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

[nbpf]

On 1/27/2018 at 7:26 PM, austinpop said:

 

That would be interesting, but I would amend it to SE chain vs. an sCLK-EX modded PC chain.

 

BTW, the SE is a 25-lb monster, so it's not something you throw into your carry-on bag.

Dimensions		70 x 420 x 320 mm (H x W x D)
Weight		11 Kg

 

So any comparisons would have to be within driving distance.

I'm wandering why Innuos do not offer a lighter version with no optical drive and perhaps with a smaller size. A driveless version would likely also simplify the design of the power supply and would be appealing to users that only need a high quality server. 

[nbpf]

14 hours ago, austinpop said:

 

Yes agree it’s not for everyone. I just wanted to give context for my results.

Thanks for providing the details about the power supply and, of course, for your original contribution!

 

I would be very interested in learning more about the potential advantages of using power regenerators like the PS Audio P5 as a means for improving poor AC power supplies.

 

This use case does not get very much discussed and the common wisdom seems to be that the first step towards a clean power supply should be laying down one or two dedicated AC lines.

 

Unfortunately, this is not always easy to achieve: I am currently renting an old and rather large flat with very old (and likely poor) wiring. Laying down a new dedicated AC line would be a mess in my case.

 

I have been wandering whether there are established means for assessing the quality of a power supply and whether a P5 or a P10 would be a meaningful  way of achieving a high quality power supply without breaking the bank.  

[nbpf]

1 hour ago, afrancois said:

I don't own a P5 or P10, but as far as I've read you will not always benefit from a re generator. Much depends on the quality of the AC that enters your home. ...

 

 

Do we know some reliable method for assessing the quality of a given AC power supply? Regenerators tend to be both expensive and heavy. It would be nice if one could measure or estimate how good (or how bad) a power supply is (perhaps at different points in time) before embarking in testing heavy and expensive gear.

[nbpf]

34 minutes ago, Solstice380 said:

 

An AC regenerator will ALWAYS improve the power delivery to your components.   Delivery includes lower impedance as well as clean.  

I guess the question is by how much? Can you elaborate a little bit on your experience with AC regenerators? What was your initial setup? Which regenerators have you tried or added to your initial setup? Thanks, nbpf

[nbpf]

1 hour ago, austinpop said:

Even though I own a P5 regenerator, I will shy away from making blanket statements. As I have reported, I do find it makes a positive impact in my system. Before I got my dedicated line, it sounded best with everything plugged into the P5. After my dedicated circuit, I found the dynamics improved by plugging my amp directly to the dedicated circuit, while leaving the rest of the digital chain on the P5.

 

I have not personally tried all the approaches - and there are many! This (AC power conditioning) seems to be a space with strong, conflicting opinions by the "experts."

 

Paul at PS Audio swears by the regenerator approach. Caelin over at Shunyata swears by their Denali. Audioquest has their Niagara. Synergistic has their Powercell. Jim Weil at Sound Application has his super-expensive power-factor-correcting conditioners. Then there's the DIY crowd here on CA that swear by the use of repurposed industrial, extremely low capacitance isolation transformers like the Topaz, with clean power strips with no filters. Of course, these things can sometimes hum like a substation in your listening room.

 

It's enough to make your head spin!

 

The one constant that no one usually disputes is a dedicated circuit. I would say if at all possible, start there. Nothing fancy, just 10 gauge Romex, and a decent power outlet. Get an electrician out to take a look. Ask around for referrals for one that has done "audio" dedicated circuits, so they won't just argue with you about how you don't need 10 gauge wire, blah blah. Also, don't rule out the idea of running outdoor-rated wire on the outside in a (PVC) conduit, if internal access is too disruptive.

 

As usual, Roy @romaz has tried more of these options than I have, and you can read his views here: https://www.head-fi.org/threads/chord-electronics-blu-mk-2-the-official-thread.831343/page-196#post-14103671

Thanks, I understand that a dedicated circuit should probably be the first step to do. In my specific case this is problematic, I have already check it with an electrician. Independently of my specific situation, I think that it would be nice to have a way of measuring or at least estimating how good or how bad an AC line actually is without having to lay down a dedicated alternative and perform listening tests or demo a very heavy and expensive regenerator!

