Regardless of who has the superior algorithm, heavy lifting tasks can be performed much more efficiently and with much higher precision on chips and FPGAs. This is a well known fact....

[Blizzard]

Not a good design for me...

 

It's going to be purpose built for my system. I'm not going to sell it as a stand alone unit.

[Blizzard]

Kool Aid.

 

Almost every one of the "purpose built" servers on the market are general purpose computers, dressed up to look nice. You can certainly say they sound better to you, but claiming they are objectively better and sound better and so on - that is very shaky ground indeed. It is also why the "objectivist" guys around here sometimes feel themselves alienated and become the target of intense criticism.

 

Most of those folks can clearly see the point I just made, and the criticism drives them into a bit of tunnel blindness.

 

-Paul

 

I know what they are. What your paying for is the time they put into them, not the parts cost. If the end result isn't worth it for you, it's not a product for you. But some of these guys still sell tons of them. Not everyone is a computer guru. So to them these devices aren't computers.

[YashN]

It's going to be purpose built for my system. I'm not going to sell it as a stand alone unit.

 

But if you want the ultimate SQ, there are a few requirements for architecture...

[YashN]

but claiming they are objectively better and sound better and so on - that is very shaky ground indeed.

 

Not if you consider what EMC is and how ground plane noise affects SQ. c.f. John Swenson.

[Blizzard]

But if you want the ultimate SQ, there are a few requirements for architecture...

 

Yes I know. My sound quality reference will be the Mirus SD Card transport. So it has a lot to live up to.

[Paul R]

Still not good enough for me, if you're doing a direct connection to a DAC.

 

That's why I'd prefer a custom audiophile motherboard, with the least circuitry possible to use as an NAA in client-server mode.

 

EMC implications are huge for SQ.

 

Well, part of me agrees with you, and another part of me wonders how much difference could possibly be detected between any similarly powered host machine - when using a NAA. The whole point of the NAA is to get a low power unit out there playing the music, while all the noisy brute force stuff is done on a server where EMI doesn't make a bit of difference.

 

EMC may not have *any* implications *at all* in a case like that, at least so far as the server goes.

 

-Paul

[Paul R]

Not if you consider what EMC is and how ground plane noise affects SQ. c.f. John Swenson.

 

That depends - a $150 REGEN box eliminates almost all of that issue, pretty much regardless of the box that is originating the USB data stream.

 

And again, in a Client/Server setup, such as when using NAAs, I cannot posit any possible connection between EMI,RFI,or other electrical hash in the server and the sound at the NAA. Other than of course, if such noise reaches a point to cause bitstream errors.

 

-Paul

[YashN]

The whole point of the NAA is to get a low power unit out there playing the music, while all the noisy brute force stuff is done on a server where EMI doesn't make a bit of difference.

 

Exactly, so you do see why having a smaller unit near the DAC, and which is specialised in just getting the info for the D/A part.

 

The principle we're using is just that: this can be used in client-server mode which seems to be near ideal for me, or in the less than ideal straight computer to DAC as most people are presumably doing.

 

So in the case of computer-direct-to-DAC, you would still want to minimise both the hardware and software footprints for EMI/RFI.

 

EMC may not have *any* implications *at all* in a case like that, at least so far as the server goes.

 

Easy to say off the bat like this, but needs system testing with the following:

 

1. No component should interfere with others

2. No component should be susceptible to other components' emissions

3. No component should interfere with its own operation

 

Easier said than done!

[YashN]

That depends - a $150 REGEN box eliminates almost all of that issue, pretty much regardless of the box that is originating the USB data stream.

 

Fairly certain both Superbad and John will tell you the Regen doesn't solve a system's complete EMC. It is in fact also dependent on what Signal Integrity comes before it which is why USB cables also have an effect even with the Regen in (or adding another Green Regen before an Amber one).

 

And again, in a Client/Server setup, such as when using NAAs, I cannot posit any possible connection between EMI,RFI,or other electrical hash in the server and the sound at the NAA.

 

You forget anything that is airborne nowadays as well as the pollution that travels through the mains in most systems.

 

Other than of course, if such noise reaches a point to cause bitstream errors.

 

It doesn't even have to affect the bits to cause large SQ changes.

 

Bit-perfection is necessary but not sufficient.

[Blizzard]

 

Exactly, so you do see why having a smaller unit near the DAC, and which is specialised in just getting the info for the D/A part.

 

The principle we're using is just that: this can be used in client-server mode which seems to be near ideal for me, or in the less than ideal straight computer to DAC as most people are presumably doing.

 

So in the case of computer-direct-to-DAC, you would still want to minimise both the hardware and software footprints for EMI/RFI.

 

 

 

Easy to say off the bat like this, but needs system testing with the following:

 

1. No component should interfere with others

2. No component should be susceptible to other components' emissions

3. No component should interfere with its own operation

 

Easier said than done!

