Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[MerlinWerks]

Hi soundcheck,

I'm afraid I may have been mistaken in my reply to you on the effect of enabling the "Output Format" DSP.  For some reason I was always under the impression that the "Overflow Handling" was not in use until you enabled one of the DSP functions. Upon closer inspection it appears that "Overflow Handling" is on full time, so it is unlikely that it is the reason for less distortion when you enable "Output Format" in DSP Studio. 

I downloaded the ramdisk you recommended and will give it a try this weekend. I also intend to ABX it, ramdisk vs. HDD. ABX software is free and easy to setup if you care to try it yourself.

Available here:
http://www.pcabx.com/training/getting_started.htm

Have Fun!

[soundcheck]

I'm afraid I may have been mistaken in my reply to you on the effect of enabling the "Output Format" DSP. For some reason I was always under the impression that the "Overflow Handling" was not in use until you enabled one of the DSP functions. Upon closer inspection it appears that "Overflow Handling" is on full time, so it is unlikely that it is the reason for less distortion when you enable "Output Format" in DSP Studio.

I downloaded the ramdisk you recommended and will give it a try this weekend. I also intend to ABX it, ramdisk vs. HDD. ABX software is free and easy to setup if you care to try it yourself.

Available here:
http://www.pcabx.com/training/getting_started.htm

The clipping protection you just need if you play with the preamp, equalizer and automatic replay gain.
If you play the tracks pure as they are, you can forget the clipping protection. That's the way how I see and use it.
So just turn it to "NONE" if you're not playing with above mentioned features.

By the way did you notice the sound difference when, tagging the "output format" box? It can't be just me.  ?

I'll have a look at ABX. THX for the hint.

[Alex B]

The DSP Studio shows on option for "output format". If I tag it the sound is cristal clear. If I untag it the sound gets distorted and very close to foobar playback quality.
I am running 16bit 2channels 44,1 and ASIO.

Does that mean that the original data are somehow treated with the DSP?
Any ideas? Any confirmations?

The clipping protection you just need if you play with the preamp, equalizer and automatic replay gain.
If you play the tracks pure as they are, you can forget the clipping protection. That's the way how I see and use it.
So just turn it to "NONE" if you're not playing with above mentioned features.

By the way did you notice the sound difference when, tagging the "output format" box? It can't be just me.  ?

If the source file is 2-channel/16/44.1 and only the 2-channel/16/44.1 DSP options are selected the output is not altered. That can be verified by bit comparing Disk Writer output files.

Also, if all other DSP options are disabled the plain clip protection does not change the output. It kicks in only if the used DSP settings would make the file clip.

[MerlinWerks]

soundcheck,

I will give a closer listen when en/disabling output format.

You really owe it to yourself to setup the ABX test, it can be very revealing 

[Alex B]

I will give a closer listen when en/disabling output format.

It's needless to ABX that. If the source format and the selected output format are the same (e.g. 2-ch/16-bit/44.1 kHz) the actual output is exactly bit to bit identical in both cases. I have tested this with a bit compare tool.

MC's Disk Writer is a nice tool for making test files with other DSP settings. The resulting wave files can be blind tested for example with ABC/HR or Java ABC/HR. These are more comprehensive than the plain PCABX.

Here is tutorial for ABC/HR: http://ff123.net/64test/practice.html

[MerlinWerks]

Hi Alex,

Thanks for the info and the links. Since the ramdisk is thought to sound better by some, my main intent is to simply encourage an ABX test between playback with the ramdisk vs. hdd as an alternative to relying on (organic) memory alone.

[soundcheck]

It's needless to ABX that. If the source format and the selected output format are the same (e.g. 2-ch/16-bit/44.1 kHz) the actual output is exactly bit to bit identical in both cases. I have tested this with a bit compare tool.

Again " Bit Identical" does not necessarily mean "Sound Identical". As soon as you are catching latency-jitter, you got an issue in the time domain for realtime applications, even if the bits are correct. And this is audible!

Refering back to the ABXing. One should go for complex, well recorded material like a nice orchestral track  (e.g. from Chesky records, OPUS3 records in XRCD quality to name my personal references)  for ABXing. Otherwise it is rather difficult to hear the latency impact easily. 
If you own a real good stereo you'll hear easliy reverberations of the room, where the recording was taken. Yes - You shouldn't need a DSP
faking Church or Jazz Club reverberation. It's usually already on the life recording.  If you can hear it well it is a good sign that your stereo is able
to reproduce micro-details of the recording.  Again- Don't forget to switch of the Church DSP, you'd just fool yourself.

The good thing though. If you don't know what's on the recording, it won't bother you that you can't hear it. 

[LesC]

Again " Bit Identical" does not necessarily mean "Sound Identical". As soon as you are catching latency-jitter, you got an issue in the time domain for realtime applications, even if the bits are correct. And this is audible!

so true.. yet hard to prove

[Alex B]

Again " Bit Identical" does not necessarily mean "Sound Identical". As soon as you are catching latency-jitter, you got an issue in the time domain for realtime applications, even if the bits are correct. And this is audible!