[nbpf]

5 minutes ago, austinpop said:

... But you missed my point about dedicated lines. You don't need a test to determine if you "need" a dedicated line. They're almost always a good idea because you're replacing thin (14 gauge typically), old wiring with a heavy gauge with very low impedance. Remember - it's more about the gauge than it is about  being "dedicated."

You are right, I missed the point. On the other hand, since laying down a new line would be very elaborate and/or expensive in my case, I would like to know that I have reasonable chances of succeeding before I embark in the enterprise. Also, according to  http://www.msbtechnology.com/faq/how-to-wire-your-house-for-good-power/, it seems that it would be wise to lay down two lines, one for the amplification and one for the rest: again more options ...  

[nbpf]

@BigGuy, @austinpop, @afrancois, thanks for your replies, I very much appreciate your feedback! Best, nbpf 

[nbpf]

9 hours ago, Evo-No-Revo said:

I read a couple of reviews about the http://www.lab12.gr/gordian.

 

The 6Moons review was partially in reference to the Kii Three speakers I now have.  That is what got my attention in the first place.

 

http://www.6moons.com/audioreviews2/lab12_2/1.html

 

http://www.monoandstereo.com/2016/07/lab12-gordian-power-multifunctional.html

This is from the Mono And Stereo review for a quick description,

 

"The LAB12 Gordian power is a multi-fuction power distributor/conditioner that is a very cleverly constructed from the beginning and showed a great sonic potential. According to the developer, Stratos Vichos, it is "basically a unit that is designed to clean up your mains supply without affecting the dynamics of the system" - which I can fully confirm.  Gordian incorporates an industrial grade power analyzer which lets you inspect all the aspects of the AC power signal. Among other things it provides data about:

 

• Frequency (accuracy 0.0005%)

• Total Harmonic Distortion (THD - accuracy 0.005%)

• RMS Voltage/Current

• Power Consumption

• Power Factor

• FFT (Frequency analysis)

• DC Voltage

 

I see the price is around 1400.00 Euro.  That comes out to a little over 1700.00 USD.  I am not sure of the exact price for US though.. 

I just now noticed that they are actually available from Audio Archon.   I just purchased a Bricasti M5 from from Mike and it should be here tomorrow. 

I will have to ask Mike about the Gordian and get his take.

 

If this can clean the AC lines and keep the dynamics, the analyzer can be a very interesting tool/toy.

 

I thought about waiting to get the Gordian after getting my new systems setup.  But after reading the reviews again, I think that I will go ahead and order sooner than later.  It could be very useful in testing the affects of different additions to the AC lines.

 

Evo-No-Revo, the Gordian really looks very interesting. Many thanks for posting and have fun with your new M5! Best, nbpf

[nbpf]

On 22.3.2018 at 1:35 AM, Solstice380 said:

 

Wow, I was traveling today and there were 15 or so more posts since then!

 

I’ve found the PS regenerators helpful in my system / location.   We tend to have poor quality power - up to 4% THD and definite power factor issues.   So time of day and phase of the moon would change the SQ. 

 

The dedicated circuit wasn’t feasible at the time but I did make that move eventually.   Ted, you can have many runs but see which type panel you have and keep them on the same phase.  And put the fridge and microwave etc on the other!

 

i started with a Chang Lightspeed and then a Transparent Ref

conditioner.   I then tried a PS Power Plant Premier and ended up with 3 of them.  @austinpop mentioned not putting your amp on an UNDERSIZED regenerator.  I added the undersized caveat. Despite the specs for current delivery etc that are great with these things, I think they over rate them.  I now have a P10 for the front end a P10 for the amp only.  The P5 isnt even big enough for a front end with very many components.  I wouldn’t put

more than a few hundred watts load on it for it to perform its best.  The PPP and P10 have the same 1500 watt rating but the P10 is fine for my amp where the PPP wasn’t.  There is a reason why PS has introduced the P20!  I’ve also heard a solid demo on the big Stromtank unit out of

Germany.   Similar idea but wicked expensive and probably better than the PS.  Also, PS is now using DSD to generate the signal and not PCM.  ?  When it comes to

power, more is always better.  Oversized PSUs in components are usually better for SQ.  Heat be damned. 

