Head-Fi.org › Forums › Head-Fi Special Forums › Premier Sponsor Forums and Sponsor Forums › Sponsor Forums › DIYHiFiSupply.com › Music Server Computer transport - Hop On!
New Posts  All Forums:Forum Nav:

Music Server Computer transport - Hop On!

post #1 of 7
Thread Starter 
I first saw one of these demo'd about 5 or 6 years ago at CES. It was a big brute floor standing computer, loaded with rips of CDs and driving some really nice high end gear. Since then they've become a common sight at high end shows everywhere but I've always found the performance spotty and not that attractive. It's been mentioned in several show reports too.

I don't really understand all the issues and problems at work but we've had someone with us this week who does. Well, my CDpro-based transports are by now getting a little long in the tooth, there's nothing comparable on the horizon AND my CD collection is showing wear and tear -- I decided it is time. So I was given a shopping list.

Here's what we want to do:
1. Use a computer to rip, store and provide playback functionality. Preferably a touch screen with remote as this one can not be used for any other purpose once optimized for music playback. We need about 1 terabyte of storage to start.
2. connect the digital playback stream from the computer to our DAC (in this case the Satch) and then onto all the usual tube based linestage, amplifier and full range speakers. BUT we don't want the terrible-ized sound of the computer's onboard sound card. It's really ugly, folks. So we will take the signal, bit perfect from the USB through ASIO.

Here's what I bought:
1. MSI Wind Neton All-In-One, 19" Touch Screen
2. WD MyBook 1 terabyte (2 pcs, one for backup)
3. Musiland Monitor 01 USD USB (much discussed in the forums as it converts the ASIO usb output to SPDIF and allows 192khz/24bit)
4. Windows MCE Remote with receiver

And here is the picture of the MSI Wind Neton as set up. the black box contains the usb to DAC interface. More about that in the next blog.
post #2 of 7
Thread Starter 

Music Server Computer transport - Round 1!



OK so all the bits arrived, we connected them together: MSI Wind Neton>>>Musiland USB to SPdif>>> the Satch DAC>>>UTS-UVC linestage>>>Fusion>>>LCAO full range speakers -- something like the picture save the black box in the centre which came later.

First task: configure the digital playback stream to bypass the computer on board sound system and output through the USB>>Musiland. I download an ASIO music player and got some music to play that way.
OK, better than using the sound card, but not as good as our standalone transport. But there is still much we can do, so -- promising.

Now aforementioned expert arrives on the scene. Here's how he explains the issues:

" ...both the Apple Macintosh and PC Audio core build into the operating system are not designed for absolute quality, but for
convenience, which raises some interesting issues.

By default, none of the existing computer configurations offer a “Bit Accurate” digital output, regardless of the connection method used. Instead CD audio is always resampled to 48KHz to comply with the Windows Audio standard set down under the name Audio Codec '97 (AC97).

This means in order to get a PC to perform as well sonically, as a good quality CD Transport (or indeed better than most, to which it has the potential) we need to bypass this operating system audio section."

The computer we choose for this task is an All-In-One, meaning screen, cpu and all hardware is mounted in one chassis. Like a notebook it has a fan which comes on based on system temperature. In this case we don't want a fire breathing screamer as those things are a Dante Inferno of heat and noise, which is bad for the perfect bits we want.

He goes to work stripping out bloatware from the OS (XP) and setting bios to use minimum of system background resources and background processes. In the end, playback only uses about 5% of CPU. Now all the Media Portal software, skins, plugins and latest Musiland drivers are installed. Not a small job and it takes the expert most of the day to get it configured.

CD's are now ripped with the correct plugin and we listen. Better! Still not up to the separate CD transport, but we've taken a big step forward and we still have the hardware to attack! Interesting we try an A/B playing a CD by the on-board CD reader compared to the same ripped and playing from the hard disk. Very similar but after listening for a while it is apparent that the CD reader is just a little less good than the ripped CD, less space, less openess and detail - just a little but definitely noticeable. The expert explains this is probably because the CD reader puts a load on system resources which dulls the sound.

OK now for a little reconstructive surgery - not for the squeamish!
post #3 of 7
Thread Starter 

Music Server Computer transport - Hardware Attack!

OK the Magic Black Box.
Much has been written about the Musiland computer-to-external DAC enabler. Just plug in the USB data stream from the computer and out the provided SPDIF comes our DAC friendly perfect bits, even up to 192khz/24bits.

That's pretty good news. However there are 'issues':

1. the Musiland is powered through the USB cable, drawing trashy DC from the computer's power supply
2. the on board 24mhz clock signal is a total disaster
3. the electrolytic caps are undersized
4. the 2 pulse transformers massacre the data stream
5. some parts values are not ideal
6. 75 ohm digital friendly cable is not used
7. the input side circuit can be greatly improved with some simple changes

So we remove the (cough, cough) clock, 4 of the SMD resistors, the 2 pulse transformers, the 5 10uf caps, the offensive wire and replace with suitable bits plus some hocus pocus to the circuit performed by the expert.

