Rafe Arnott over at Audiostream just figured out that a dedicated server is better than a laptop

[jabbr]

23 hours ago, davide256 said:

What we buy off the shelf is cheap crap for allowed errors, compared to purpose built machines that need to integrate into electrically complex environments with high reliability. I wonder if for audio some of NASA's design requirements for control PC's might be applicable. At least we don't have to worry about radiation effects so much in earth atmosphere.

 

Whaaaaat? 

 

I can tell you with certainty that both NASA and our national research laboratories (think nuclear weapons research etc) have been pioneers in the use of cheap of the shelf computers clustered together https://www.hq.nasa.gov/hpcc/reports/annrpt97/accomps/ess/WW46.html

 

The concept that these cheap computers are designed to allow errors is not even wrong.

 

Do you think NASA and the military have their own foundaries? Design their own CPUs ? ... maybe but just like a CPU becomes specialized with software, so are FPGAs. 

[jabbr]

19 hours ago, Albrecht said:

Uh no...

The difference is single-purpose computers vs multi-purpose computers. Everyone knows what people mean: they just don't write out "single-purpose-computers." There is nothing WRONG with multi-purpose computers. Their GOAL and DESIGN is not to do one thing well. This is what high end audio manufacturers bring to the table, - single purpose machines designed to playback digital files. This means eliminating unused, noisy, and lower quality components that either get in the way, or can add unwanted artifacts in enhanced audio reproduction.

Uh no. 

 

The trend end these days is to put more function on an integrated CPU called a SoC ... ClearFog is an example that I often use as a specialized network I/O board...

 

There is no one design for “audio” this concept is frankly absurd.

 

The requirements for a streamer (eg ClearFog with essentially one chip) are vastly different from what is needed to run HQPlayer upsampling ... in my experience the video ouput has the most noise, but what about folks who want integrated video + audio ... similarly my NAS has terabytes of storage ... why do you think there are so many motherboards all made by SuperMicro as well as ASUS and then NVidia/AMD etc ... each have different options. Design the system that has the options you need. All off the shelf.

[jabbr]

56 minutes ago, davide256 said:

And none of them would be trusted as part of the control systems in a spacecraft... the standards for  spacecraft are even higher than commercial aircraft because

of the extreme environment and because you can't just land anytime when in difficulty. Your speaking to throwaway tech...

Oh c’mon this is nothing to do with computer audio. 

The person here who has actual experience with space system design is @marce so it’s funny that you are arguing with him.

 

I’m just saying that commercial computer equipment hardly has a high error rate even with trillions of calculations. 

 

Yes you want low noise. So called “high end” audio designs use commercial off the shelf parts (digital) although Pass has commissioned custom transistors but regarding digital audio, and specifically the chain before the DAC none of the parts need to be “special” 

 

Real high end electronic design may use $80 - 100,000 software packages and the fact that audio companies can’t even afford mid-range measurement equipment suggests to me that practically none of the board design is done with the resources available to eg SuperMicro and certainly not that of Intel or Apple.

 

If I’m wrong let’s see the data  

[jabbr]

3 hours ago, Jud said:

 

Some of what's being done is selecting commodity components on the basis of different goals. 

 

Exactly. In some cases using a commodity board with low noise power supply. In other cases selecting from among commodity chips.

 

Quote

I'm sure something like the Sonore streamers use components they didn't design and fabricate themselves, other than designing the boards, but component selection was done with a view toward low noise rather than highest fps on a popular game.  Even artisanal companies (or individual DIYers) have the tools to design boards, and can have small runs fabricated by vendors.

 

Of course they do. I think the biggest value they provide is packaging a custom Linux that anyone can use so that you have a plug and play product. Like a power supply, you expect to plug it in and have it work.

 

 Designing for the lowest EMI can be very challenging. The talent in most audio firms tends to be analog, not high speed digital. I wouldn't underestimate the expertise that goes into a SuperMicro server board, or an Intel NIC, for example. I wouldn't describe either as "cheap, throwaway" or "error tolerant"...

 

 

Quote

 

Now (leaving out speed), whether they have circuit design tools as sophisticated as Apple or Intel, very possibly not, though they might not need them (the components don't need to fit in a phone or a chip).  And I'd imagine measurement equipment is a very real difference - not that small companies don't have and use measurement equipment, but companies like Apple and Intel can afford advanced and costly measurement equipment far beyond the means of artisans.

 

Intel undoubtedly uses its own in house proprietary tools in many cases ... who knows ... but lower noise translates into higher working clock rates ... likewise lower power draw translates into longer battery life so we shouldn't underestimate the resources that are devoted to both low noise and low power.

