My Digital Audio Journey (So Far)

Hi everyone, This is my first post on SBAF. I hope the experienced people go easy on me. As such, I don't have any measuring gear, and all of these are done only with my ears and some correlation to existing measurements + Some math knowledge. While I do correlate some of my audible experiences with measurements from audiosciencereview, innerfidelity etc, I don't fully believe all of them and have lots of disagreements.

If I have to tell my audio gear journey as a whole, it'll take ages and I am not going to bother with that since it is barely of use to anyone. However, I find that quite a lot of people have not yet found enlightenment with regards to digital audio, and hence am sharing my subjective experiences and "predictions" of how it went for me.

Like most general people, I started off with my laptop output (surface book). It used to be fairly clean on full sized headphones but has some hiss with sensitive stuff. One thing I noticed often was how "electrostatic" how the output of the surface was. Every headphone I plugged in sounded similar to two of the staxes I heard - L300 and SR009. I had few other devices at the time - cayin n3, a few phones - iphone 4s, iphone 6, oneplus 3, a few lumias but the surface was the one with such a unique tone. The cayin was equal in resolution and power but different in tone and texture (some mild dip around 5-9khz, a mild boost in the bass and around 10Khz). I never got around to why the surface sounded that way until after I bought the nx4 dsd. http://archimago.blogspot.com/2014/09/measurements-microsoft-surface-pro-3.html . I realized that it was the low passing that made it sound like estat texture. After getting NX4 DSD, I simulated a low pass in matlab and I confirmed my doubts. http://www.learningaboutelectronics.com/Articles/Square-to-sine-wave-converter-circuit.php . And modified the r and c values. Makes sense since electrostats are capacitors. NX4 introduced me to "dynamics". Full Initial Impressions here: https://www.reddit.com/r/headphones...ew_there_was_a_difference_but_i_didnt_expect/
It didn't sound louder per se, but the impact, detail and just overall sense of surprise was absolutely mind blowing for me at that time - "mountains" from Interstellar made my hairs stand up. Little did I know how big of a rabbit hole I'd get into.

For long I had been using Microsoft's Groove as the music player. A friend of mine shared me this song - "Back to business" from the Violet Evergarden OST and asked me to check for faint thumps during 0.25 to 0.45. This was the first time I was A-B ing music players.

There's a faint ring of the bell at 0.25. There's another instrument for a split second during 0.28 and 0.29 (comes fades and comes. Plays two notes). Regarding the faint kicks at just before 0.35, just before 0.37, twice just after 0.38 and once just before 0.40. I was also able to hear something similar around 1.30 region and especially at 1.36. The more interesting inference was, above timestamps were wrt foobar2000. On groove player the timestamps switch by like 1 second earlier (0.25 shifts to 0.24 and so on). But that's not where it stops. On foobar 2000 I could replay from 0.25 seconds how many times I want and I'd be able to figure out that thump. But in groove player, I need to play it from beginning again by opening the file again. If I try to replay it back from like 30 seconds, all the above details are lost. Realized that groove had some super sucky algorithm and buffer stack. (The recording used is : Back in Business (track no 13), from VIOLET EVERGARDEN: Automemories. Year - 2018. More details. Dynamic range as measured by dr tool for foobar 2000 - 10. Left and right both 9.6. peak left is -1.48 and peak right is -0.99. RMS left is -14.3 RMS right is -13.6). That was the last time I used mircososft's groove player.

Then I started playing around with buffer settings in foobar. Damn it each buffer setting in foobar sounded different. The short buffers sounded fairly solid to me while the long buffers sounded softened/rolled off. Guess what, I had been running it driver-less all the time, and I guess it went to isosynchronous mode. I thought the drivers come packed in to windows since I was getting sound and the device didn't come with any manual etc. Went to the site, downloaded the driver and installed. Instant improvement. I couldn't point it to one thing. It sounded just a lot more "raw" for lack of a better word. It sounded more dynamic but also less harsh. Switched from directsound to ASIO, and it felt even better. The buffer changes no longer made any sound difference. Any buffer I set sounded the same.