We power the pcb with the DIYHFS VCS regulated supply driven by its own power transformer

We feed the clock signal from the Ultimate Clock and Power supply system with a shiny new .5ppm 24mhz OSC.

All is installed in the Black Magic Box as shown.

Now we have the Musiland DAC enabler where we want it: untethered from the computer power supply and operating with lowest jitter from a proper clock which has its own power supply.

We put the box together and reconnect the system. Is it worth it?

post #4 of 7
Thread Starter 

Music Server Computer transport - Is It Worth the Effort?



So we finish the previously described modifications to the Musiland and install the Magic Black Box. The system boots up and recognizes all hardware. After a ten minute warmup we play music.

Playing 'Dreams of Likembe' we hear bass notes that weren't noticeable before. Soundstage is more focused and detailed, less starined. In fact a pervasive 'hardness' that was there before in varying degrees is gone.

This is a big step forward. The effect of swapping bad clock/good clock is sometimes subtle, sometimes way obvious depending on how much it's needed. In this case it is very obvious. If you've used vinyl playback it most reminds of what happens when you tighten up a cartridge that is loose in it's headshell -- things snap into focus, bass goes deep, there is more dynamic spread and less noise/more music.

But the real question is how does it compare to a standalone CD transport. Are we close?

We have 4 listeners present. We cue up the CD transport and the Computer transport - each playing through a Satch DAC. For about 20 minutes we play tracks from All For You (great voice) and Mozart Salsburg Sympnony (great strings), switching back and forth without knowing which is the source.

We can't hear a difference. This is the first time that happens (usually we hear a difference but may not be sure which is better until after a long listening session).

It may be that something else is masking the difference because surely there must be some. I plan another listening session with fresh ears and music (actually that listening session fell through as the 'ears' didn't show when expected so will do this later). We will either confirm or refute the result.

In either case this is encouraging because there are still more things we can do to make it better:
A linear DC supply for the computer to replace that SMPS provided.
Underclocking the CPU (we are only using 5% so if we underclock by 50% or so we can get rid of a lot of heat/noise, ie turn the hose on Dante's Inferno).
One other thing which we'll try before mentioning as it's a little 'unconventional'.

The question was asked if the Musiland really is asynchronous? Something to help understand the issue and food for thought:
asynchronous:

"In a synchronous system, operations are coordinated under the centralized control of a fixed-rate clock signal or several clocks. An asynchronous digital system, in contrast, has no global clock: instead, it operates under distributed control, with concurrent hardware components communicating and synchronizing on channels."

So either it is or it isn't. The Musiland uses an onboard local clock and its own driver instead of an MS driver. I'm not sure that is an issue relating to sound quality.

The key here is how good the audio clock is. Being asynchronous means its output jitter depends on how good the Musiland audio clock is.

In the Musiland a simple 24MHz clock generator is used that runs off the main powersupply, which is fed directly from the PC's USB power line. This will create a very jittery reference clock for starters.

This clock forms the reference of a "fractional division PLL". Such a PLL is used to make non integer multiple clocks from a reference clock. It can never be any better than the reference clock. The DIYHFS EZ Clock with a 0.5ppm clock generator provides as close as possible to the ultimate as reference clock, eliminating this issue.

There are many varying PLL implementations, however a crucial thing here is that the PLL will contain another clock generator which is MUCH more subject to jitter from power supply noise. So by feeding this PLL a clean supply, jitter is minimized.

Purely from a viewpoint of potential jitter (which we cannot verify as having been achieved or not) the Musland system (the chips used) seems to be capable of sub 100pS jitter results, something the Musiland guys also claimed (but their actual design does not achieve it).

All that said, the common Cirrus Logic (Crystal) Receivers and most others add over 100pS jitter of their own, so I should think the SPDIF Receiver would be the limit in the DAC.

If receivers used have low jitter of their own and equally good suppression of external jitter, all of this is moot anyway.

The Musiland is cheap and the mods done do not cost the earth either (there are more expensive USB2SPDIF boxes out there that do less). So I think one can easily defend the Musiland USB to SPDIF converter with modifications as a good way and economical way to make a computer transport for a pre-existing SPDIF DAC capable of 24/192KHz.
post #5 of 7
Thread Starter 

Music Server Computer transport - More!



Musiland USB to DAC:

We left off last time with the resistors in place of the pulse transformer to the SPDIF. Getting rid of the transformer and adjusting the output to 75ohms really opened up the sound.