[jabbr]

1 hour ago, Jud said:

 

Hmm - @PeterSt, @jabbr, interesting, eh? 

 

Right. Logic circuits have the same issue, and when there is a mismatch between the incoming signal's rise time, and the ideal rise time of the receiver, ringing can occur. When you are stitching together ICs there's only so much you can do, and that's another reason why testing is necessary ... or perhaps some people like the sound of ringing logic switching in the morning ...

 

Now this entire thread presupposes that different digital circuits have different "SQ" ... ok so how could that be? The standard answer is "noise" and that leads folks toward assuming that if they use a "magic" LPSU that this will solve the "noise" problem in digital circuits. Nowadays there are many great chips that can be used, the "LT3045" is a currently popular one, but also very easy to create an excellent LPSU from a few transistors, resistors, diodes, capacitors etc... this isn't magic but ok let's do it. But realize that going from 1nV RMS to 1pV RMS does not make the SQ 1000x better (or perhaps it does ? )

 

Then we realize (hopefully) that using the magic LPSU doesn't magically transform our listening room into Carnagie Hall ... so hmm... "jitter", femtosecond clocks ...

 

Ok so lets assume we've totally licked both the voltage noise and phase noise problems and our digital signal is perfectly "clean" -- what are we missing? RF entering our circuits? Ok so we shield, and now we have no noise entering our circuit ... what are we missing? Logically at this point you may think you are perfect (and the SQ will be perfect ? ) but still the problem of intrinsic circuit noise remains, which means the logic has an intrinsic voltage and phase noise. So no matter how perfect your LPSU and clock are, you aren't directly addressing the circuit ... and you still hear the sweet sound of logic ringing in the morning.

[jabbr]

18 hours ago, Albrecht said:

“Now this entire thread presupposes that different digital circuits have different "SQ" ... ok so how could that be? ""

 

There is a big difference between the "digital circuit" and the complete, overall device. The power supply powers the digital circuit, - but it is not the digital circuit.

 

Substitute “device” for “circuit” if that helps you understand. I always consider the power supply as part of the circuit. The “device” can be subdivided into subdevices or assemblies. 

I’m using the term “circuit” in a common way that people who understand electronics would use the term. No need to quibble.

 

18 hours ago, Albrecht said:

I think that you are incorrect, - and it decidedly does not. Although both the SMS-200 and a MacBook Pro, and a Razberri Pi, are all computers, - they are all very DIFFERENT kinds of computers with different design goals. With more than a few calisthenics, you can get your Razberri Pi to run email, - but sheesh, - why bother?

 

Actually you are conflating a packaged product with both a digital circuit (which is what I was discussing) as well as a device — both the MacBook and Raspberry Pi are considered devices which the customer modifies with software. Actually most OS distributions including Mac OS, Windows and Linux shift default with email — specialized distributions (perhaps SMS-200) can remove. But what are you trying to say? Are you worried that the presence of an email reader which isn’t running but is on the device,  alters SQ? 

 

A “custom” server might have custom assemblies (eg PSU or motherboard). 

 

My post was about the actual digital circuit as delivered, not about the intention of the designer. As @marce says, the details of the board layout for CPUs/FPGAs is constrained by requirements.

[jabbr]

1 hour ago, Albrecht said:

Yes,

No need to quibble. But all I'm saying is that the layperson,- who audiophiles are, - wouldn't consider the two types to be the same even though they both are computers. Again, - a "computer" that has just 2 types of I/O ports is not thought of by the layperson as the same type of computer as a laptop. ...

The relevance to this thread is that it's a straw man to try to assert that the author is saying that all digital circuits suck, - when the author is in fact, - speaking of "packaged products," - like most everyone does. An external linear power supply plugged into an ultrarendu is certainly not considered part of the digital circuit by anyone.

 

I consider the power supply as part of the circuit. Perhaps I'm strange, who knows, but when you buy an amplifier, the contained power supply is part of the amplifier. 

 

You responded not to the OP, rather to my discussion of the ways that digital circuits generate noise. Perhaps you don't understand what I wrote, but if so, don't lecture me about circuits and devices and why there are different types of computers. I'm well aware.

 

My point wasn't that nothing is important, rather that one very important source of digital noise has been significantly neglected or not even dealt with. Details are actually important, and what I am saying is that there is a limit to the ability to reduce noise in a server if all you are doing is upgrading a clock to the single femtosecond accuracy, or reducing external power supply noise down to the picovolt range.

