Tír Na HiFi

2.29.1
think the stage got smaller than norm. compare to 2.20 or 2.27.

jesuscheung wrote:2.29.1
think the stage got smaller than norm. compare to 2.20 or 2.27.

right, a bit reduced.

uploaded v2.30 256 Fibonacci , very prefer this sound. try it.

lekt wrote: uploaded v2.30 256 Fibonacci , very prefer this sound. try it.

and v2.30.1 256 Fibonacci.

2.30x use buffer size 256 frames with 3 loop: loop in loop in loop.

uploaded 2.30x mod

2.30.x
-more positive tune, more direct, more energy/power.
-2.29x very polite(too mellow sometimes?). 2.29.1 best piano flow. good treble notes.
-2.30 is less bright(in piano), more real. 2.30.1 seems more refined.
-slightly prefer 2.30
-soundstage sounds very correct in many albums! not flat. not 3D either. just sounds right.
-vocal is good... something is missing. prefer 2.29.1's vocal. not sure why. doesn't feel the same.

very good! thx
like both 2.29.1 and 2.30

jesuscheung wrote:uploaded 2.30x mod

2.30.x
-more positive tune, more direct, more energy/power.
-2.29x very polite(too mellow sometimes?). 2.29.1 best piano flow. good treble notes.
-2.30 is less bright(in piano), more real. 2.30.1 seems more refined.
-slightly prefer 2.30
-soundstage sounds very correct in many albums! not flat. not 3D either. just sounds right.
-vocal is good... something is missing. prefer 2.29.1's vocal. not sure why. doesn't feel the same.

very good! thx
like both 2.29.1 and 2.30

2.29x good vibrates but more dry, 2.30x more water. i prefer 2.30x, soundstage more correct.
i'll come back to 2.29x later, there's new method, may also optimize again, its sound width a bit not enough.

jesuscheung wrote:lekt, you don't use any hardware prefetch, right?

i am seriously considering disabling "adjacent cache line prefetch" in bios. so much better sound with lekt player.

too mellow with "adjacent cache line prefetch". disable it, more revealing. piano is 99% perfect. treble is 5% better.

my bios setup of Dell very simple, don't let see many things. there's laptop mainboard, can not do anything.

2.30.1>2.30 in vocal. 2.30.1 has great vocal! sorry, only listened to 2.30.

forgot the buffer is changed. updated setting for new buffer 256.
2.30's piano flows amazing.

this size of buffer is very good. much easier to tweak with timerresolution. now bass is best i ever heard! 1024 is terrible to tweak. keep using 256!

thx. 2.30>2.29x

jesuscheung wrote:forgot the buffer is changed. updated setting for new buffer 256.
2.30's piano flows amazing.

this size of buffer is very good. much easier to tweak with timerresolution. now bass is best i ever heard! 1024 is terrible to tweak. keep using 256!

thx. 2.30>2.29x

i confused version name, 2.30.1 not used Fibonacci ratio, only 2.30 used.
2.30 bass interesting, and temp. sometimes it sounds as water flow. made cut by Fibonacci ratio.

think buffer size don't affect to sound, depend on how player load data to buffer, if loading is regularly and speedly then sound is good. buffer size is number, can be used its characteristics to optimize code. for example, 256 can use UINT8, 255 + 1 = 0, 0 - 1 = 255 (very interesting); 4096 have benefit if use chunk size, page size,... developer takes care about these characteristics.

lekt wrote: i confused version name, 2.30.1 not used Fibonacci ratio, only 2.30 used.
2.30 bass interesting, and temp. sometimes it sounds as water flow. made cut by Fibonacci ratio.

think buffer size don't affect to sound, depend on how player load data to buffer, if loading is regularly and speedly then sound is good. buffer size is number, can be used its characteristics to optimize code. for example, 256 can use UINT8, 255 + 1 = 0, 0 - 1 = 255 (very interesting); 4096 have benefit if use chunk size, page size,... developer takes care about these characteristics.

256 is good coz 256/44100=58050 = 5.8ms. almost 6ms. very good sound using with timerresolution=1ms.
1024/44100=23220=23.3220ms. bad bad. doesn't go well with timerresolution = 1 or 0.5 or 2ms.

i tried timerresolution=0.58ms or 0.575ms for 256 buffer. the sound has improved but.... but because timerresolution is not normal as in 1ms, or 0.5ms, the rest of windows suffers. the sound is also compressed. both good and bad.

i find that if buffer has clockrate rounds to nearest ms, the sound will be better.
e.g.
352/44100 = 8ms. good for timerresolution=1ms.
640/44100 = 14.5ms. gooood for timerresolution=0.5ms.
256/44100 = 5.8=6ms. ok for 1ms. not perfect. timerresolution=0.58 is ok.
1024/44100 = 23.220ms. not good, always issues. no timerresolution seem posible. 0.23 is impossible.
etc.

basically
i prefer lekt2.30 + 256buffer + 58050clockrate + 0.575 or 0.58ms timeresolution. not perfect. but i like it.