Over time I realized the shortcomings of NX4 DSD - high frequency noise, grain, low frequency cut off (some coupling capacitor etc), not enough current push in high gain. I don't have measurements to back it up but you can see a few of these even on "resistor load" measurements on audiosciencereview. I bought the apogee groove, and never looked back on the nx4. The differences were exemplary (dynamic range, realism of sound etc). I don't want to delve deep into measurements or subjective impressions of groove, but I can guarantee that it sounds wildly better. It also powers high impedance high sensitivity cans with ease (HD800), but for low sensitivity stuff its not up to snuff (Planars, my SRH1540, HD425). Has something called current feedback which shapes the current instead of voltage at output. I think I can hear the effect on HD800 (which typically sounds artificially wide, but not on groove), but can't confirm. I currently use groove + hd800 directly. Or groove (-8db on its control panel to give 2v line out, pushes 5V at 100%) -> Burson Fun with V6 Vivid opamps -> Shure SRH1540. The latter is what I typically use since I find it more resolving (I find the midrange of HD800 to be a little off).

I used foobar for a while until curiosity struck. I downloaded in installed winyl, another music player. Oh god, difference again. This time I couldn't understand which is better. Both sounded different. Foobar was softened. Winyl was harsh. There were a few new details in echo when using winyl, which I was unsure if winyl was artificially adding or was it present in the recording. But the differences are super profound and I was sure I could ab-x them 10/10.

Notes from my Listening session:
Comparing Winyl and Foobar2000 with ASIO by Peter. (ASIO by Case clips at 100% for some reason). Asio foobar fluid - low pass? winyl sounds closer - v shape? or reverb? At most only one of the above two is right.

https://winyl-player.github.io/ (link for the player, it is free)

I tried checking which of them was better for a couple of weeks, till my supra USB cable came. I bought it since it was cheap and honestly had not been expecting much difference out of it. Guess what, Difference again! This setup now established winyl as the reference for me, and I also came across a post at headfi which measured these two players in asio/wasapi and found winyl to be more accurate, and foobar to show harmonic aberrations. Since the post was older, I requested a friend of mine to measure it (he measured only directsound of both) and it showed similar changes so confirmed that nothing has changed all this while. He measured it on different versions of foobar. I honestly don't know why the cable sounded better, but its not noise. If I can guess, it is polarity consistency. I feel like the normal cables feel like I constantly am often bombarded with more positive polarity signal than negative polarity ones.

I want to stress that this difference in this specific measured parameter (-130db added harmonic) is certainly not audible and I was pretty sure that this was not what I was hearing. But it did give me confidence to know that there is something happening that sets both these players apart. I believed that there was a single major phenomenon that was causing multiple changes, one of which was this measured distortion level. I am still trying to figure out ways to measure that.

While on the supra I feel like it sound more like the real thing - positive and negative as per what is there in the recording, or how it sounds in real life. My prediction is 2 ways - 1. since USB audio doesn't support error correction cables might actually be able to destroy the sonic information either by glitch between d+ and d- lines or if their impedance is not up to match with the USB spec, could cause issues with latching (metastability etc). 2. The same as above, but if Phy Layer supports error correction, if the data is incorrect, it will push more stress in phy layer triggering electrical noise int he rest of the circuit. I can't ascertain to the above since I don't have measuring gear, and I'd be highly obliged if anyone can recommend me a way to analyse this. Pretty much all that is claimed on supra audio's page is that they comply to the USB standards properly, and no magic buzzword.

In the meantime I also got the geek out 450 as for a reference that most people are accustomed with but honestly it felt super underwhelming. It was give and take vs the nx4 dsd - nx4 dsd was noisy, harsh but dynamic. Geek out was clean and extended but compress (always loud). Against my apogee groove, it was ladders and floors below, just like the NX4 was. I could actually simulate geek out on my apogee groove, which we'll get to in the following passages. I also tried the two filters in geek out and while they sounded slightly different, it was marginal in the grand scheme of things.