 

You missed the point of my post. its not that audiophile products aren't making an effort to reduce noise, its that they, by and large, use the very same chips as all other mainstream devices and these chips have a certain degree of intrinsic noise.

 

The fact that these chips have intrinsic noise levels, and generate a certain amount of EMI,  isn't me making this argument out of the blue. This is known in textbooks.

 

1 hour ago, Albrecht said:

Regenerating & reclocking the USB signal through just one USB port won't make the device NOT a computer, - but no multi-function computing manufacturer would ever consider doing such thing, - in fact, - they would eschew something like that as absurd. Just read the many trolls who consider themselves (so-called) computing engineers, - who attack high end audio designers, asserting that their designs don't work and they are bamboozling audiophiles.

 

No regenerating and reclocking and noise reduction are done ALL THE TIME in high performance high speed digital circuits. Read about the FPGAs of which diagrams were shown. Programming an FPGA is very very significantly about signal integrity, clock domains, rise times, propagation delays etc. These issues are not specific to "high end" audio, in fact I don't see any evidence i.e. specs & measurements that many of the >$2K so called high end streamers have gone to any extraordinary effort (and I mean that literally) to mitigate noise. 

[jabbr]

31 minutes ago, barrows said:

 

Just because you "don't see any evidence" does not indicate that high end streamers do not address the issues which you point out.  Unfortunately you appear to be posting as an "expert" (albeit you have not made such a claim yourself) but in reality you seem to have little to no actual knowledge of what is done to make high end Ethernet Renderers perform.

 

I don't claim to have actual knowledge of how your own products are designed or how they perform (in an electronic sense). By "evidence" I mean measurements. You know, the way network manufacturers supply eye diagrams, you might supply some type of objective measurement which indicates that, for example, your renderer has less RF/EMI production, or an eye diagram which demonstrates the jitter on the USB output.

 

Right, I don't see evidence. I made no claim to have trade secret knowledge, but if you have evidence then publish it! That's all.

 

Quote

While I agree, no one in high end audio is likely to ever have a custom ARM style processor chip built (as the cost could never be re-couped), all the other issues which you mention are indeed addressed in some of the better high end streamers.

Certainly what good high end Renderers address is not just power supplies and a good clock, as you seem to suspect, every detail is attended to in order to reduce noise levels and increase the integrity of the signal output: including paying attention to things you and @marce point out here like skilled high speed PCB design.

 

I didn't mean to imply that you didn't use skilled high speed PCB design, just that motherboard manufacturers such as SuperMicro, Gigabyte, ASUS and well Intel from time to time, have considerable resources that they direct on skilled high speed PCB design as well. 

 

Quote

 

While I agree with much of what you are saying when it comes to most of the custom (full featured) Servers out there, which indeed use OTS MoBos, etc,  This does not apply to (all) of the purpose built Ethernet Renderers.

 

I've never said the purpose built Renderers aren't well designed. As you know there is an entire and growing industry devoted to custom CPU/PCB designs and even custom power supply ICs and my understanding is that with many of the complex CPUs/FPGAs and with the high number of board layers, and given the very specific routing requirements, that the latest generation of board layout tools are almost required to get a reasonable design ( @marce can correct me if he still hand routes Virtex-7 designs  ) and so firms like Digilent/AVnet and even Element4 etc are increasingly offering core designs that can be modified. There is also an emerging industry of audiophile add/ons to the RaspBerry Pi (eg mezzanine boards, linear supplies etc) -- not that I'd start with that particular design but nonetheless ... in any case the various ways to design and layout a "custom" CPU board are not all that mysterious.

 

Let me also say that your own boards don't seem out of line for what you are offering because there's a price for design and manufacturing and some of the parts I'm aware you've used are good values e.g. the Crystek clocks, likewise with Uptone and the LT304x regulators (or whatever similar). People also want packaged products and there is value in that... well until y'all can get the "Geek Squad" to come set up your linux boxes ... hmm ...

Really I'm more inclined to caution folks not to spend $$$ on outboard clocks or very expensive power supplies to power a PC or renderer because the PC/renderer itself has an intrinsic noise/jitter that an asymptotically  better clock/psu can't fix.

 

So no, I'd say that @marce is an actual expert, not me --- he has decades of experience, but y'know I've done some stuff and done some stuff and I know what is often needed for very low noise designs.  I think his main point is that high speed digital design is high speed digital design, and low noise, low EMI is low noise EMI (that can be both low voltage noise and low phase noise). So yeah if you have actual evidence that this isn't the case, I'd be happy to see.