 

I snagged a Topaz off eBay to try.  750VA 0.0005 cap -31 model and put my whole front end on it in place of the P10.  Hundreds vs thousands of $ is worth a try.  It collapsed the soundstage noticeably.  No question.  The newer PS units have some measurement and graphical capabilities so I plugged it into the Belkin power strip that I hardwired into the Topaz to see what I could see.  The THD was a little lower than the wall outlet and the waveform was just as bad.  Flat topped, skewed sine wave.  I dedicated the Topaz to my HQPlayer server and LPSU powered drives.  My NAAs are powered by Uptone JS-2s into a PS regenerator

 

I now also find find no difference with or without any DC neg grounding.  We do it in our industrial equipment because we never know what type of power environment it will live in, and factories are baaaaaad.  I plugged in a system once and watched the cable start to melt from the outlet being wired incorrectly at the panel!  We make a “transceiver” for 5 Hz - 5 MHz incorporating both a DDCDAC and an ADC so we also use a pretty serious Powervar power conditioner.  But, I think by now we all know about good grounding and power.   

 

So for my system, which consists of generally well engineered components, the benefits are there. Better soundstage, longer decays.  Does everybody’s system let you hear the ambience after a piece is over and before they cut off the recording?  Like the sheet

music fluttering and the whum whum whum of the HVAC. 

 

Maybe we should start a thread on Regenerators but I won’t do it, I’m convinced they are good in my system.  P10s are almost always available at US$2999.  P5s are almost never on sale and I can’t figure out why.  Maybe they cost to much to make?  Wouldn’t be the first product to suffer that fate.  I don’t know if they even make the P3 any more.  Maybe you could

power a headphone setup with it.  

 

Apologies in advance for typos, run on or incomplete sentences, etc.  I hate phone typing!

 

Thanks Solstice 380, I very much appreciated your post!

[nbpf]

On 22.3.2018 at 1:16 PM, flkin said:

 

I use the Alphalab Power Line Meter https://www.alphalabinc.com/content/power-line-meter/

 

to measure total noise in my AC line. It's in milli-volts. 

 

 

My AC is pretty dirty especially in the evenings as I live in a condo and the device measures 840mV of total noise.

 

 

 

 

After passing the AC through my Topaz 91001-21, 1KW isolation transformer the noise drops to 23mV so it's doing its job.

 

It's interesting to see how appliances around the house affect the power line noise like air conditioning and lighting. The reading changes during the time of day also.

 

The Topaz doesn't repair bad wave forms and other bad wave shapes but is very effective removing DC noise in the power line.

I'll have to get one of those power line meters, thanks for posting! Best, nbpf.

[nbpf]

12 minutes ago, austinpop said:

...
So the working conjectures to explain Roy's findings are (let's not even call them hypotheses):

• eMMC over SSD
• Intel CPUs over ARM
• in-memory "Dream" OS (Linux based)
• SR-7 DR rail
• something about the NUC mobo.

It should be easy to support/confute conjecture 4 (SR-7 DR rail) by replacing the SR-7 with another PSU. Perhaps it would also be useful to check whether loading the OS in RAM makes a difference or not. In the latter case the kind of storage (eMMC vs SSD) might again become significant.   

[nbpf]

I am currently running some tests (using `cyclictest -l1000000 -m -Sp90 -i200 -h400 -q`) to get latency measures for my minimal Raspbian distribution that I am using to run MinimServer and upmpdcli on a (RPi 3B+ & DigiOne Signature) server + renderer solution. 

 

For meaningful comparison with the results shown on http://www.audio-linux.com: are the tests meant to be run on a an idling system or on a system that is replaying some music?

 

Thanks, nbpf

[nbpf]

Here are the results of `cyclictest -l100000000 -m -Sp90 -i200 -h400 -q` for the dedicated Raspbian system that hosts my DigiOne Signature. The system runs on a  RPi 3B+ with cpu frequency set at constant 1.2GHz.

 

The kernel is the standard Linux rpi6 4.14.71-v7+ #1145 SMP Fri Sep 21 15:38:35 BST 2018 armv7l GNU/Linux with only a few optimization options. Are these results good? Are they bad? Are what one would expect?

 

 

 

 

[nbpf]

1 hour ago, misterspense said:

What are the experiences with the original Intel NUCs with respect to the noise of the fan? I'm following this thread and contemplating buying a NUC, but am not keen on a fan equiped computer in my livingroom. I know there are fanless cases available aftermarket, like Akasa, or fanless mini pc's from other brands.

Do these Celeron NUCs produce some noise? Or are they virtually inaudible?