I went around forum after forum trying to understand the full sequence of what happens when it comes to digital audio transfer - The mixers, audio libraries, kernels, buffer, etc. Few blogs I found to be enlightening.
http://insanecoding.blogspot.com/2009/06/state-of-sound-in-linux-not-so-sorry.html
https://www.electronicdesign.com/te...ticle/21801786/achieving-bitperfect-usb-audio

You can read about the above when you have time, but tldr is isosynchronous over USB is inherently lossy since it averages and discards packages. 125us per transfer and 8000 frames per transfer. Asynchronous is the way to go but it has its own timing restrictions. It uses 1ms poll time with tonnes of buffer (defined between the dac and host during initial communication) and the clock is now determined by the dac. In async, the slave (dac) sends a request for data and it expects data to reach its end within specific time from that. Host needs to serve the slave as fast as possible. Hence file system, error correction etc are very important since it'll reduce the amount of cache request attempts due to reduced number of data transfer failures hence reducing the time of overall transfer and getting it more in sync. Kinda asks it at the last moment and says, specifies a very short deadline to meet and says, you give it or not, I don't care. I'm moving on. It's like a professor who gives you an assignment and says you gotta complete it in 2 hours and submit. Now you're gonna be scrambling out to do it, might skip things etc. some may not even submit. But some, might have predicted this and might have prepared something in already. If the professor gives you a little more headroom and splits it into multiple assignments each with specific buffer it is the safe mode. It is important to note that the asynchronous data happens through an isosynchronous pipeline. Phew!! Also there is no error detection, as I mentioned above.

Something about Libraries, kernels etc that I mentioned above: To have an in depth look, it'd take quite a few pages and a few courses to explain every stage in a computer operating system, and I personally only have limited knowledge on it. So I'll cut short to the most relevant points. Basically, an operating system acts as an interface between you and the internal logics/hardware. To perform every action - be it addition, transfer of data or comparison, etc. you need a sequence of instructions to be given to the hardware and the instructions typically contain an operand, an operation and an address or it will take the stack pointer. These are very basic instructions and modern processors support more complex instructions and theres more complex things like SISD/SIMD (eg AVX, SSE). For a personal computer, you might be having multiple tasks at once - reading a web page, downloading content, playing music, watching videos etc. And there will also be other tasks to monitor battery etc. Now all of these tasks have sub tasks like when you want to read a web page, it will have to decode the info, render the image and pass the data to the display controller. An operating system schedules this task in an optimal way so that you don't visually perceive any lag - CPU is very fast. Now the way these sub tasks work is that OS will have handlers linked to the hardware through pieces of codes known as drivers. Basically these drivers interact with the hardware to initiate the task. So, if CPU is very fast and it can be scheduled well enough, we're good right. Except that we're not. NOTHING IS INSTANTANEOUS, EVERYTHING TAKES ITS FAIR SHARE OF TIME. Now the thing is, there are certain tasks that are order sensitive and certain tasks that are not order sensitive. If I want to send a file it doesn't matter what order I send as long as I say where the packet is to be reconstructed at. But in audio and video, which are real-time applications, you need to get the data in proper order. There will be a different protocol used for the same.

I did some testing on winyl to figure out what happens on a software level. Did some test in putting the os to sleep and bringing it back. It resumes playback without error, the time runs, scrubbing works but doesn't make sound. Even if I switch to another song it doesn't make sound but doesn't throw error either. I need to restart winyl to get sound. I did this on directsound, firing data to my laptop speakers straight. I didn't want too many variables, though it should repeat on other stuff, in an even more complicated manner. I think the way winyl sets up page tables is very rigid and hence very stable with respect to data integrity but needs exact same address locations. Need to check how memory is utilized/changed when putting to sleep state.

This is what happens to the system when you close the lid. after 10 secs it goes to sleep. but it does enter two stages the first 10 secs. 3-4 secs = active. next 3-4 seconds inactive but not sleep (almost like a freeze+safety delta buffer) the moment you open the lid, the playback starts in an instant. After 10 secs, or sleep, the ram has gone a complete change.
Resuming playing from winyl after sleep doesn't result in an error dialog but it doesn't produce sound output. I need to close and reopen winyl for it to send out an useful data. Someting with regards to output reference or memory sampling reference or paging reference which winyl looks for is now corrupted/shifted.

A shorter crisper explanation of the whole thing in digital domain.
1. Winyl uses BASS audio library - audio library is what handles system calls etc and the instruction sets, sequences, memory handling. BASS library is probably the best for playback.
2. ASIO is probably the best playback kernel for windows. Developed by Steinberg. Linux has alsa which I'm not sold on yet, but I'm trying to get OSS working. OSS is apparently superior to asio and i can believe it. It's kernel doesn't have too much junk. But I need to try it out to confirm and also the fact that usb audio in OSS is iffy makes it harder.