[jabbr]

1 minute ago, crenca said:

 

 

What @jabbr said @barrows, where is the what, when, and how of this "every detail is attended to"?  You complain of folks just making assertions "without actual knowledge", and then you simply assert that "every detail is attended to".  Show us the money...

 

Yeah let me echo ... if "every detail is attended to" show me the EMI/RF output pattern at the microRendo USB compared with some of your other devices ... e.g the Solid-Run derived ... that would be just one of "every detail" that I'd attend to. and show an EMI pattern on the board, etc ... ok, so you don't have it and realistically no one attends to every detail, and most good engineers who build commercially available SBCs attend to many details.

[jabbr]

Hey also let's not get off track here ... the thread is about dedicated servers vs laptops ... I do in fact expect that a high powered server with GPU would spew out high levels of EMI from its ports and every other orifice ... perhaps not too high to pass regs but who knows, and high enough to matter...

 

My advocacy all along has been to isolate the server with a network, and consequently use a low powered and consequently low noise network render.

 

I did say that I doubted that high end audiophile renderers go to any extraordinary measures to mitigate noise. I consider good board layout, good power supply design and good clocking to be an ordinary measure. I mean, yeah, if I were paying >$2K, I'd really expect these techniques to be used. By ordinary I mean textbook, like if you pick up a textbook they tell you to do this stuff --- common knowledge that clocks need good power. Textbook.  Let's see, yeah a GaN custom ARM chip would be extraordinary  

[jabbr]

This is also extraordinary: https://www.nature.com/articles/s41566-018-0236-y .... hmmm wonder the SQ? Jitter?

 

and if you think DSD512 is the ultimate: https://www.lightwaveonline.com/articles/2014/03/semtech-announces-ultra-high-speed-adc-dac-cores.html

https://www.fujitsu.com/downloads/MICRO/fme/documentation/c60.pdf

 

and if you think I was joking about GaN processors ... https://phys.org/news/2017-12-gallium-nitride-processornext-generation-technology-space.html

[jabbr]

7 hours ago, marce said:

how do you know its less noise it may be more noise with a LPS

Good point ... some people tend to learn somethings 30 years ago and fixate on them forever as if gospel.

 

Best to continuously reevaluate your dogma. There are many smart young whippersnappers as Analog Devices etc who have developed clever new chips.

 

The current technique of placing a highly responsive regulator such as LT3045 following a DC/DC converter has merit. 

 

Keywords don’t equate to SQ.

[jabbr]

13 hours ago, sandyk said:

 

How about you give a well respected E.E. like John Swenson a bit of credit ?

John normally provides measurements wherever possible.

 If you bother checking the Uptone area of the forum you may even find appropriate measurements.

Alex, rather than invoke someone else’s name in this debate, because John & UpTone both use/ship combinations of SMPS / linear regulators, you can make the argument yourself and rather than handwave that someone can find some measurements somewhere, please link to a measurement that supports your assertion. In many cases you may be misstating someone else’s position, or out of context. 

[jabbr]

19 minutes ago, sandyk said:

 

 It's annoying  to see certain members continually rejecting huge numbers of reports by other members as purely anecdotal, and demanding measurements.

 

"Everybody knows" ... nah ...

Quote

This forum is NOT a sub branch of  an engineering forum, it is an Audiophile forum where members should be able to post their impressions and freely exchange ideas without such continuous demands. 

 

Seriously? No one is interfering with your ability to post your impressions. Your classic argument is either "everybody knows" or "some other well respected guy agrees with me" ... uh ok, perhaps that's worth saying one or twice but don't place that in your signature...

 

You aren't merely claiming to be an audiophile, rather you continuously discuss engineering topics e.g. which IC regulator is best, which type of PSU is best, which circuit is best. Yeah in engineering circles there are established ways to settle disagreements so let's see some beef and stop whining that "continuous demands" are being placed. Those are normal ways to help settle debates.

 

Quote

 Of course, if measurements are also readily available, then that is a bonus.

 

You said they readily available, so let's see?

 

also understand that what you consider "proof" of something isn't what everyone, or even most people, would consider "proof". If you aren't being believed, then you have to put forth a more convincing argument, rather than repeating the same argument over and over.

 

"Everybody knows" and "huge numbers of reports" is called: 

Argumentum ad nauseam

[jabbr]

30 minutes ago, sandyk said:

The continuing demands aren't mainly aimed at me, they are aimed at the vast majority of members ( including yourself) that post subjective impressions due to technical based changes.

 

I don't see these demands of the vast majority of members, nor myself. 

 

I have no issue with someone asking for a measurement if I make an assertion. I don't prefer to publish measurements of other peoples work nor do I publish comparative listening tests lightly. 

