A novel way to massively improve the SQ of computer audio streaming

[marce]

15 hours ago, JohnSwenson said:

Grounding the output opens the possibility of a good old fashioned ground loop (magnetic induction from hot wires to ground wire causing voltage difference between branches of the electrical wiring or long distances on the same circuit). But not grounding lets the high impedance leakage pass through to the output.

 

Nobody else even seems to be looking at high impedance leakage, I didn't even know about it until two months ago. It is so high an impedance that normal test equipment shorts it to ground and you never see it. The problem is that its effects ARE showing up in audio systems (not all and not in the same way in all systems, of course it can't be anything simple). I had to build my own test equipment to even see it.

 

A good part of the problem is that almost all the original work on leakage currents took place many years before the advent of SMPS so they were just looking at linear power supplies, which don't have the high impedance leakage. Its a fairly recent issue, significantly exacerbated by computers and computer networks involved in our audio systems. (since these almost always use SMPS)

 

John S.

Class 1 or class 2 SMPS's or on both?

[marce]

Slide 10

https://www.et.byu.edu/sites/default/files/electrical_grounding_0.pdf

Double isolated electrical equipment should not have any connection to the PE (protective earth), I would recommend checking your local regulations. Though when connected to class 1 equipment it is often effectively grounded. The PE is connected to the negative, so grounding double isolated can cause a shock via the added ground connections, whereas when un-grounded you have to be between the hot and neutral line to get a shock.

Just a thought on safety, worth checking with someone who is more au-fait with the regs.

[marce]

4 hours ago, Johnseye said:

Individual chips, at least a minimum should be shielded.  It isn't limited to high powered transmitters.  Here's some reading on the subject.

 

Were you referring to integrated circuits and not interconnects?  Electromagnetic Interference and Digital Circuits: An Initial Study of Clock Networks goes into some detail about integrated circuits.  This is actually some pretty interesting stuff.  Solid evidence and pertinent to digital audio applications.

 

Understanding and Eliminating EMI in Microcontroller Applications goes into some detail.

 

This PowerPoint discusses RFI in audio, but it's a ppt so doesn't flow like a paper.

 

And another paper.  From that paper:

 

"Conductive enclosures can be used to shield sensitive circuits from the effects of these external fields. These materials present an impedance mismatch to the incident interference, because the impedance of the shield is lower than the wave impedance of the incident field. The effectiveness of the conductive shield depends on two things: First is the loss due to the reflection of the incident wave off the shielding material. Second is the loss due to the absorption of the transmitted wave within the shielding material. The amount of reflection loss depends upon the type of interference and its wave impedance. The amount of absorption loss, however, is independent of the type of interference. It is the same for near- and far-field radiation, as well as for electric or magnetic fields."

 

So, how effective are the chassis of our equipment in shielding outside noise?  What material is it made of?  How effective are they at keeping that noise in?  Is covering integrated circuits enough or are there other other places EMI/RF can cause an issue?  Can you cover every IC with the 3M shield?

 

 

More from that paper:

 

"Thus for high-frequency interference signals, lightweight, easily worked high conductivity materials such as copper or aluminum can provide adequate shielding. At low frequencies however, both reflection and absorption loss to magnetic fields is low. It is thus very difficult to shield circuits from low-frequency magnetic fields. In these applications, high-permeability materials that exhibit low-reluctance provide the best protection. These low-reluctance materials provide a magnetic shunt path that diverts the magnetic field away from the protected circuit. Examples of high-permeability materials are steel and mu-metal. To summarize the characteristics of metallic materials commonly used for shielded purposes: Use high conductivity metals for HF interference, and high permeability metals for LF interference. A properly shielded enclosure is very effective at preventing external interference from disrupting its contents as well as confining any internally-generated interference. However, in the real world, openings in the shield are often required to accommodate adjustment knobs, switches, connectors, or to provide ventilation. Unfortunately, these openings may compromise shielding effectiveness by providing paths for high-frequency interference to enter the instrument."

 

But best of all, I can without a doubt hear a difference.  That to me is really all that's necessary, but there appears to be some proven science behind it.

This is standard everyday stuff in electronic product design...

Look at the books and notes by the likes of Keith Armstrong (UK EMC club, Cherrytree consultants) practical guides and notes on both EMC testing and how to avoid EMC problems.

Henry Ott, the main man. 

Ralf Morrison, another good source of theory especially regarding the waves of electricity.

