter to make sure signal is not distorted by filtering especially at end of packets.
- USB precision clocking is not important for Async USB application but nice to have it in place

D. SOUND QUALITY:

the music server allows Wyred4Sound DSDSe DAC Async USB to operate at its best. High range and low end are improved significantly and very noticeably. Music is much cleaner and 3D imaging is very satisfying. I use an Audioquest very short USB cable to save money but it has enough shielding to keep RFI out.

E. FURTHER POSSIBLE DEVELOPMENT IN MIND:
- Build an Ethernet filter adaptor to filter out common mode noise and EMI from LAN Ethernet going to the music server.

 

F. CONTROL SOFTWARE:

- I am using iPhone so iPeng8 on iOS is a good control software. It costs 10$ but it does its job well.

 

G. SOME PHOTOS: 

Edited July 19, 2015 by Tam Nguyen

[Guest myrantz]

Wow.. Real nice! What's the second USB card (below SoTM?).

If your SATA cables have the metal clips (SATA 2.0 and above), in theory they would already be shielded. So you probably don't need that aluminium foil.. 

[Tam Nguyen]

I made the USB card first using usb out from motherboard but I filtered out EMI and common mode noise on 5V and USB signal; then I installed SOtm card to compare. I filter the output of Sotm as well.

 

My card has clean 5V out for usage in future.

 

Agreed! some are not very necessary such as aluminium foil on sata cable but I have bought a 50 meters of it so why not use some for those sata, no harm , just in case.

I have to admit, some are over built but nice to be that way than under built.

Edited July 6, 2015 by Tam Nguyen

[Guest myrantz]

I made the USB card first using usb out from motherboard but I filtered out EMI and common mode noise on 5V and USB signal; then I installed SOtm card to compare. I filter the output of Sotm as well.

So the DC plug on the 2nd USB card, it's to power which device?

 

My card has clean 5V out for usage in future.

Nice..

 

Agreed! some are not very necessary such as aluminium foil on sata cable but I have bought a 50 meters of it so why not use some for those sata, no harm , just in case.

Do you use kitchen alumimum foil or something more exotic? If the former, you can always use 'em for baking

 

I have to admit, some are over built but nice to be that way than under built.

True. If it's worth doing, it's worth over doing. Love that PSAudio P10 (I think) and LS50 in the background.   

Having said that, these Streacom cases are breeding like rabbits..

 

You probably don't need PCI-e. But just in case you do, you can get those mini pci-e to PCI-e 1x adaptors (Your motherboard should have 1)... They work well but YMMV. I'm using one just so I can use an Intel NIC, and I disable the onboard Realtek chip. This is probably snake oil but I really prefer Intel NICs over anything else. Good upgrade to have even if you do store your music locally. I'm going abit far by getting this, but most cases you just need 1 ... Obviously you have no PCI slots left now, but something you can consider in the future if you find you don't need the 2 internal USB ports.

[Tam Nguyen]

"Do you use kitchen alumimum foil or something more exotic? If the former, you can always use 'em for baking :P"

The best is 3M EMI tape ordered from Digikey. Kitchen one should stay in kitchen

[Guest myrantz]

"Do you use kitchen alumimum foil or something more exotic? If the former, you can always use 'em for baking :P"

The best is 3M EMI tape ordered from Digikey. Kitchen one should stay in kitchen

Ha! I stand corrected!

[Tam Nguyen]

 

So the DC plug on the 2nd USB card, it's to power which device?
 

The DC plug can be used as 5V output out. Not used now but can be used in 2 cable setting - one carrying USB signal only and one carrying 5V only. This can be implemented  when USB cable is long but I need to make an adaptor with EMI and common mode noise filter to put signal and power from these 2 cables together right before the DAC and feed to the DAC. Not very necessary though. Only need this setting when people want to go extreme.

 

a note that Sotm has 5V on/off switch and ground floating jumpers. These can be set as well if the DAC does not use 5V from USB and/or ground floating is needed. Good grounding design does not need this ground floating anyway, I believe. My W4S Dac does need 5V from USB in, may be for USB part or both USB part and clocking part. Not knowing in very details about it but it needs 5V in from USB so I let this Sotm switch on.