Perhaps, but that does not apply in this case. AFAIK, the DSP part is simply bypassed when the selected options do not differ from the source format so all processing inside MC is identical (= non-existent). The next audio component after MC gets exactly the same PCM stream in both cases.

[LesC]

oh ok. thought you were talking about file buffering..

anyways, before, when i'd play flac files (they're 16/44.1) with or without output format checked i'd still get the same output in my stereo system.

now.. since i'm on ogg (aotuv), since lossy has no bitdepth, i have to select 16bit or i'd get grainy highs.. (same as output to 32bit)

[Alex B]

MC's Vorbis decoder outputs PCM stream in 16-bit integer. So you may be experiencing the effect of DSP processing from 16-bit to a higher bit depth.

[Alex B]

On second thought, I'd guess you are merely experiencing the effect of Windows Kernel or the sound card driver reducing the bit depth from a higher value to 16-bit.

AFAIK, you should not be able to hear the effect of MC's DSP when going from 16 to a higher bit depth if no other DSP options are selected and the sound card supports the selected bit depth. I cannot, not even with high end headphones.

Also, if your sound card does not natively support the 44.1 kHz sample rate the output will be resampled to 48 kHz after MC. In that case the best sample rate option in DSP Studio would probably be 48 kHz. MC has a high quality SW resampler. It produces better quality than Windows Kernel resampling.

[LesC]

thanks but i'm using a DMX 6fire (based on the Envy24 chipset) that does not resample. and then i'm running windows vista rc1 with ASIO...

[Alex B]

So we have the same soundcard. I have a Terratec DMX 6fire 24/96 on my HTPC.

Vista's audio handling is a completely new factor. I have read about it, but I have not seen any independent test reviews or tried it personally. Can an old ASIO driver still work properly with it? Can a driver bypass all the new processing that Vista supposedly does?

[LesC]

Alex B, i'm sorry but you're right, i tried it and it sounds exactly the same.. and come to think of it.. how could it output a 32 bit signal to my dac since the dip is only 24/44.1. i really don't know what it was that i heard before... apologies

i'm using xp drivers for the 6fire in rc1's x86 version so i guess that explains why it doesn't do the  processing you mentioned.. my onboard on the other hand with vista drivers.. all this weird stuff about being able to set output format (dolby ?) and sampling rate..

could you post a link about all that extra processing vista does ? i'd like to read it..

thanks

[MerlinWerks]

Here ya go... 

http://arstechnica.com/news.ars/post/20060907-7682.html
http://www.avsforum.com/avs-vb/showthread.php?t=713073
http://blogs.technet.com/windowsvista/articles/450038.aspx

[LesC]

many thanks ! interesting read

[soundcheck]

Here ya go... 

http://arstechnica.com/news.ars/post/20060907-7682.html
http://www.avsforum.com/avs-vb/showthread.php?t=713073
http://blogs.technet.com/windowsvista/articles/450038.aspx

Hi MerlinWerks.

Great AVSforum thread, thx.
As far as I understood MS still gives MC space for soundimprovement on Vista.
Will MC use in the future , I think they called it, "Exclusive Mode" to replace e.g. ASIO?
The MS audio tech specs of Vista are quite limited and far away from being audiophile.
Will MC do something about it?

I was a bit disappointed that the MS guy did not jump on the Jitter issue somebody mentioned.
They were just hammering on the "Bit Perfection" issue all the time.
As I mentioned more than once, Latency-Jitter is to me the well hidden enemy of audiophile audio from PC.
I don't have a clue, why this issue is not really being discussed. 
Further, a comment was made that streaming from HD, while audio-playback is going on, would not
be the best setting for audiophile PC-audio. Pretty much confirming that what I figured when introducing my RAM-disc.
Just to mention it- i mean the ramdisc-thing- , since this thread, if I recall it right,  is about file buffering. 
I am wondering if Vista at all addresses file buffering.

[Mike Noe]

fwiw, after reading thru this thread (and checking my skepticism at the door), I modified the "PlayerCore Min MS" setting to 12000 and I noticed a couple of things:

- I too could get the occasional "click" when browsing with FF and playing music.  I no longer get this

- Task Manager would always cause the same "click", now it does not anymore

- I notice no difference in audio "quality", it's still as great as ever.

Two other things I was able to do as well.  I set the latency switch on the RME Digi96 to it's lowest setting and I was able to set the MC setting on playback to 1 sec.

Computer setup:
XPSP2, Celeron D, overclocked to 2.86GHz, 512MB RAM, RME Digi96/8 PST using their lastest driver, 2.11 (ASIO)

I run a low priority Folding At Home session, SpeedFan and CachemanXP.  MC runs without visualizations and I do not use the DSP or any "PN" plugins.  MC usually peaks about about 5% and hovers at 2-3% in task manager.  My CPU is always pegged at 100% and runs about 45C.