SIgnal Integrity and EMC comparability engineering has been studied extensively for many years, I have collected GBytes of related documentation and books over the years and as we keep increasing the signal rise times and clock frequencies the more problematic layout and system design becomes. The kit required for audio reproduction is pretty common place, how well some of it is made in terms of EMC would take an EMC lab to determine, you need a shielded environment to determine the system noise, not the environments background noise. 

Often do stuff with a requirement for immunity up to and greater than 18GHz and currently working on a hand held medical device with a blue tooth trasmitter (4 board stack, 15mm x 20mm two processors, memory, sensors, RF antenna all in close proximity, plastic shielded case). Its all down to the layout mainly, get that wrong and as the papers pointed out you have multiple antennas (there are numerous papers on antenna structures on PCBs, dipoles for dummies, Henry Ott is a good start. Its down to good design and testing, the problem is EMC testing is expensive, how many small firms do full comparability testing?

[marce]

21 minutes ago, Summit said:

 

 

 and painted circuit boards, 

 

What is a painted circuit board...

[marce]

Printed circuit boards are the most critical part of any electronic assembly, they hold the components, transmit the signals in a controlled manner, handle the heat of today's modern bottom terminated components, and for digital and analogue are critical to it working, without them you would not have today's electronics.

Ah re-read it, yes I have to concur that some motherboards can scrimp on layers, but still have to work, probably not as well as say bespoke server boards (12-16 layers compared to 8 for a basic PC (and in some extreme cases 6 layer, though these are becoming rare as smaller and smaller BGA components are used.)

 

[marce]

1 hour ago, Summit said:

switching regulation is maybe more important than a reasonable good switching ATX PSU?! Placing all components like processor, ram, hard drive, mother board, switching regulators, clocks, inputs/outputs very close together means EMI and RFI will infect each other more than in a full- size box there they can be placed further apart.

 

All the interfaces, especially memory have maximum defined lengths.

Actually having the components closer together is better for signal integrity and EMC, you are minimising loop areas, which with digital signals is critical due to the wide band of frequencies involved. Being closer you can use a weaker drive strength for the the signal, slower rise times, again limiting the amount of energy in a circuit and thus again reducing noise and other coupling mechanisms (cross-talk). The shorter trace lengths means they are less likely to act as antennas and if they do it will be at a much higher frequency. Power planes will be smaller, again putting any resonances higher up and due to the smaller size again reducing coupling mechanisms (in more critical designs, close coupled power ground planes with special dialectics are often used to increase the power plane to ground capacitance, a critical element of the initial switching current supply of any high speed digital circuit). We go to great lengths on every design to place the digital (and analogue) components in the smallest space possible... Obviously you separate functional circuitry to avoid interference between say the DDR memory and your USB ports, where space dictates the use of shielding cans around circuitry...

[marce]

1 hour ago, Johnseye said:

 

It isn't low power as much as clean power, meaning as little noise as possible.  Low power helps achieve that.  I think clean power could have a bigger impact than many components.

 

 

This is a good question.  I don't know what challenges there are in building a higher power PSU.  This would be a good question for @Superdad

12kW @ 230V or 35kV @ 20ma

More focus on the complete power delivery system would yield far better results. With digital its the power available at a devices pins when it switches that,s critical, not whether the front end is a linear or a switched, though either must regulate well and be free of noise and filtered.

[marce]

Snap

[marce]

38 minutes ago, Johnseye said:

 

 

So how can we determine when a motherboard or an external component has a high quality power delivery system?

At the design, simulation (if done) and testing phases of the design cycle. All I can say is look at the price of some motherboards... even allowing for volume production costs there has to be some compromise, a generic PC only has to work as a generic PC, warcraft, a bit of browsing it does its job... a laptop, the only function of a lap top is to allow you to work on trains, anyone who uses them as a main system is... Look at 3Ms EMC foam etc catalogue, its all about reducing laptop noise...Even I have heard a difference using headphones between my laptop and my main system (i7 loads of ram, loads of disk space).

My own view would be a compact system, one box, but more similar to enlarged mobile phone design/satellite electronics design, minimal distance between the sections, but each section isolated to the highest degree for both conducted noise and radiated noise, The best choice of PCB would be an impedance controlled flexi ridig design, so you have total control of a signals propagation from transmitter to receiver and the added advantage of no added connectors and the parasitic interference they can add. Not a popular idea as it does limit the after sales market, a little. 

[marce]

20 minutes ago, rickca said:

For motherboards, I find reviews on Anandtech a pretty good source of information.  The motherboard industry is so competitive that I think price is indicative of both features and design/build quality.   