 

W4S claim its Async USB part has good isolation. Sotm claims its USB output is ESD filtered but still I filter ESD/common mode noise/EMI further to make sure each stage should give a clean signal to the next; no dependent.

Edited July 6, 2015 by Tam Nguyen

[davewantsmoore]

Focusing on

- Low noise, low ripple, clean EMI and powerful PSU

- EMI control at many stages

- Good and shortest grounding paths

- Eliminate common mode noise and EMI as much as possible; eliminate ESD to the DAC

- Power line filtered at various stages; USB signal line filtered is done via top industry USB filter to make sure signal is not distorted by filtering especially at end of packets.

- USB precision clocking is not important for Async USB application but nice to have it in place

 

 

 

still I filter ESD/common mode noise/EMI further

 

   These things can be pretty loud inside a PC... and they can cause quite extreme amounts of jitter later (sometimes difficult to measure due to its transient nature)

Edited July 6, 2015 by davewantsmoore

[Tam Nguyen]

Anyone in Canberra are welcomed to come over to audition.

Or can arrange time to bring it to your place to test with your system.

I have a Pioneer N50 music streamer and Wyred4Sound Remedy reclocker but i must say with this experience that spdif is hard to get near to Async usb if DAC and music server/source is implemented well, even if reclocker is used.

This combination ensures more warmth, clarity, deep bass, musical and sharp high end, and wide open sound stage. Also dynamics are maximised. The sound is 3d, and very smooth.

Edited July 6, 2015 by Tam Nguyen

[andyr]

Sensational - well done!

 

Andy

[Tam Nguyen]

Thanks, Andy

[Equaliser]

Very nice, I have the identical Streacom case that I purchased for $15 that I want to put to good use and am not a fan of the inferior PicoPSU's. I'm very interested in your Power supply design. 

Edited July 7, 2015 by DefQon

[Tam Nguyen]

Very nice, I have the identical Streacom case that I purchased for $15 that I want to put to good use and am not a fan of the inferior PicoPSU's. I'm very interested in your Power supply design. 

Thanks.. You had a very good buy with 15$ for a Streacom case. I spent 450$ to get mine

[davewantsmoore]

I am also interested in both those things  ($15 streamcom cases, and power supply used)

[mattjtaylor2809]

What was the cost of all the parts Tam? I assume the SSD's swallowed most of your budget?

Sent from my iPad using Tapatalk

[Tam Nguyen]

Re. the cost:

1/ SSD is around 550$ but I got it for 450$. I chose Samsung simply because its price is quite good and I had it on a deal.

2/ Case: Over 450$ (not necessary but hifi people like something beautiful as well  : )). I was thinking very hard before ordering it.

3/ Sotm Card: if you get a new USB PCi card, the price for it to land in Australia is around 450$ but it is quite hard to get the PCI one because they are discontinued.

4/ This might sound strange but the electronics parts cost the most. For this particular music server, they cost around 700$. The quality parts are not cheap especially buying small amounts. Because I don't have to pay for marketing, hifi reviewing fees, sales, labor costs (I did it with funs) etc., I could spend that money into high quality parts. An example, a high precision resistor alone costs me more than 17$ not including GST and shipping (more than half of the price of Crystek 957 femto clock used in DACs over 2000$); parts used to filter EMI, common mode noise and ESD in signal line and data line are also not cheap. However, they pay back in the sound quality.

 

I had to buy a Oscilloscope to measure : ). But I can use it for other projects in future then.

 

I could have gone with more cost effective solutions such as a multi-purpose/multi-output external Hd-plex 100 watts Linear PSU, landing here for about under 500$ with beautiful case or Paul Pang PSU but I didn't think they can come near to this PSU, at least by the quality of parts used so I decided to make it. I have a teradak PSU; it is much better than Switch mode PSUs but still it has noise at 2Khz down and the ripple is not very excellent. Switch mode PSUs have high frequency noise and linear psus normally have low frequency noise in audio frequency range if not filtered out properly. Looking into eBay and I could find the same parts I was looking for with very low prices but never know if they are counterfeits or low quality ones (for example, with 4$ I could buy 10 connectors from ebay but with that amount, I could get only 1 here in Australia). All the parts I used are from Digikey and Element14, some PCb boards from jaycar. I could have ordered PCB boards from Element14 to make the music server look nicer .