 

I try to give practical advice and if I make a claim try to support it with a reference. I don't ask anyone to take my word alone for anything and I don't ask anyone to take my listening impression as gospel. Generally when I publish a listening impression, its just that I liked something, or perhaps wasn't impressed.

 

Also, while we all have technologies that we tend to be fond of, as do I and I've said so, and while I've never had the pleasure of listening to a Bruno Putzey's creation, there is much admiration and I tend to think that a well done "SMPS" could compare to a well done "LPS" just as well done Class D could compete with well done Class A ... that said, if you give me a few transistors and diodes and capacitors I can knock out a great LPS and similar for Class A, so perhaps its easier to get good results with LPS ... that said I'm not into dogma.

[jabbr]

16 minutes ago, mansr said:

I see, the Latin needs to mixed up with an equal amount of Greek.

Ooooohhhh! I will use that!

[jabbr]

43 minutes ago, sandyk said:

 The main problem with most generic SMPS in many types of consumer equipment is shown in the attached diagram.

 

Right, the common theme we have been discussing... you are getting into engineering though (which is fine) ... "meeting the relevant FCC emission specifications" because specifically what?  

[jabbr]

14 hours ago, sandyk said:

I presume that you are also talking about things like the permissible level of leakage currents etc. going back into the A.C. mains from the inverter , double insulation etc. ? 

Actually some SMPS place capacitors to reduce EMI so tradeoff. Presumably the caps in the schematic you showed?

 

Dont compare $5 SMPS with $500 LPS ... not fair. Let’s use equal cost... there are medical grade SMPS with lower leakage currents and lower EMI ... and higher cost (a bit) ...

 

Just trying to keep it real here ...

[jabbr]

14 hours ago, One and a half said:

My employer's laptop, one of those rolled out, used to give me tingles on my wrists while typing. I thought it was the material of the case being magnesium alloy touching the skin and irritating it, until one day got jack of hurting and measured the case to me and found 95V AC. I dare not measure the current with me in circuit.

The leakage voltage across the primary and secondary caused by the X class caps was enough. The information on the web, regarded the leakage currents for these two prong power supplies to be of a low value and unimportant. Pfft. The case measured 95V to earth, hello should be 0! The two prong power DC power supply was changed for an earthed power supply and all was good again, no more tingles.

 

So clearly that product would not meet permissible leakage current levels  

[jabbr]

42 minutes ago, sandyk said:

Yes, the caps in the schematic serve that purpose, as well as being the cause of the high impedance A.C. leakage.(tingle )

 ...

My E.E. friend has even designed such a

 

Perhaps you should ask your friend exactly what the predicted impedance would be and report back to us whether you consider that “high” or “low”  

[jabbr]

4 minutes ago, sandyk said:

Jon

 That was quite a few years ago, and I doubt that he still has that info available .

 He retired recently too, as too much downsizing in this area of employment without moving living locations again.

 Presently he is constructing a shack for his electronics.

"  I've finished doing all the cabling to the house now, with six runs of gel-filled cat6 cable and 3 phase power, all buried in a 600mm trench in separate conduits."

Ok here you go, you too can tell us ... https://www.allaboutcircuits.com/tools/capacitor-impedance-calculator/

[jabbr]

35 minutes ago, sandyk said:

 Jon

 I have better things to do than play these silly games. I know the capacitive reactance and inductance formulas without needing to use an online calculator. I also have charts that I can get quick answers with when I am lazy.

 

@JohnSwenson introduced the term "high impedance leakage" to us which he states is approx 300-400 MegaOhm ... that equates to a capacitance of about 2 pF at 1kHz. Lets assume the capacitor in your schematic is ??? you posted it so tell us. 

 

Now assuming you place your audio section behind a Topaz type isolation transformer e.g. 0.0001 - 0.0005 pF interwinding capacitance, and realizing that leakage currents would need to flow from the server PC we are discussing here, through this transformer (which isolates the audio ground), and then plugging that value into the above equation, you get an impedance of 30 or so giga ohms ... and now explain to me why we are concerned with the SMPS in the PC power supply? Surely not leakage current getting into the isolated DAC ... so let's keep this discussion on topic, you have access to these basic formulas like everyone and claim to understand how to use them, so really, why are we discussing this???

 

Not everyone takes what I've recommended at face value as the way to isolate an audio system, so here's a bit of math which supports why I do what I do: everything outside the audio system area is isolated from everything inside the audio system by 30 giga ohms or so depending on the quality of your iso transformer.

 

This isn't a "silly game", its really fairly basic math ... and everyone has access to these formulas