 

As an example, here are some details about the ASRock Z370 Taichi

 

- 12 phase power design

- high density glass fabric PCB

- Nichicon 12K black caps

- aluminum alloy heatsink

- premium 60A power choke

 

 

Means nothing...

Most PCB's are glass fabric, the ultimate build is the critical thing... Or are they referring to 7268 weave, which compared to the more generically used 1080 has a higher glass density, thus lower resin content so a lower overall dielectric constant, and a more even impedance path for the signal (that travels in the dielectric) has to follow.

Ref:

https://www.altera.com/content/dam/altera-www/global/en_US/pdfs/literature/an/an528.pdf

Page 5 for the various glass weaves.

[marce]

23 minutes ago, mansr said:

In many SoCs, drive strength and other parameters are configurable for each pin or group of pins.

Yep, the IBIS models that come with the devices have different drive strengths and using something similar to Aldec active HDL or similar the whole design can be configured for the correct signal drive strength and thus lowest noise, costly though, in design time, simulation and testing... Often quicker to fix on the fly, with a scope and re-configuring the device.

[marce]

16 hours ago, lmitche said:

"Means nothing" not exactly the type of reply that is going to make friends and influence people.  Time to pull your head out of your rear.

It does mean nothing, nothing to do with my head up my A***, considering the not insubstantial mark-up over a stock Netgear router, I would at least appreciate some relevant information... Hence why I also provided a link to PCB materials to back up my comment...

Now if I was going to mod my netgear PSU I and sell it as an audiophile version I would probably use power modules from Vicor, combined with active EMC modules from Picor (or similar)

http://www.vicorpower.com/

This would give me the advantage of being able to feed 12 or 24V input into these modules from either a linear supply, a wall wart or a battery. If I used a linear supply I would have it in a separate box near the mains, with a DC umbilical, the best way to reduce the transformer magnetic coupling, also the noise can be kept separate in its own little loop. So you provide good clean power and allow the user to tailor his main power to what he believes is best for his system, Similar to what Uptone have done with the Regen etc. And you can claim Mil spec power supplies...

Clocks, not a fan of messing around with clocks,  they are critical to digital systems and have to be considered at the design stage... From my own experience I know how hard it can be to distribute a clock a few mm around a PCB, moving the clock off-board is something I can not comment or reccomend. I cant comment on it because without some serious kit, you just do not know what is going on, and the faster the clock the more tetchy it is to added parasitic's, longer cables add more by definition. That how do you determine the drive strength for the clock, the output of the crystal has to be buffered by a line driver... How do we determine how much energy to put into this, not knowing its exact parameters... How critical is this to signal integrity, very, let me illustrate:

Here is a pretty slow clock going about 8.5mm down a PCB, because of the drive strength there is a lot of ripple on the line, this caused some problems later on (Jtag clock being distributed by a clock distribution device, cant remember which one). Now stadard engineering rule of thumb says just slap a 22r resistor in the line, as shown this did not provide the attenuation needed for the line. After some simulation and measurements (shown) the final choice was an 82R resistor, the simulation showed a clean line, the residual ripple is caused by the scope loading. As is clearly shown, the series resistor controls the energy fed into the line, it also slows the rise time, limiting the amount of upper frequency harmonics and thus energy that is m ore likely to cause ripples. That,s the trouble just one measly clock can cause, when you get on to spread spectrum channel hopping radio communications (very low jitter required) then it becomes a nightmare. SO whilst I understand the value of a good clock, I cannot personally see how you can transmit a clock down and add on wire and get the claimed improvements in the jitter spectrum, to many unknowns, both in terms of parasitic influence, impedance mismatches and most critical "return path discontinuity"

Here is some info on return path discontinuity, it shows that even if the ends of the wire are just a couple of mm apart it will have a big effect on the signal. From Eric Bogatin's excellent site, be the signal.

http://bethesignal.com/wp/wp-content/uploads/2014/01/PPT-240_Return_Path_Issues_-MGB_Interconnects.pdf

 

https://www.bethesignal.com/bogatin/

 

On mobo and sCLK-EX server I must look into it more before I can comment. Currently trying to decide which way to go my-self , and trying to formulate a basic system becomes more remote the more I look into the choices. Starting of at a very basic level though, just putting mains runs into my new house, so the main IT for the house is on its own feed, I have a separate feed for audio to my room... The first choice is do I keep all the main digital together and limit it to just a DAC on the audio feed... So a lot to think about.