The new transformer parts arrive and we install one in place of the resistors. The other output still has the resistors so we are able to A/B:

Nov 20
Today we installed the upgraded transformer into 75ohms and had a preliminary listening A/B between Rout and TXout. I'll write more later after we do something more extended but briefly:

There were 4 of us and 3 didn't know which was which, Rout or TXout. We listened to one track, about 8 passes. After the first pass 2 of 4 listeners choose Rout as better detail and bass, 2 listeners choose not to comment.

By the third pass the tide changed to favour the TXout and by the last pass it was unanimous the TXout was best. It took a few passes to identify that the extra detail was really a kind of aggressiveness and the bass was somewhat exaggerated.

All agreed the TXout had more real detail and more tuneful bass. Dynamics were about the same. The listeners liked the word 'natural' to describe the TXout. Next week I'll check out some bass heavy tracks to find out if the lows are rolling off at all. I'll then install a TX in second output with a little different loading. Once we get a winner there then I'll do the BNC connection.

Nov 23
We confirm all the bass is still there, just less overhang. More listening also supports the differences noted above.

Comparing CD transport to Computer transport -- Round 2:

Today we compare the current setup to the CD transport. Earlier we were unable to hear a difference which is a first for our listening panel. We have 5 persons in our panel of listeners today,-1 female, 4 males and ages 25 to 40 except for me, saying goodbye to my 5th decade. I do the switching between CD and Computer transports, the others don't know which is which.

Music is a track from a Stereophile disk: Rendezvous. The first track leads in with some very delicate cymbal sounds. The second round was Michel Buble Fever and then Robbie Williams Straighten Up and Fly Right.

Once our ears 'warm up' we do hear differences between transports:
3 listeners feel the Computer is more natural and retrieves more hall sound, ie ambience. One listener says the CD transport has a faster pace and is more lively. The #5 listener can hear no differences. 3 feel the Computer is better overall, one chooses the CD transport -- all by a small margin.

Both sound very good. It seems we are now at a level where differences are more a falling of the shadow rather than markedly better or worse.

Next we'll turn attention to the computer power supply. We want to get rid of that SMPS brick and power from a regulated linear supply. We need 19v/3.9A and I think we can do that with a substantially heat sinked filament supply.
post #6 of 7
Thread Starter 
OK an on-the-ground tech report:

"I had an unmodified Musiland for use as simple SPDIF generator and found my receiver had problems locking consistently (I was debugging the software control for this receiver). Putting a 'scope on the signal out of the musiland showed this, left picture.



The actual signal level is around 200mV peak-peak and there is a huge 100mV spike on each transition.

First, this is WAY below the IEC 60958 type II specification of 0.5 - 1V peak-peak (and low enough that some receivers designed to this standard will not trigger reliably). Secondly the waveshape is UGLY. And NOISY. Brrrr... No wonder the poor receiver was constantly complaining about unlocking.

Okay, out come the original transformers to be replaced by the "magic part" (sorry, cannot reveal which, it is pretty decent but most important it's parasitics are consistent and I know what they are).

Remove the original disastrous resistive divider. A corrected divider was applied AFTER the transformer, with an impedance that correctly terminates the transformer; drive is now from a very low source impedance (the buffer in the Musiland is < 10 Ohm Z-Out).

After that the trace was much more like it, correct levels at around 0.8V Peak-Peak, minimal rining on the edges, but the traces where still furry. So I added an SMD CLC filter (1uF/0.1mH/1uF) between USB input and main circuit (musiland has a place for a choke on the PCB but does not fit, shame, things would be much quieter with choke) and I added a 220uF/10V Os-Con as bulk filter capacitor. Now the traces are not furry anymore. As shown, right picture, but with different scale to allow for the different output level.


A quick before/after listening test showed that most of the hard and mechanical sound quality has gone, now we do not only get all the bits, but most of the music as well.

Doing the full mod with clock and seperate supply is still quite a bit better, but the basic mod alss gives a major step up in sound quality. Oh yes, and I can go back to debugging my receiver control software, for now we have a lock absolutely rocksteady (and the clocks out of the receiver are SO MUCH less jittery, before you could see the jitter with the naked eye on the 'scope!)..."
post #7 of 7

Dear bcherry!

 

Only one question (for now) about the mods, maybe a silly one but i feel better asking than destroying what i don't even have yet :) 

 

The Musiland applies a 24Mhz clock as I see on the pictures (mine have not yet arrived). In the DIYHFS there is only 24.5760 MHz Clock.

Is there something I missed, or did you have one made for the specs?

 

thanks

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: DIYHiFiSupply.com
Head-Fi.org › Forums › Head-Fi Special Forums › Premier Sponsor Forums and Sponsor Forums › Sponsor Forums › DIYHiFiSupply.com › Music Server Computer transport - Hop On!