Audio Setup:
(Transcendent Sound Balanced Power Supply) => RME => Denon DVD5000 DAC => TS GG Pre => TS T8-LN monoblocks => Hammer Dynamics Super12s

Also, I've tried Vista a few times on this machine and the RME drivers appear to work fine.

[MerlinWerks]

Also, I've tried Vista a few times on this machine and the RME drivers appear to work fine.

Cool 
Good to know, I have the DIGI96/8 PAD.  I may be giving Vista a look (on a spare box) earlier than I thought I would

[Tuckers]

I know Arnies ABX comparator tool.  We found it very useful when evaluating codecs and rip quality.  It's a very good sounding player too.

I did give a listen to some files played from RAM and from Hard drive with this it, but I could not hear a difference in a non-blind series of tests, so I also obviously could not get anything beyond random on the ABX either.

It's a completely different piece of software and handles files completely differently too.  So I don't think it can be used to make a valid listening test for this issue.

I will try the ABX in Foobar, but fro some reason I can't find the required dll foo_abx.dll for 0.83, which is my preferred version of the program.  I might yield a more significant result.  Anybody know where I can find this dll?  Most of the 0.83 stuff is gone from the web.

[CadErik]

Bit-Perfect does not mean Sound-Perfect!! All kind of processes, task priorities, hardware drivers and hardware can add more latency to the signal. This is causing more or less distortion at the ouput, since we are talking  of a realtime audio bit-stream.

I'm not sure where you take that info that latency causes distortion... When you play audio files there is way enough buffer for the system to counter balance any kind of task switching. The page on the RME website refers to real time playback, on pro systems where you can't afford to have large buffers (1 ms buffer at best).  MC has about 6000 times this buffer size by default so I'm not sure if any of this info does really apply.
At worst, if you had some buffering/task switching issues, you would hear some crunching not distortion. Btw, 6 second buffer is *not* real time.

[soundcheck]

I'm not sure where you take that info that latency causes distortion

I am not saying latency is causing distortion. Latency is just another word for delay and that's what's all about it. For playback it is not a real issue as long as the latency is kind of linear.
What I am saying is that non-linear latency-jitter or jitter is causing distortion, if it is induced to the data-stream which goes (realtime) right into the DAC-chip.
The RME side explains very well the differences.
If your soundcard and its drivers are able to cope with this non-linear-latency-jitter-in-the-datastream fact, you shouldn't have a big problem, as long as your soundcard does a real great job.
I know that some high-end soundcards/DACs do some buffering and reclocking before the stream leaves towards the DAC-chip to avoid these kind of Latency-Jitter or Jitter problems.

The major source for latency jitter is the interrupt handling. And you can't get around it.
This is by the way the reason why UNIX systems are the prefered systems when talking about real-time applications.

Example: Let say you got an USB soundcard. Playback and  datastream is running towards your soundcard. Suddenly "Interrupt" pops-up  saying - "Hold on - I have to read the data from HD first" to fill the 6000ms buffer. This one has higher priortity than USB. Next One. Graphic Card is saying: Stop if have to refresh the GUI.
And there are at least 40 other processes running from which a lot of processes have higher prority than the USB-bus in charge for your Latency-free realtime datastream.

My theory:  If the buffer is 6000 or 12000 that's  not the key issue. The key issue is to avoid long interrupts from e.g. your Harddisk during playback, which potentially influence negatively the USB realtime stream at a point in time where there is no chance to get the Latency-Jitter effects cleaned up anymore.
 
Anyhow. I think this is well known fact - at least to MS. I hope that the "exclusicve mode" in Vista will limit these kind of impacts.

[CadErik]

I know that some high-end soundcards/DACs do some buffering and reclocking before the stream leaves towards the DAC-chip to avoid these kind of Latency-Jitter or Jitter problems.

Pro sound cards have 1ms buffer, as more regular soundcards do usually have more, about (or more) than 5ms buffer. Also with WDM drivers, there is another buffer in the windows kernel in the mixer. Since this buffer is running in the kernel, it would be less prone to interruptions.

I really don't believe in this interruption theory. Unless you have a really slow computer where interrupts are taking lots of time, with a modern computer, I'd be more than surprised that they are visible/audible. To affect the audio stream - given that there is 1ms buffer in the sound cards - it would require that interruptions take around 1 ms?!

[dlone]

I've tried this but have come upon a couple of problems

On short tracks (about 3 minutes) if I'm on the playing now screen with a track info template, the display changes to the next track about 2 thirds of the way through the track and shows the wrong lyrics etc
Also whilst trying to correct lyrics using 'doofs lyric editor' showed the same problem
When you get towards the end of the track and then move back a bit to re-check the words MC plays the next track instead (but shows the origional track name etc in the top section of MC)

[Rambaud]

I have been running a RAM disk for several years - it is currently 1Gb.

I shall try loading the tracks (APE format) onto  the RAM disk and run MC from there to see if ther in an improvement in SQ.