Have fun, I'm off to slave pulling wires and installing floor temp sensors!

 

[marce]

The latter part of this video shows the effects on clock signals of return paths...

 

[marce]

5 hours ago, ElviaCaprice said:

Not looking for a lecture on clocks.  We already know empirically that the sCLK-EX clock replacement improves audio performance immensely.  Keeping an eye out for better mobos to perform better once the clocks have been replaced.  Probably handle the power within the mobo better for audio purposes is what I would be on the lookout for.  Smaller is better, 12V DC input.  Ability to have or add one PCIe lane, actually 2 would be better. 

Just thought, @marce, with all your technical experience would have a thought on such a mobo.  Didn't hurt to ask.  Sometimes you can overthink the process.  Practicality comes into play here.

Just a bit of practical information, so sorry, would you rather I just be an objective with a negative attitude to all this, or one who joins in the fun and puts his own views and a little technical knowledge I have picked up to help develop a greater understanding of what is happening...

You don't know empirically though...

You can't win...

[marce]

1 hour ago, Johnseye said:

 

This is helpful in understanding the importance of short leads and short runs on a board.  Evidence of why a smaller board would be better.  I can see its implications when using the sCLK-EX.  As it's an external board we're using leads between our mobo, switch, tx-USBultra, tx-USBexp, etc and the sCLK.  This video shows the impact those leads can have on the signal, and our SQ.  Evidence as to why we've been discussing using the shortest leads as possible.

 

As I'm getting my devices modified with the sCLK-EX this has me thinking whether I want to use the straight sCLK-EX board inside my PC case so it can be closest to the motherboard and tx-USBexp PCI card.  It will have a longer run to the switch if I do it that way which is less important.  I'm now wondering if I use a tx-USBultra for the sCLK, I can somehow install that inside my PC case so the runs can be shorter.  The Streacom case is small so I don't think this can be done unless I remove it from the tx-USBultra case.  I will ask SOtM if this can be done.  The bottom line question is how close can I keep the tx-USBultra to the PC, minimizing the lead lengths.

 

Another thought came to me from this video.  @JohnSwenson has been writing about grounding the negative of the SMPS DC barrel.  I suspect the length of the ground wire used in doing this has an impact.

The length of the cable, can be catered for, its more important to minimise the return path discontinuity, ie the ground needs to be next to the clock, withing a mm if possible.

When I get chance to look at my EMC stuff I have some data on what is happening with what JS has suggested, trouble is my memory is not up to scratch at the moment, so much stress, I can't remember where the stuff is, so I am going through my documents. Briefly it has the effect of standard practice in instrumentation etc. of having the ground pin connected to the chassis where a signal enters a unit. What interest me is that JS's testing has shown that on some units this is not happening, so the high frequency noise can get into the equipment. But that is for another day...

[marce]

1 hour ago, ElviaCaprice said:

No, I don't mind a bit of practical information, but you we're asked a question and instead made posts that seemed to be more of a lecture than helpful in answering what was asked.   If you do not know, just say so, no problem.  Yes, we do know through empirical evidence within this thread.   Anyways continue on, I'll not ask again or question your posts.  Anything new is a breath of fresh air.

Sorry I did answer your question in my rather long rambling post, as I said i am not currently qualified to answer that question directly, my post was a rather rambling explanation why.

Please do question my posts... very often a differential view leads to new lines of discovery, we should always question a post, especially one that is so objectively based.

[marce]

Well, that was interesting.

I will bow out myself as I don't want to get in any bun fights.

One thing though, a couple of you are using the word empirical incorrectly, you don't want any science so realise that you have NO empirical data, it is subjective opinions only...

Have fun.

[marce]

Pathetic and childish...

Does your attitude move the audio hobby forward, no, 

proof of nothing, I could have put up a one line reply and achieved what you accuse me of, instead I spent more time than I have to spare carefully formulating my reply and hopefully adding to the collective understanding... And especially to avoid simple mistakes that will bugger up your super clocks, but ...

All you have done is pull people down and act like a bit of a bully... 

 

[marce]

Sorry for getting tetchy but as I said if I wanted discredit, it is easier to do with less words. I have major differences in beliefs between myself and those of a more subjective nature, by the same token I have major disagreements with my missus regarding furniture, decor even the time of day... That said we can still discuss our viewpoints and each take a little extra away.