I have been using fitPC products (http://www.fit-pc.com/web/) and they are well built and fanless. I do not particularly like NUC cases and I find the black, ripped fitPC devices aesthetically more appealing. They have a new, modular series of devices called fitlet2 with, among others, Intel® Celeron® Processor J3455 CPUs.

[nbpf]

2 hours ago, misterspense said:

What are the experiences with the original Intel NUCs with respect to the noise of the fan? I'm following this thread and contemplating buying a NUC, but am not keen on a fan equiped computer in my livingroom. I know there are fanless cases available aftermarket, like Akasa, or fanless mini pc's from other brands.

Do these Celeron NUCs produce some noise? Or are they virtually inaudible?

Depending on what you want to do, a Raspberry Pi might be a valuable alternative to more powerful machines. With a real-time kernel and the usual optimizations (headless, no X, no desktops, no HDMI, no bluetooth, etc.) it exhibits very good latency measures and is very easy to setup and maintain.

[nbpf]

To those who have recently experimented with low-latency or real-time systems: would it be possible to summarize your findings? Do we have an idea of which measures of latency have a more significant impact (assuming that they do have an impact) on sound quality? Minimum latency, average latency, maximum latency or perhaps something else? On which systems?

 

I'll try to make the first step: on a Raspberry Pi 3B+ & DigiOne Signature running MinimServer and upmpdcli, I have tested the standard kernel (4.14.71-v7+) against a real-time kernel (4.14.74-rt44-v7+) with the same optimizations: no bluetooth and HDMI output, fixed CPU frequency (1200MHz), internal audio off. The differences between standard and real-time kernel can be summarized as follows:

 

standard:

# Total: 100000000 100000000 099999995 099999949
# Min Latencies: 00004 00004 00004 00004
# Avg Latencies: 00006 00006 00007 00006
# Max Latencies: 00335 00278 00283 00269

 

real-time:

# Total: 100000000 099999949 099999824 099999700
# Min Latencies: 00006 00005 00006 00005
# Avg Latencies: 00009 00009 00008 00008
# Max Latencies: 00091 00059 00052 00058

 

The corresponding latency plots are

 

 

, respectively. The results clearly show that the real-time kernel does very effectively limit the maximal latency to values well below 100ms. This is done at the expense of the average latency that increases from about 6ms to 8.5ms. The minimum latency also slightly increases with the real-time kernel.

 

These result are consistent with the many reported at https://www.osadl.org/Latency-plot-of-system-in-rack-1-slot.qa-latencyplot-r1s1.0.html. From the point of view of sound quality, I have not been able so far to perceive any difference between the two kernels.

 

[nbpf]

4 hours ago, diecaster said:

I am not looking for a science experiment...just some sort of logical hypothesis as to how latency could matter if the DAC’s buffer never becomes empty. Is no proponent of this low latency concept able to be an apologist for it?

Good question! As reported, I have not noticed any obvious difference in sound quality between the two kernels. That said, there is more than one logically plausible explanation of how latency could affect the sound quality of a DAC. First, the rate at which a DAC buffer is consumed is not constant. The old Naim DAC, for instance, does select an internal clock (according to which the buffer is consumed) on the fly and on the basis of the speed at which the buffer is filled in. If this rate varies because of latency, it is conceivable that latency can have a direct impact on clock selection and hence on sound quality. Another, perhaps less direct way of impacting sound quality could be through electrical or electromagnetic noise. We should assume that, under conditions that grant bit-perfect data transmission, the stream of bits that fills in a DAC's buffer is independent of the choice of S/PDIF cable, USB to S/PDIF bridge, upstream computer, etc. Still, S/PDIF cables and USB to S/PDIF interfaces are known to have an impact on the sound quality in spite of filling a DAC's buffer with the same bit stream at more or less the same speed. This suggests that, in spite of filtering and isolation measures, electrical noise and/or electromagnetic radiation can have an impact on sound quality. If this is true, then it is not surprising that latency can indirectly impact the sound quality: the results that I have posted suggest that the average latency increases by about 50% when one replaces the standard kernel with the real-time kernel! Moreover, real-time kernels tend to consume more resources and we know that electric noise and electromagnetic radiation do depend, among others, on the power consumption. These are all hypotheses that need to be supported or confuted by empirical experiments. But they seem logically possible, at least to me.      