I think we could measure a lot of these even by running some latency monitor application in the background or by using a tool like wireshark. But what I'm going to tell now is something that is quite impossible to measure using any tool other than a really high precision mic and good test tones (not sine sweeps) or an I2S logger that plugs into the Slave side to check data in the receiving end. I came across this tool called minorityclean. It is free.
http://www.mics.ne.jp/~coolverse/MinorityClean/

The tool promises two things - removal of jitter in the CPU level, and some wierd stuff that he calls electronic bit reference points for memory access or whatever. I'm able to understand what he means by the second one (I worked in VLSI physical Design for a while) but I don't know how he is able to mitigate it neither can I attest for its audibility since what happens inside the companies stays inside the companies (I really don't know what logic is used inside to compensate for such issues though there are a lot of protection logic to make sure such stuff don't occur). But the first one I can understand and I find it audible in my setup. It's quite a simple thing. If you have repeating instructions, the x86 CPUs introduce jitter in an attempt to parallelize it (pre fetch etc). The tool just makes an alternate compiler that spaces these repeating instructions and eliminate this instruction level jitter. Also uses 64 bit registers instead of 16 bit registers to make sure all the data for a particular sample is finished in one shot.

An intuitive example of what I mean by instruction level jitter (Please note that this is not the code used in the program).

mov rax, 1 can cause jitter since it has to fetch 1 from memory.
xor rax, rax inc rax is mostly jitterless since the entire thing is inside the processor accumulators/registers, assuming the instructions are cached for the processor.

Improvements I hear with Winyl (ASIO) + Minorityclean (version 8 rev 23): Lower Midrange/Wetness of sound (not necessarily bassier or compressed), Sustain Detail and Echoes, Sibilants, Spatial Panning accuracy.
Problems I hear with minorityclean : not sure if it is accuracy, but it sounds a bit slow, and maybe a tad warmer but I don't think so since it feels better for the most part. For sure there is a difference, with and without minorityclean that I'm confident.

Recently I tried a few other players, and winyl still remains the most transparent player I've tried yet. It has two friends - xmplay, and HQplayer both of which sound identical to winyl under identical settings. Also tried Jriver. It's fine but not as good as winyl/hqplayer/xmplay. But definitely gives a tougher fight on me when trying to compare. Foobar vs my reference players were profound. Jriver wasn't as profound. Overall jriver sounded like my apogee groove was taken to sound like geek out a little, a bit compressed and loud.

I also played around with HQplayer for a while and some notes on the player:

1. When bit depth and sample rate are matched, winyl and hqplayer sound identical. If music is 16 bit, Hq player at 16 bit and winyl sound same. Hq player at 24 bit for 16 bit input music without proper filter sounds worse. Making bit depth higher on hqplayer without doing any processing makes it sound a little worse. Somewhat soft. I am unsure of the reason for this but it might have something to do with how the groove handles dither on its side.

2. Making bit depth higher and increasing sample rate to 192khz (max supported by my unit) and adding processing makes it sound "different and interesting". Sinc-M and Sinc-M + couple of dithers worked fine for me. The mp (the graphs without pre ring) made it sound somewhat closer to geek out but since it retained dynamics it didn't sound as full/dark/compressed as geek out. Maybe if i added a compressor on top it could sound identical to geek out. Either way you can get the same filter on groove as you do on geek out with this processing. Most filters worked fine. Very few (the first one I told sinc-mp- long) didn't work well enough. I think they hold the most accurate rendition and a little sad I can't explore it now.

3. Most 16 bit music in Winyl/HQ player at song's default processing have a sense of harshness to them with the setup I have. (remember the harshness I told when comparing to foobar?). I am not sure, but I think I found an explanation for it. (https://www.stereophile.com/content/apogee-electronics-groove-da-headphone-amplifier-measurements) - take a look at the dithered vs undithered 16 bit/24 bit sine measurements (one of them is annotated wrongly 24 bit annotated as 16 bit). You can see that 16 bit data gets a little clipped. But I am unsure since the level is of the order of -130db again. Could be something else as well - the type of recording etc, but I have done my homework and I find that 16 bit pretty much always has this issue. I can process it up using HQ player (see point 2) and could get it to sound less harsh without smearing detail. No issue of loss of accuracy in doing that since we are looking at points beyond resolution of input sound, and it's more about achieving sweetspot in missing data here.