[marce]

On 05/11/2017 at 12:20 PM, ElviaCaprice said:

Not looking for a lecture on clocks.  We already know empirically that the sCLK-EX clock replacement improves audio performance immensely.  Keeping an eye out for better mobos to perform better once the clocks have been replaced.  Probably handle the power within the mobo better for audio purposes is what I would be on the lookout for.  Smaller is better, 12V DC input.  Ability to have or add one PCIe lane, actually 2 would be better. 

Just thought, @marce, with all your technical experience would have a thought on such a mobo.  Didn't hurt to ask.  Sometimes you can overthink the process.  Practicality comes into play here.

!

[marce]

As I pointed out in my previous post, you have to measure a clock line and pick a termination that matches the line, distributing high speed clocks is not that straight forward if you want to maximise signal integrity and in your cases achieve minimum jitter... Using a generic line and resistor will go some way, but especially for the higher frequencies is tailoring any termination to the full line, which does require some direct observation of the signal.

Clocks can be distributed using a clock distribution amplifier, it does not need to be a $1000 device, there are chip distribution IC's.

Daisy chaining can be horrendous to get right, again direct observation and simulation is the best option if you want to be sure the clock that is at your device pins still has low jitter...

 

[marce]

A scope will usually suffice, that is the general tool for looking at waveforms...

Generally adding a resistor to a line will swamp the lines mismatches making for good signal transfer, when you get into the high frequencies though things can get a bit fun... The illustration I put up was a 16MHz clock going 8.5mm, just to show what can happen to clocks.

 

[marce]

1 hour ago, Johnseye said:

We've broached EMC/EMI/RF concepts in this thread before.  After experiencing positive results from shielding with the 3M sheets I started looking further into this.  Roy was talking with me about a conversation he had with Rob Watts who recommended using ferrite on the USB cable.  This isn't uncommon with computer based USB, display, mouse, etc cables.  I did some digging about ferrite in the audio realm and came across this article.  It's in depth, but at a minimum have a look at the summary and conclusion points.  I've pasted the 4 common sources here.  There's solid evidence behind this.  I'm going to begin adding ferrite cores to my cables and will share my opinions on audible impact if any.

 

http://www.audiosystemsgroup.com/AESPaperFerritesASGWeb.pdf

 

1. An audio cable will be excited as an antenna by radio signals that surround it, causing current to flow along its length. Most of the current will flow on the shield, but the current can also induce a common mode voltage on the signal conductor(s).

 

2. Improper termination of cable shields within equipment (the Pin 1 problem) injects RF shield current directly into the equipment, [6] where it is detected by several well known mechanisms. [2]

 

3. Imperfect inductive coupling between conductors of a shielded twisted-pair cable converts shield current to a differential voltage on the signal pair (SCIN).

 

4. Inadequate low-pass filtering of signal input and outputs lets RF present on the signal conductors into equipment. Equipment can be sensitive to both differential mode voltage (between the signal conductors) and common mode voltage (an equal voltage on both signal conductors).

Nice to see ferrite's get some good press in the audio world, for high frequency noise reduction they are a useful tool, I have posted some pi layouts with a distinct moat somewhere. Will dig out some notes later, also Murata's got some good catalogues worth looking at. Mainly SMD devices but some good info.

Mains cables (and others) can be looped through the ferrite more than once, this does increase the effect of the ferrite.

 

 

[marce]

11 hours ago, jean-michel6 said:

Hi

I have also experiment with ferrite . However you need to be careful on which cable you use them.

All current are electromagnetic waves as well as Emi-rfi that we are trying to get rid of. 

Ferrite is effective with positive results ( from an audio perspective ) on digital cables ( USB, lan,...  ) on power cords. 

On interconnect , it has a negative effect as it just kills the sound making it softer and less engaging. 

One very effective way to shield all cables from emi-rfi is the use of mumetal , with two layers to passively cover cable. 

One cable company in france HiFi cable is using this  technology on all their top line cable with very good results. 

I have treated my digital cables with very good results. 

The main difficulty is to find mumetal for this at a reasonable cost. 

Ferrites are frequency dependent in their effect... generally in the MHz

[marce]

This may be of interest...

https://www.murata.com/~/media/webrenewal/products/emc/emifil/knowhow/26to30.ashx

 

Looking at the ferrite stuff, there is a whole range including some low frequency ones deigned for SMPS's with a noticeable effect from 150kHz - 30MHz perfect for the output.

Trying to find some figures for the high frequency ferrites (100MHz-500MHz) all the graphs I have found so far only go down to 1MHz.