10 replies to this topic

[JonP]

I have been frustrated by an inconsistency in the sound quality of my digital setup (Core i7 with 8GB of RAM and JPlay xtreme mode). In all the testing I have carried out over the last month, I have been happy with the results - I had been able to obtain a sound after weeks of tweaking and listening tests which - by the standards of digital - I would now consider to be satisfactory (which I consider to be a very good result given that I am not a fan of digital except for the sake of convenience). The problem has been, however, in that subsequent "real world" listening with JPlay (to complete-length tracks as opposed to the test tracks) the "real world" listening produces a worse sounding result. A solution (and an explanation) for this discrepancy eleuded me until this week, when I decided to figure out why the test tracks all sounded satisfactory, but the real world tracks did not - this is despite me listening to those exact same tracks in both cases, the only difference being that in the test scenarios, the test tracks I was listening to were no longer than around 45 seconds to 1 minute each (so file sizes were less than 20MB at 24 bit, 48 khz), whereas the real tracks can be as long as 30 minutes (I listen almost exclusively to classical). In other words, my test tracks were merely excerpts from the real world tracks rather than the full length tracks. After three days of testing, I believe I have now isolated the reason the real world tracks sound worse - quite simply they are much larger files. I confirmed this to be the case by re-creating a number of my very short "test" tracks, but instead of keeping them small (i.e around 20MB maximum), I padded them with silence for another 25 minutes so as to make them around the same size as the full tracks. This enabled me to listen immediately to the particular test sections I am interested in (using JPlay mini), without having to listen through partial tracks to get to those actual critical musical test points. And sure enough, there was a clear loss of sound quality with these new "1 minute" tracks, seemingly because I had increased the size of the files to be similar to that of the normal tracks (up to 30 minute) tracks that I listen to. In other words, once the file sizes were the same, the sound quality became the same (i.e not as good for large files, better for small files). And I am not talking about a slight difference either - the sound quality of even a 16 minute file drops off quite noticeably compared to a 1 minute excerpt from that same file. These results are nevertheless is a bit mysterious, as although my machine does not have a large amount of RAM, it has sufficient RAM to run JPlay at anything from a 500MB to 4GB memory allocation whilst still having enough for the "rest" of the machine to function without problems (I normally have the allocation set to 1 GB, since that is plenty for a 30 minute file at 24 bit, 48 khz when using the Xtreme engine). So then I began experimenting with different memory sizes in JPlay, wondering if simply increasing the memory allocation would ameliorate the sound quality losses from playing larger files versus the tiny test files. But again, I experienced results that were contrary to my expectations - the larger the memory allocation in JPlay, the worse the sound. So this could produce a double whammy of bad sound - playing a large file with a large memory allocation versus a small file with a small memory allocation. After a full afternoon of listening tests, I concluded that increasing the meory size of JPlay in my setup beyond 2GB produce worse results. So, given that the amount of memory did not seem to be helping the situation, I decided to play with the memory speed in the BIOS. I had always run the memory at the safe default speed of 1333 Mhz with 9 timings, even though I had always known the memory had an XMS profile of 1600 Mhz with 9 timings. I had never bothered, even though my mainboard supports these XMS profiles - I only bother with these sorts of tweaks on my gaming rig. In any case, i thought I had read previously that it was better to run the RAM at conservative speeds when it comes to audio). Anyway, I decided to set the XMS profile to the 1600 Mhz setting and then repeated my listening tests. This time around, the quality loss between the large files and the small files was extremely small - infact up to files of around 16 minute length, I could well say that the 1600 Mhz XMS profile actually sounded slightly better than the standard 1333 Mhz profile - or at the very least it sounded as good. So it seems that in order to retain the same sound quality in a 16 minute file at 24 bit, 48 khz as I do with a 1 minute file at 24 bit, 48 khz, I have to increase the speed of my RAM by 20%. These experiments have led me to the following hypothesis with regard to JPlay in my setup:   1. Adding more RAM of the same specification won't improve sound quality. 2. Exchanging the RAM for a kit with a faster XMS profile will likely improve sound quality so long as the mainboard supports that XMS profile. 3. The larger the size of the file being played, the worse it sounds, all other things being equal. 4. Faster RAM speeds will ameliorate the sound quality loss referred to in 3 above.   These findings seems to be different to what I thought I had read. As I mentioned, I had been led to understand it was better to run the RAM at a slower speed, but in my testing, not only does the faster speed improve sound quality, but it is actually neccessary to run the RAM as fast as possible (consistent with a programmed XMS profile) in order to reduce as much as possible the sound quality losses incurred in playing large files.

[Marcin_gps]

Hi JonP, Very interesting report, thank you!   That's why I never set Memory in jplay above 500M However, in my setup, going from 4GB RAM to 8GB yielded significant improvement (Memory in jplay set to 500M in both cases), so I'm not exactly sure whether more ram is not partial solution to this issue. I have to add some fuel to the fire – not only track's size has impact on SQ, but path to where the files are located on your drive matters too! The longer the path is, the worse sound becomes. And the difference is not small! The best results are achieved when the track lies in root folder on partition and has only extension, without name (e.g. the path is C:.wav). You can't remove file's name completely using Windows Explorer, but it is possible in total commander. Not knowing this would be probably better

Edited by Marcin_gps, 10 July 2012 - 08:18 PM.