All that said, I prefer winyl because I find keeping things stock has an element of accuracy I don't find otherwise. 24 bit music with good dynamic range is pretty much plug and play and sound great. Winyl + Minorityclean is on a whole another level. I'm still trying to experiment with different variants of minorityclean but the default version 8 brought marked improvements on most fronts. He now has a rev24 that I'm yet to try. Hqplayer + minorityclean could sound identical to the above you don't upsample and don't add any processing but the point of hqplayer is those features. No processing on hqplayer even comes close to what minorityclean does. Overall hqplayer is really nice, more like a feature packed variant of winyl, but minority clean is just overall a higher fidelity medium.

If you could wait for minutes before playing music try the other software from the same guy - pinkHQ, nontallion,etc.. they do a system cooldown and voltage referencing (the parts I don't understand) and set the memory pointers in most optimal jitter free locations. I'm not bothering much with them because i tried it once and it took 15 minutes to start playing one song and I was really bored. Didn't try to find differences/improvements since i was super frustrated at that point. But I remember that minority clean had noticeable changes till maybe 15 minutes from start (can't say if it is real or placebo, but there seem to be other people who have heard similar).

Also, gave a shot at macbook recently. Put up with it for over an hour and gave up. Was kinda horrible. Returned back to windows. Using macbook, felt legit like a hipster trying to re-learn everything that used to be intuitive otherwise. Was disappointing since I liked FreeBSD when I tried it on my virtual machine. Seems like it's not freeBSD based. Just some 20% of the code used for networking were based on BSD. Couldn't get usb audio to work properly. It made sound and everything, but didn't sound right. Felt like snakes and ladders where i try my best to climb to stage 90 and the snake pulls me down to stage 3 again - sounded similar to how windows media player sounded, if even worse. For an os that prides itself to be a professional environment I expect it to just work no excuses. I'll give it a better shot sometime soon and see if something was broken in the setup I used.

Chain used in comparison:
Windows pc (surface book 2015 8gb ram 256gb SSD gt940m GPU) -> supra usb cable -> apogee groove -> burson fun + V6 vivid opamps -> shure srh 1540.
I also used Sennheiser hd800 and urbanite xl and got same results. But I find that the shures are more resolving overall so I reference it when I post. Other gear used are - HD700, Geek out, and a dozen other negligible headphones.

 

Hopefully someone can guide me on how to go about measuring these differences since it for sure is audible in my setup and I have a few friends who have confirmed this in their setup as well. All I have now are focusrite scarlett 2i2, a couple of presonus prm1 mics and a hantek 6022BE oscilloscope and none of them have particularly great resolution. Hantek can't even measure audio band signals properly (bit depth not enough I guess).

My only scopes at this point of time are latency monitors and wireshark. Wireshark will be destructive testing minorityclean and can't be used for that analysis.

Also a word of warning, be careful if you use winyl for playing back uncompressed wav files through asio configured with large buffer. I often got some sharp white noise, either due to buffer underrun or overrun. It doesn't happen with mp3, flac etc, only with wav files.

 

It did anyway. TL/DR

 

Oops, thanks for reminding.

tl,dr : a random guy's journey through digital audio. For usb dacs kindly install drivers and make sure it is running in usb asynchronous audio mode. The music players ranking in terms of fidelity imo - winyl/xmplay/hqplayer > Jriver >> foobar2000 = windows media player >>> windows groove player >>>> android and it's it's resampling screw up.

 

Putting it kindly...

4305 words.

Well, we do ask people to introduce themselves, so...what can we say? I have a certain amount of awe that you wrote so much, and I think it is just a bit sad that tl;dr is the answer you are going to get. From me too.

Like... Introduce yourself, yes. But nobody is going to read 4,000 words, which is huge for any forum post anywhere, without knowing you a bit first. Internet forums is one of the things that has reduced my attention span (but at least, not to Twitter levels) and if I am going to read a book, or even a long article, I'm going to have to know that I really want to first.

But hey, thanks for the effort. And welcome.

 

Thank you and my sincere apologies. The post has personal experiences but also technical stuff relating to filter realization of mechanical circuits. I thought people from EE background or Signal processing background might find it interesting. It is something that is supposed to be 4-5 separate posts but I went ahead and made it a single post. Realized now that I should keep things constrained.