 

5/ Others (RAM, motherboard, cable, EMI foil, etc.): a couple of hundreds more.

 

Total cost is not low, but the sound quality is amazing to me. And still I save heap comparing to the cost of Antipodes DX or the likes reaching Australia, especially when the AUD is going down.

 

Because I made it for myself, I made things which make sense the most in terms of electronics, not based on unreal theories/stories. To compare, personally, it is hard for me to believe the PSU used in Antipodes can be as good as mine. You can Googling for the photos and look into their PSUs to see. Noise, poor power, and jitter are the enemy of hifi. The other things some hifi gear makers usually make a fuss about them in specs, marketing materials and "technical papers", "educational papers", FAQ, such as what software used, what SSD used etc. can make the sound different are, to me, nothing but non sense. Or they don't know what they talk about or more correct, they do it on purpose; they need to sell enough stories to make the prices look reasonable. Maybe the Titan can hear out the difference made by the brands of SSD used. I am human, no I can not hear out. Maybe pride of ownership makes people's brains hear out things. Even measuring of the USB output signal for 2 scenarios with 2 different SSDs used cannot tell the difference let alone hearing.

 

The clean 12V combining with Sotm USB PCI card produces good sound quality but filtering the output from Sotm adds another level of sound quality improvement. I don't think others like Antipodes do it. In those, Sotm is the  final output.

 

I have not calculated in details yet but it may cost me up to 2400$, but it is very satisfying. I am in Canberra so very hard to compare it with other top music servers or streamers but I am going to hook it into a PS Audiso Direct Stream Dac to test, maybe next week. My Wyred4sound DSDse DAC is very good and I was not happy with all the sources I used so far with it even I added reclocker for spdif with good interconnection cables. But this Async USB thing working directly with DAC's Async USB to I2S opens up the whole new Wyred4sound DAC capacity which I have never experienced before with that DAC. The music servers with digital output like this ones do only one job - giving clean digital signal and power to DACs (more important) (+5V via USB) and less jitter (lesser important with Async USB DACs). Actually DACS are supposed to do cleaning-up and isolation job well but unfortunately, in reality we cannot expect much from them in this aspect. I had a good sounding and well-known DAc in the past with very poor USB isolation and ESD, common noise filtering. Turning on/off any light in my room will affect the sound due to the DAC handling noise/ground loop poorly. After a while it's gone wild because ESD from laptop degraded the components inside the DAC. The rest are decided by DAC, pre-amps, power amps, signal cables and speakers and power conditioner. Each thing has its job and  for me any of these has the same importance in hifi. Any weak point in the chain will degrade the whole system in terms of sound quality. Lots of people focus on analog but forget about digital source.

 

Wish I could do listening tests and compare with other best music servers in future.  

Edited July 7, 2015 by Tam Nguyen

[davewantsmoore]

I had to buy a Oscilloscope to measure : ). But I can use it for other projects in future then.

<snip>

Because I made it for myself, I made things which make sense the most in terms of electronics, not based on unreal theories/stories. To compare, personally, it is hard for me to believe the PSU used in Antipodes can be as good as mine. You can Googling for the photos and look into their PSUs to see. Noise, poor power, and jitter are the enemy of hifi. The other things some hifi gear makers usually make a fuss about them in specs, marketing materials and "technical papers", "educational papers", FAQ, such as what software used, what SSD used etc. can make the sound different are, to me, nothing but non sense. Or they don't know what they talk about or more correct, they do it on purpose; they need to sell enough stories to make the prices look reasonable. Maybe the Titan can hear out the difference made by the brands of SSD used. I am human, no I can not hear out. Maybe pride of ownership makes people's brains hear out things. Even measuring of the USB output signal for 2 scenarios with 2 different SSDs used cannot tell the difference let alone hearing.