[JonP]

I had always run it at 500MB myself, except that it would not play some of my larger files (> 25 minutes). So I had set it to 2GB not putting much thought into it. I actually did further testing tonight on lower figures (500, 1000 and 1500) and actually found that 2GB seems to be a sweet spot for me. It definitely sound sworse with larger files at anything below 2GB, however I don't really hear a difference with small test files regardless of the settings I use. I did not know about the difference when the file paths are different - that is something I can of course easily change. Then again, I have heard differences depending on whether the music comes from a USB thumb drive versus the hard drive (and a solid state one at that), so I guess anything is possible (or more correctly, nothing surprises me anymore). I'd been thinking of buying 16 GB of RAM for my main PC given it is relatively cheap these days but it would seem any result as regards audio quality is unpredictable. I would never have thought even a year ago that the amount or speed of RAM should make the slightest difference to sound quality, but now I realise that simply buying a different RAM kit will definitely change the sound. But there is no way to know if it will be better or worse. I bought a G.Skill RAM module for my notebook last month and was astonished to find it sounded different to the Kingston module I already had, despite them being the exact same size, speed and timings. The G.Skill sounded relatively muddy and the Kingston was clearer sounding. I long for the old analogue days - things were much easier back then. Digital is just too hard to get right. Or to put it more precisely, it is much easier to damage the sound in digital audio than it ever was in the analogue days.

[Zan.Tiago]

Hi JonP, Very interesting report, thank you!   That's why I never set Memory in jplay above 500M However, in my setup, going from 4GB RAM to 8GB yielded significant improvement (Memory in jplay set to 500M in both cases), so I'm not exactly sure whether more ram is not partial solution to this issue. I have to add some fuel to the fire – not only track's size has impact on SQ, but path to where the files are located on your drive matters too! The longer the path is, the worse sound becomes. And the difference is not small! The best results are achieved when the track lies in root folder on partition and has only extension, without name (e.g. the path is C:.wav). You can't remove file's name completely using Windows Explorer, but it is possible in total commander. Not knowing this would be probably better

Hi, Marcin

Sorry about the silly question, 
How do you set Memory in JPLAY ?

Thank you in advance!

[Zan.Tiago]

Hi JonP, Very interesting report, thank you!   That's why I never set Memory in jplay above 500M However, in my setup, going from 4GB RAM to 8GB yielded significant improvement (Memory in jplay set to 500M in both cases), so I'm not exactly sure whether more ram is not partial solution to this issue. I have to add some fuel to the fire – not only track's size has impact on SQ, but path to where the files are located on your drive matters too! The longer the path is, the worse sound becomes. And the difference is not small! The best results are achieved when the track lies in root folder on partition and has only extension, without name (e.g. the path is C:.wav). You can't remove file's name completely using Windows Explorer, but it is possible in total commander. Not knowing this would be probably better

The longer the path the worse woud aplly to JPLAY mini ? Since we are copying to the track to the player ?

[Marcin_gps]

Hi, Marcin

Sorry about the silly question, 
How do you set Memory in JPLAY ?

Thank you in advance!

This is minicache in JPLAYmini. You resurrected a very old topic

[billgr]

"I long for the old analogue days - things were much easier back then. Digital is just too hard to get right. Or to put it more precisely, it is much easier to damage the sound in digital audio than it ever was in the analogue days."

 

Yeah, right (smirk)! 

Clean the vinyl. Clean the stylus. Ignore the crackles, pops and rumble. Don't fret about the rolled-off frequencies or lesser dynamics. Ignore the sound quality changes from edge of a LP record to the centre. Adjust the azimuth and tracking weight. Try a different cartridge. Get a new drive belt. Tweak everything. Get a special table. Then a vibration platform, etc. It wasn't hard to get a good sound from vinyl was it?

 

But I loved it just the same 

[Zan.Tiago]

This is minicache in JPLAYmini. You resurrected a very old topic

Thanks Marcin,



As well I'm tryin to reduce the root path in windows in order for my audio player to access my music files directly.

"C:\Users\Santiago E.A\Desktop\Alexis Cole\A Kiss In the Dark\01-Ain't We Got Fun.aif"

Is there any way to have something like this:
"C:\esktop\01-Ain't We Got Fun.aif"

Or at least something shorter ?

Thank you in advance!

 

[Marcin_gps]

I replied to your PM

[bodiebill]

I have to add some fuel to the fire – not only track's size has impact on SQ, but path to where the files are located on your drive matters too! The longer the path is, the worse sound becomes. And the difference is not small!

 

It is getting crazier and crazier :-)

But I know: reality is crazy.

 

Question: in your experience, does the path length also make a difference in a dual PC setup where the files are stored on the control PC or even on a NAS? 

[Marcin_gps]

It's not sth I'd worry about