I didn't actually write all these just for this post. I typically make a lot of notes (since I like playing around mix and match) and I just "organised and made them (relatively) concise".

 

I made an effort but my eyes glazed over a little bit in

Kidding aside, you do seem to know your way around design if your posts so far are any indication (actually enjoyed your post in the Etymōtic thread, it was educational to me), which is more than I can say for a layperson like myself. Don't mind the dislikes, if you aren't actively being evil or happen to be a really bad fit then people will eventually get used to you and the walls of text.

Well, maybe not the massive walls of text.

There's an introduction page: https://superbestaudiofriends.org/index.php?threads/new-members-introduce-yourself.17/

And some talk about measurement systems here: https://superbestaudiofriends.org/index.php?forums/measurement-setups-systems-and-standards.25/

The forum's grown large since I signed up, and I've not even been here for very long! There's no real shortage of stuff to take in if your Insight's high enough; I spend my free time occasionally trying to digest posts here that are well above my pay grade and social sciences intellect, haha.

 

I need to get myself one of these apogee groves, I find it very hard to believe you can tell the difference with that 2 in 1 of different instruction sets and that the differences are so large, larger than some reported amp changes.

I'm not one to blind A/B but you should try it, might shake things up, the power of suggestion is strong.

Next:
-clean power
-I2S
-ditch USB, listen to the CPU directy.

 

I have and I didn't.

No offense, but it seems to be mostly finding excuses to explain your perceived experiences.
Any modern PC OS will have thousands of threads running at any given time. They go through at least one scheduler you can't control before being handed to the CPU and there they'll be sliced and diced, pipelined, speculatively executed out-of-order... And this is not counting the BIOS.
I'm sure you'll be aware that such time sensitive control of execution on a multi-threaded system over a multi-purpose CPU is a holy grail of real time computing and, as such, would be a very expensive technology if it worked!
Also, where did you find the explanation of what MinorityClean is supposed to do? What you state does not match what its author says.

In any case, this should not have any effect on audio playback of 192 kHz 24-bit streams on GHz+ systems. I guarantee you that the drivers don't do any kind of audiophoolery with voodoo memory alignments, register zeroing or whatever, so what you do in user space is useless. And the Windows scheduler is a tyrant from whom you cannot escape.

 

Tried your player on ASIO and everything else controlled by Lynx, has to be the softest player I've listened to, kills the impact of percussions and strings, never heard such a thing, made the detail nicer but sounds fake-ish, like a smoothing filter and ironed out imperfections.
I kind of hate JRiver but can't really get rid of it

Leave this crap for later, focus on the macro first, HP, dac, amp then when you're getting bored and have literally nothing better to do fiddle with micro, buffers, fuses, cables, isolation, quantum stickers etc.

 

Lynx conversion is known to be soft, wide, and slightly warm but nothing like an Apogee. It’s not gonna hit you like Dangerous will or punch like Prism. You’re just hearing it at work better than with Jriver and foobar2000. Similar to a good daw. It varies less than DAWs. You can pretty much control stereo width and center depth in most daws by changing the panning settings.

Winyl has depth that is destroyed on other players. The distortion of the windows limiter is very apparent in foobar2000 now. Foobar2000 1.5.1 and up is majorly distorted out of the box. It sounded more immediate at first but then I realized it was harsher and distorted. Jriver is super heavyweight and is best for video. Great for blu rays. Can mess with the sound a bit depending on cpu load.

Winyl detail is next level. You can hear tape saturation much more clearly on bass and kicks. You can pick out the gear used to compress different drums and mics on the kit. You can’t on foobar2000. It’s clearly just better. All of your recordings not modern classical done on rme or infamous pro tool hd digidesign 192 have analog warmth baked in. Even overdriven SSL 4000 mixes like typical 90s music will have it due to the converters and tape.

I will test it with something I did myself later today.

 

Yes I am aware. I worked at a Semiconductor company for quite a while, and even then I was skeptic of this. But since it was free, I gave it a shot. And Phew! It did make a difference to the sound (as described above). Like you just said, modern processors can have thousands of thread running at a time, but I doubt any of them is actually real-time or in uniform speed throughout (jitter-free).