<snip>

Actually DACS are supposed to do cleaning-up and isolation job well but unfortunately, in reality we cannot expect much from them in this aspect. I had a good sounding and well-known DAc in the past with very poor USB isolation and ESD, common noise filtering. Turning on/off any light in my room will affect the sound due to the DAC handling noise/ground loop poorly.\

 

    

 

 

Wish I could do listening tests and compare with other best music servers in future.  

 

Just keep using that 'scope, IMO       (although listening test can be fun....   music!)

 

Have you looked at the analogue output of your DAC under version conditions?   It is very interesting to look at the effects on the final signal when you do some extreme things to the source  (like add or remove filtering, or inject a lot of noise).

[Guest myrantz]

Because I made it for myself, I made things which make sense the most in terms of electronics, not based on unreal theories/stories. To compare, personally, it is hard for me to believe the PSU used in Antipodes can be as good as mine. You can Googling for the photos and look into their PSUs to see.

Be careful saying things like that.. People may insist on you to do a blind test, else you risk the dreaded term snake oil   .. Having said that, will be interesting if you can pit your unit head to head against an Antipodes and see if you can pick it .

[Tam Nguyen]

Be careful saying things like that.. People may insist on you to do a blind test, else you risk the dreaded term snake oil   .. Having said that, will be interesting if you can pit your unit head to head against an Antipodes and see if you can pick it .

I don't sell it so I can say this . Besides, I used the term "personally". Just my personal idea. Not a statement then . Safe! If could be, I would love people help me do listening tests including blind tests. I am very confident it makes people happy.

Edited July 8, 2015 by Tam Nguyen

[Tam Nguyen]

    

 

 

 

Just keep using that 'scope, IMO       (although listening test can be fun....   music!)

 

Have you looked at the analogue output of your DAC under version conditions?   It is very interesting to look at the effects on the final signal when you do some extreme things to the source  (like add or remove filtering, or inject a lot of noise).

I did look at the USB signal in action with and without filter, cutting off a standard USB cable for that measuring purpose. The filtered signal of course looks better from the CRO.

[davewantsmoore]

Be careful saying things like that.. People may insist on you to do a blind test, else you risk the dreaded term snake oil 

 

Gosh ..... someone might even ask to see some measurements of the performance of the power supply .....  so they could verify for himself if this power supply was different from another power supply.   Duck and cover.

 

 

I did look at the USB signal in action with and without filter, cutting off a standard USB cable for that measuring purpose. The filtered signal of course looks better from the CRO.

 

 

Having a look at the analogue output of the DAC can show you how much of this is making it through.   Use a jitter test similar to the one they use in stereophile.    In can be very instructive for evaluating changes to things.

[Tam Nguyen]

Gosh ..... someone might even ask to see some measurements of the performance of the power supply .....  so they could verify for himself if this power supply was different from another power supply.   Duck and cover.

 

 

 

 

Having a look at the analogue output of the DAC can show you how much of this is making it through.   Use a jitter test similar to the one they use in stereophile.    In can be very instructive for evaluating changes to things.

"someone might even ask to see some measurements of the performance of the power supply":

 

Measuring during making a DIY to adjust things is simple measurements; measuring to come with specs to show to others are not that simple, especially when you sell things to people. This involves good measurement methodologies/industry methodologies used to be able to get good measurements, very good test devices, and measurement devices, etc. These devices are not cheap; a DIY person cannot have an advanced lab to work with . How many audio gear makers give their measurements/graphs out to public/consumers? None. Does guys like Antipodes or the others who sell music servers for big prices have measurements? Do 2000$ up PSUs in the markets have measurements? None. Some serious makers of course go through a loop of designing,  prototyping, measuring, and testing. Some small companies cannot have that or very limited capacity. Big names can give good specs. Some give not very correct specs. Some list the features/parts used etc. in specs. They are not real specs actually.

 

If somebody wants to compare 2 gears using measurement data/graphs, they need to have both measurements from both 2 gears and they need to some basic knowledge in electronics to understand them before they can compare them using measurement data.

 

The guys selling music servers for thousand dollars, can they provide measurement? Nope .