 

My experience has shown CPU load to have a detrimental effect on sound quality. On my media player, I have CPU down to 1% max while playing music up to DSD 256. It buffers the data into RAM (which takes about a second, and brings CPU Load up to around 10%) then stays at a ruler flat 1% while the track is played.

 

On any standard Windows or OSX system I've tried, there's always other stuff going on that affects the CPU load (and sound). What I want to determine is if this no longer matters when using a NAA.

[YashN]

On any standard Windows or OSX system I've tried, there's always other stuff going on that affects the CPU load (and sound). What I want to determine is if this no longer matters when using a NAA.

 

List the assumptions that could make you believe it would no longer matter with a NAA.

[Blizzard]

List the assumptions that could make you believe it would no longer matter with a NAA.

 

Well it seems with HQplayer that nobody is taking into consideration the other sound quality repercussions that come with higher CPU load when performing these heavy weight algorithms. So I was wondering if the NAA eliminates all of this?

 

So with a NAA you can run your server computer at 99% CPU load and you won't experience any sound degradation?

 

This is the main reason I'm not getting better sound with HQplayer. It's causes too much CPU load for my light weight media player, and forces me to use my more powerful NUC that doesn't sound as good and still runs at a pretty high load.

 

Overall doesn't sound as good as my purpose built media player. And that's with using the USB interface for both.

 

My media player just has a low powered Atom processor.

[YashN]

You definitely would need a powerful computer server-side.

 

Once that's taken out of the equation, there's a whole world to explore.

[Superdad]

Well it seems with HQplayer that nobody is taking into consideration the other sound quality repercussions that come with higher CPU load when performing these heavy weight algorithms.

 

REALLY? That is one of the most popular topics here at CA.

Spend some time on the HQ Player and NAA threads and you will see lots of discussion about the balance between computational power, performance with Miska's more intensive SDM algos, as well as what is the right capability for the NAA device (hint: even on the NAA the lightest weight computer is not necessarily the best).

And yes--on the process-intensive HQP desktop player side--OS, power supply, and other h/w and s/w issues still make a difference.

 

--Alex C.

 

P.S. You need to ditch that Atom machine.

[Jud]

Well it seems with HQplayer that nobody is taking into consideration the other sound quality repercussions that come with higher CPU load when performing these heavy weight algorithms. So I was wondering if the NAA eliminates all of this?

 

So with a NAA you can run your server computer at 99% CPU load and you won't experience any sound degradation?

 

This is the main reason I'm not getting better sound with HQplayer. It's causes too much CPU load for my light weight media player, and forces me to use my more powerful NUC that doesn't sound as good and still runs at a pretty high load.

 

Overall doesn't sound as good as my purpose built media player. And that's with using the USB interface for both.

 

My media player just has a low powered Atom processor.

 

Using an NAA will eliminate *electrical interference and noise* from server computer activity as a negative to the sound quality being fed from the client (NAA) to the DAC, since there will be no electrical connection between server and client.

 

If you've been trying to run HQPlayer with a NUC or Atom-powered computer, no wonder it hasn't been impressive. That's exactly the opposite of what it's meant to be run on.

[Blizzard]

REALLY? That is one of the most popular topics here at CA.

Spend some time on the HQ Player and NAA threads and you will see lots of discussion about the balance between computational power, performance with Miska's more intensive SDM algos, as well as what is the right capability for the NAA device (hint: even on the NAA the lightest weight computer is not necessarily the best).

And yes--on the process-intensive HQP desktop player side--OS, power supply, and other h/w and s/w issues still make a difference.

 

--Alex C.

 

P.S. You need to ditch that Atom machine.

 

I suppose I haven't spent enough time on the HQ player thread.

 

That Atom based machine is the best sounding computer I've heard yet. Beats my I5 Mac Mini, I7 MacBook Pro, and Broadwell NUC. And it's completely fanless.

[Blizzard]

Using an NAA will eliminate *electrical interference and noise* from server computer activity as a negative to the sound quality being fed from the client (NAA) to the DAC, since there will be no electrical connection between server and client.

 

If you've been trying to run HQPlayer with a NUC or Atom-powered computer, no wonder it hasn't been impressive. That's exactly the opposite of what it's meant to be run on.

 

Maybe your right. The new I7 Broadwell Nuc's have quite a bit of power.

 

http://www.anandtech.com/show/9166/intel-nuc5i7ryh-broadwellu-iris-nuc-review

[Blizzard]

What kind of workhorse would be required to process 20 channels of the best algo's, along with DSP for xovers, room correction?

 

Something that doesn't exist today?

[YashN]

Using an NAA will eliminate *electrical interference and noise* from server computer activity as a negative to the sound quality being fed from the client (NAA) to the DAC, since there will be no electrical connection between server and client.

 

But that's not counting the pollution of the mains, nor anything air-borne...

[YashN]

What kind of workhorse would be required to process 20 channels of the best algo's, along with DSP for xovers, room correction?

 

Something that doesn't exist today?

 

I can do PCM 2-way crossover per channel easily on my old iMac (2009).