"speculatively executed out-of-order" - this is what his tool tries to avoid by altering how the instruction set are organized, just by spacing the data transfer instructions by some random wait/loop instructions. I don't know why he did that but apparently this procedure inherently causes jitter that somehow propagates in the rest of the chain and impacts till the dac, sothe tool tries to fool the logic to not go by that route. This is a topic in itself that I'm just dipping my toes into, so this is the best I could explain. Also, every now and then there are bugs/exploits being found in existing processor architectures, example - "https://meltdownattack.com/". Nothing is out of possibility.

You can escape from windows scheduler if you use some other OS.

Regarding how I knew what he did, I just asked him. He has a discord server. The guy is doing this for free when he gets time, and I'm glad he took his time to explain me what is happening. He is Japanese I guess, mostly uses translator to translate to English, and lot of content got lost in translation, which is why I have no clue of how he claims to be able to do that voltage level referencing. He even showed me code snippets. You can shoot him your questions, but kindly keep it civil.

FYI, it doesn't seem to stop with audio playback either. Some CPU based image display or whatever also will be affected from what he tells. The designer told me that he used a screen wallpaper of blank color to validate when he was coding. A friend of mine also saw changes in his screen the moment he fired this tool though he couldn't tell apart audio. For me it was opposite, I couldn't see any change in my screen but could hear changes in sound.

 

Winyl looks interesting but i see it's windows only(name gave it away kinda hehe). I play my files on Jriver with tdpf dithering under Fedora which sounds good to my ears. Also has a remote app which is a win for my lazy ass.
Haven't experimented with low latency kernels yet, perhaps i will sometime when i'm bored.

 

OK so I started reading this thread and I became superbly uninterested after the first few paragraphs. My thinking was:
"If the instructions have to flow in the same order through the entire computing system, then there is nothing left of general computing. We are done with parallel IO, memory cache etc.. and back to $1000 price tag for a single run of a Cobol program. Or a perfect run through "Hotel California".
Then I downloaded Winyl. And I cannot argue with what I hear. I spent last hour playing a single track through JRiver and Winyl. My chain here is Xonar Essence SPDIF to a mid-fi DacMagic Plus, Noforce monoblocks -> HD800S .
The difference is not minor, it is very clear. Winyl is at different level: Jriver sounds congested, slightly smeared in comparison, with a bit impaired transient response and instrument separation. Both players with default settings.
I am not saying it has anything to do with the instruction sequence or whatever but you are up to something. Reading your post is a pain but it is worth it.
Except talking about USB is just waste of time.

 

I had the opposite impression at first, comparing the two I found:
-Jriver has harder hitting low with more weight and deeper
-Jriver adds slight mid warmth
-Winyl has an upper hand in detail and airer, wasn't that obvious with HD650, on MySphere it's super clear, I thought the slightly harsh treble was the AVVTs, it was actually the f'ing player, who would have thought.
-Winyl lacks punch compared to Jriver, bass heavy/focused songs sound off and lighter.
-Jriver is better for an all rounder even if now I hear the detail smear, Winyl can be better or worse track dependent.
-Winyl interface is simple and to the point, just wished it had bitrate as a column.
-Winyl suits MySphere 3.1 like a glove, Jriver is superior for HD690s.

So a player, software can make a difference, but cables of different materials and geometry is too much of a stretch...

 

My Digital Audio Journey (So Far) AKA If Homer Was An Audiophile.

 

Have you tried HQplayer? It is available for Linux. On windows, It sounds Identical to winyl, but unsure how it sounds on linux.

You don't need gazillions of speed for real-time applications. Timing is more important. An analogy I could give is Vertical Sync - the software voluntarily ignores a few frames to ensure consistency, but yeah that is a heavy handed task and not comparable to audio. Perhaps a more favorable comparison would be ping. Another example, though obsolete today - Back in the XP/Windows 98 days, with ps/2 mice, my mouse would freeze for a while and then suddenly get a jerk. Seems like windows used pretty bad in handling I/O and let CPU take too much time in other tasks before responding to the poll. So much so that whenever I attend a linux shilling session, the guys used to describe this to downplay Windows.