 

I could have taken some photos of the graphs on my CRO and list them here but they are not precise enough to be specs; they are good to support while working on the electronics parts. My CRO is 40 Mhz and 1Mv/div, 0.1 us/div. Good to support when making things but not enough to have get data out as specs. I have to be realistic with that and that is truthful as well.

 

Besides, the CRO is not the new digital one with lots of features such as graphs in frequency domain so lots of indirect measurements have to be calculated by hands.

 

In all the gears, regulators with big current cannot produce ultra low noise voltages. That why good designs go with may stages of regulators and noise filters. This is the rough data of my PSU: no ripple can be seen at 10 Hz up to 40 Mhz (the limitation of my CRO) at 1mv/div (my CRO cannot go further) so I cannot claim further than that. But that is very good already for a PSU which can give out 7-8Ampers. It also has EMI filters for 1MHz up (Linear PSU won't normally have that EMI but nice to have filters).  Sotm usb card will do further with low noise linear regulators for quite large currents then ultra low noise regulator stages for smaller currents which are fed to USB host controller, to clock and to 5V for USB outputs.

 

What Sotm may be lacking is the EMI and and common mode filter for USB from the USB controller to USB ports. I add the filter to that stage and to my ears, the sound is improved.

 

If circumstances allow, good measurement is of course a good roof; the design is another way to tell how good the gears are. Listening is the third way to tell. While many make their things to sell cannot have measurements, how come a DIY can go further and give all that with the costs of time and money

 

"Having a look at the analogue output of the DAC can show you how much of this is making it through.   Use a jitter test similar to the one they use in stereophile.    In can be very instructive for evaluating changes to things."

 

Incoming Jitter (to the DAC) is not a problem in DACs which support Async USB.

 

If you have a PS Audio Direct Dac, incoming jitter is not a problem in all digital inputs. So probably, a toslink is the best digital interconnection used with that DAC. Ted Smith of PS Audio mentioned that clearly.  Toslink can add incoming jitter but that DAC does not read clock information from the source anyway so that DAC does not get incoming jitter in. The only problem with that DAC, like all other DACs, is the noise from incoming signal and Toslink has no noise.

Edited July 8, 2015 by Tam Nguyen

[davewantsmoore]

Measuring during making a DIY to adjust things is simple measurements; measuring to come with specs to show to others are not that simple

 

This is true.   It takes a lot of effort to present measurements in a way which "has meaning".

 

How many audio gear makers give their measurements/graphs out to public/consumers? None.

 

Yep.   They're easy to misinterpret and misuse.

 

 

If circumstances allow

 

Sorry.... I think I might have confused you.

 

I wasn't asking to see measurements of the power supply... or trying to be mean to you   :-)

 

I was (trying to) have a go at the "measurements are meaningless" crowd.

 

 

... but like I said - keep in mind that the most important measurement to keep in view when tweaking - is the analogue output of the DAC.   That is one which is audible  

 

 

 

PS - I am still keen to hear more details about the circuit you used for the DC supply.

[davewantsmoore]

What Sotm may be lacking is the EMI and and common mode filter for USB from the USB controller to USB ports. I add the filter to that stage and to my ears, the sound is improved.

 

Can you tell more about this?   Where you put the added filtering?

 

(I use a firewire multichannel audio interface in my main system.... but a while back I have built a few PCs for people using SoTM PCI card)

[Tam Nguyen]

This is true.   It takes a lot of effort to present measurements in a way which "has meaning".

 

 

Yep.   They're easy to misinterpret and misuse.

 

 

 

Sorry.... I think I might have confused you.

 

I wasn't asking to see measurements of the power supply... or trying to be mean to you   :-)

 

I was (trying to) have a go at the "measurements are meaningless" crowd.

 

 

... but like I said - keep in mind that the most important measurement to keep in view when tweaking - is the analogue output of the DAC.   That is one which is audible  

 

 

 

PS - I am still keen to hear more details about the circuit you used for the DC supply.

 

 

No worries. I understood yours well . Just tried to discuss further a bit. Agreed with measuring analog output of the DAC. But I use the DAC made by W4S and I made the music server so I only measured USB output of the music server. Have to have an good Oscilloscope to be able to measure analog out.

Edited July 8, 2015 by Tam Nguyen