[nbpf]

1 minute ago, diecaster said:

 

Okay. Let me get this straight. People here are finding that loading the OS into the RAM of a NUC resulting in lower latency improves sound quality on a DAC when using a USB and Ethernet or just one or the other?

I am not sure that everyone who has tried to reduce latency measures has noticed improvements in the sound quality.

 

The aim of my post was indeed to see if, after about 15 pages of reports on experiments with latency measures, it was possible to come up with some consistent observations or findings. Perhaps just a summary the observations and some preliminary conclusions. 

 

I have only played around with Raspberry Pi devices and, as reported, I have not observed obvious differences in sound quality when running the standard kernel and a real time kernel. This might be due, among others, to the fact that my system is not particular revealing or to my old ears.

 

It should probably also be noticed that, on the Raspberry Pi,  the real time kernel decreases the max. latency by a factor between 3 and 10 but at the same time increases the average latency by about 50%!  Thus, real time does not necessarily imply low latency, at least not in average and on weak processors. This is something that I felt had been completely missed in the discussion.

 

[nbpf]

As an alternative to NUC devices, you might also consider fitPC units. They are well built, fanless and modular. The low-power fitlet2 can work with power supplies from 9V to 36V DC. It seems like a good proposition for running a network transport or a lightweight server + renderer solution.    

[nbpf]

7 hours ago, austinpop said:

 

Believe it or not, this is what most of us want very much. A DAC that is 100% immune to the upstream chain would save people a lot of money and effort. All the experimentation you read here comes from the fact that current DACs do not achieve this ideal - even very expensive DACs.

It would be very easy to achieve 100% isolation from upstream LAN sources (albeit not from internet sources) if users would accept some delays: just equip DACs with sizeable local memories.

 

If one requires the replay of data that are in the local memory, no data exchange is needed and replay can start immediately. If the track or album to be replayed is not in the local memory, first fetch it and then start replay.

 

With this approach, a manufacturer could focus on optimizing replay from (known) local memory instead of having to fight noise that is inevitably system dependent and largely unknown.

 

I have been advocating this simple solution in many occasions but I am not actually sure that so many users actually want a 100% source agnostic DAC: we seem to like fiddling with cables, optimizing sources , cleaning up streams ...   

[nbpf]

7 hours ago, the_doc735 said:

....but surely all that toxic rubbish is inherent & intrinsic to the electronics, hardware & software? Yes reduce to a minimum level! BUT surely 100% clean is impossible; like you say: "even very expensive DAC's don't achieve this". I would say it is impossible to totally eradicate the toxicity?

It is both possible, easy and actually very cheap if a user accepts that the replay of a track or of an album only starts after that track or album has been completely transferred to the DAC's local memory. It is just a matter of waiting a little bit, like when we were using a CD player and we had to wait for the tray to close for replay to start.

[nbpf]

15 minutes ago, sandyk said:

 

 Then try and explain why the same bit identical tracks saved in different storage media, can still sound different when played from System Memory !

What shall I try? Why should I explain what? What have your requests to do with my post?

[nbpf]

15 minutes ago, diecaster said:

The DAC needs to be a close to a real-time device. It can't take a lot of time buffer full tracks. Imagine picking tracks to play randomly and having to wait a very long time to hear the first note every time you pick a new track. Not very practical. Nor is it really necessary. A 1 second buffer is an eternity to a computer.

Shutting down data transfer at replay time might be unpractical but it is logically necessary if one wants to achieve replay that is by construction completely independent of data transfer! I fully agree that there are ways of nearly achieving this goal without compromising responsiveness.

[nbpf]

49 minutes ago, sandyk said:

Try looking at the images I have supplied in My Profile . The comparison videos that the images came from were even located in the same folder of the USB memory stick.

I have looked at the pictures but, to be honest, I do not understand which consequences we can draw from them. I am not criticizing you experiments, just trying to understand what the results actually are and what they could possibly imply.

 

Thus,  just to mention one example, it looks as if the first picture was spoiled by camera shake while the second one looks much better. Without further information about the f-stops, the shutter times and the stabilization mechanisms used, it is very difficult to draw any reliable conclusion, at least for me.  Perhaps I am missing something obvious.

[nbpf]

2 minutes ago, sandyk said:

I will take this to a PM to avoid derailing the thread.

Great, thanks!