Make sure you use Asio or Wasapi, preferrably ASIO if your dac supports it. No, winyl vs jriver doesn't relate much to instruction sequence. The player differences mostly stem from audio library and their function calls, buffers, stock decoders, etc. I don't think talking about usb is a waste of time once we could iron out the issues - and companies like singxer and gustard seem to know their way around audio.

People who are on-board already, here is a link to a test comparing different mp3 decoders : http://mp3decoders.mp3-tech.org/objective.html . Even a decoder can make a noticeable change to the sound, especially on a "standard" that is as ambiguous as mp3.

If I had a lot of money lying around I'd have got better DAC and amp. I did purchase a set of sparkos ss3601 opamps for my amplifier (burson fun) but it is taking really long to reach my country. Regarding headphones, I have over 20 of them, and listen to only 3 of them at most. I don't think any of these phenomenon are to be ignored, atleast imo. Before I can choose a DAC, let alone spending tonnes of money for it I need to ensure the rest of the variables are taken care of. And more importantly, these give me better learning (signals, digital systems) than I would get if I just put a lot of money to a better DAC. I am currently trying to use the same concept to create a digital transport in my arm computer, that hopefully would be better suited for audio.

Most of my listening (70-80%) happens through either my phone speakers or some 20$ bluetooth speakers. I'm not as much an elitist as my posts would suggest - convenience wins for most people, which is why hifi-focused companies are scarce as compared to feature focused audio companies. I experiment with pretty much everything that an audio purist would dread - crossfeed, binaural synthesis, equalizers (high precision ones), among others though haven't found any implementation of them convincing enough yet. I just find scope of learning electronics and computer engineering here so I experiment and when I can find differences, explanations or even predictions, I share it so experienced people can validate it for me, and correct me if needed. I was obviously prepared for a fair share of skepticism here, but glad it didn't turn out as negative as I expected. People from most forums refuse to even "try", let alone comprehend what I say.

Yes and Yes. I hear that, except that I associate the punch from Jriver to compression. Loud compression gives the feeling of punch or hit at the expense of micro dynamics. Winyl sounds odd at first, but as you listen you start to notice a sense of texture detail that the other player couldn't resolve. It feels more real (though nothing is as real as reality). I am unable to describe it properly, Psal describes one of those textures as "tape saturation".

 

Cables - I wasn't meaning specialized crystals or geometry. All I mentioned was about adherence to the USB specification - 90 ohm impedance and minimal phase skew between D+ and D- Lines. The supra doesn't even cost that much. By the time you end up with 4-5 torn cheap USB cables, 40$ for supra might even sound like a bargain.

Glitch:
http://www.designcabana.com/knowledge/electrical/electronics/digital/propag/

Speed:
https://www.tutorialspoint.com/basic_electronics/basic_electronics_transistor_load_line_analysis.htm
https://www.edn.com/understanding-the-basics-of-setup-and-hold-time/

Yes there are logic present in USB to try their best to correct all these but I'm no longer convinced of them, especially in relation to USB asynchronous audio Protocols.

 

@dietwater messaged me on another forum and we've been talking for a little while off and on. He did remind me of how important it was to not use windows audio mixer and to have an exclusive mode like ASIO direct to your DAC. I reinstalled JRiver's windows audio driver which bypasses windows sound mixer and noticed the audio was less fatiguing even with brighter recordings. I do recall this happening when trying JRiver's drivers previously. So that was a definite improvement and one I intend to keep using.

I don't pretend to know the inner workings of computer processing, timings or other core functions but I wanted to give a free program a try because why not, it's free. I installed Minority Clean Rev23x64 on my computer and waited a few minutes. I restarted my Chrome browser as I typically stream with Google Play music and similarly to above, I noticed the music to be less fatiguing and could maybe hear decay and ambiance better. Was this because transients were softened? Or quite possibly placebo. I'll play with it on for a few days and switch it off and see if I notice anything.

Listening for changes like these, at least for me, is not about A/Bing, as often I find I've changed a cable and forget then realize the music just isn't flowing as naturally as it used to or the sound stage is off. Sometimes it's pretty obvious (Pangea cables) or not so obvious, Gotham cable vs BJCable interconnects but I can usually get a vibe after a few minutes.

So, what does any of this mean? I don't know. Try stuff, see if it makes you smile more listening to music. If not, don't use it.

 

@dietwater any other settings I need to change besides this?



Downloading Minorityclean as well.