 

But using HQ Player and its best filter/modulators as well as the new matrix routing with DSD processing (this I haven't yet tested) sure needs a more capable computer than my iMac (Core 2 Duo).

[YashN]

REALLY? That is one of the most popular topics here at CA.

Spend some time on the HQ Player and NAA threads and you will see lots of discussion about the balance between computational power, performance with Miska's more intensive SDM algos, as well as what is the right capability for the NAA device (hint: even on the NAA the lightest weight computer is not necessarily the best).

And yes--on the process-intensive HQP desktop player side--OS, power supply, and other h/w and s/w issues still make a difference.

 

Indeed, these threads are full of great info. It's time well spent to go through them carefully (including the other one I mentioned).

[Paul R]

EMC may not have *any* implications *at all* in a case like that, at least so far as the server goes.

Easy to say off the bat like this, but needs system testing with the following:

 

1. No component should interfere with others

2. No component should be susceptible to other components' emissions

3. No component should interfere with its own operation

 

Easier said than done!

 

Here we depart from agreement. The Server is totally decoupled from the player with a HQPlayer Server and NAA, so no, the electrical noise on the server can have no effect on the NAA, unless it reaches a magnitude to cause data errors.

 

There is simply no mechanism to tie the two devices together. The data cannot carry the noise with it.

 

Note, those are fairly certain answers, because this is a fairly certain type of thing. Even jitter in the transmission can have no effect, since the transmission is fully buffered and the two devices are not electrically connected in any way.

 

Now, does the construction of the NAA matter? I am pretty sure it does, though, being a low powered relatively quiet device which never really has to strain at all to play the music, I do have my doubts about how much.

 

When you start talking about airborne RFI and such, that has nothing at all do do with the server. To begin with, it is limited to what it can put out these days, and to continue, the server and the NAA are probably not going to be in anything like close proximity to each other. That defeats the purpose, because with HQPlayer, you want a beefy server to do the heavy lifting. Very few beefy servers are also quiet. The NAA is, or is almost silent.

 

-Paul

[jabbr]

Well it seems with HQplayer that nobody is taking into consideration the other sound quality repercussions that come with higher CPU load when performing these heavy weight algorithms. So I was wondering if the NAA eliminates all of this?

 

So with a NAA you can run your server computer at 99% CPU load and you won't experience any sound degradation?

 

This is the main reason I'm not getting better sound with HQplayer. It's causes too much CPU load for my light weight media player, and forces me to use my more powerful NUC that doesn't sound as good and still runs at a pretty high load.

 

Overall doesn't sound as good as my purpose built media player. And that's with using the USB interface for both.

 

My media player just has a low powered Atom processor.

 

The idea with HQPlayer is to isolate the noisy heavy lifting number crunching on a quad i7 3.5 Ghz or better machine. I do this via optical networking.

 

My NAAs are low powered -- these are also very inexpensive. The first one I've built has a DN2800MT board and a PPA V2 USB card. I've also built a mini ITX based J1900 with a fiberoptic NIC. I'm thinking of using my Regen Green as an output USB reclocker. I'm building a J1900 micro ATX which will accept 3 PCI-e cards, so 1) fiberoptic NIC, 2) PPA V2 ... all these are low powered LPS picoATX type machines.

 

So yes, the HQPlayer machine is isolated from the NAA machine. There is some work that can be done to reclock with low phase noise clocks ala what PPA does. He has had good results reclocking a $10 ethernet switch using a low phase noise TCXO which he sells for $159. I'm looking into doing this with a $120 fiberoptic switch -- and in case anyone thinks this is voodoo, the high precision clocks are actually used for high performance networks and communications. Clocks that can be made more precise are those for USB & Ethernet ... the PPA USB card does this.

 

So, yes, YMMV but people are indeed taking these issues into account!

 

Assuming that a 4 core i7 3.5 Ghz is good for HQPlayer DSD256 upsampling + room correction (or DSD512 without room sampling) and this is stereo (2 channels), then 10 channels would need 20 cores. You could run 5 4-core machines in parallel. or 2 10-core machines. etc.

[Jud]

But that's not counting the pollution of the mains, nor anything air-borne...

 

Re the mains, there's nothing preventing the server from being on a different circuit than the NAA, different isolation transformers, etc.

 

Re Wi-Fi, you'd want to have a look at how the receiver in the NAA might electrically affect other circuitry in the NAA and/or DAC, and see what can be done to minimize it.

[Jud]

Maybe your right. The new I7 Broadwell Nuc's have quite a bit of power.

 

Intel NUC5i7RYH Broadwell-U Iris NUC Review

 

See jabbr's comment for a better idea of what you'd ideally want to run HQPlayer. You do *not* want to minimize anything (other than possibly mechanical noise from fans if the server is located nearby) associated with the server machine, unlike what you're accustomed to with other player software or DACs, because Miska has set things up specifically so computing power in the server is desirable, and it won't matter re the DAC because all the latter sees is the NAA (client) machine.