Stereo Tool
https://www.forums.stereotool.com/

Stereo Tool 7.51 BETA
https://www.forums.stereotool.com/viewtopic.php?t=5635		 Page 42 of 44
Author:  Bojcha [ Thu Nov 13, 2014 2:09 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Yep, for pre-limit is no go under -3,5dB (for non-fm). I use 16 bands because i like how/where bands are. Still at -3.5 there is some IMD on bursts even if constant sounds looks ok with that settings.
Author:  hvz [ Thu Nov 13, 2014 2:14 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
On another matter, the dropped presets, which all of mine were dropped, what happens if someone is using a preset that has been dropped? If the user switched away from a preset for whatever reason, unless previously exported, there does not seem to be a way to get back to the original preset without reinstalling an older version, exporting, installing the newer version, then importing it back in. That would seem to be a hassle, particularly if it happens in a broadcasting environment.

As far as mine that were dropped, I know that you prefer for people to not use the "classic" part of the code, however, particularly for the "generic" presets, the CPU load is still significantly less if you use the older functionality. I am still having to use my aged AMD hardware. I do have a Core i5-4460 in-house now (replaced a P4 2.53), but that is not my system. Not concerned about me specifically, because I know what to do to use the older code, but IMO, there still should be a few choices to pick from that use the older code path(s).
I got lots of comments from people that there were too many presets. If someone switches to the new version of ST the preset they have loaded will keep working. I've just done a comparison and the new MB with 7 bands uses the same amount (actually slightly less) of CPU load on the highest loaded core as the old one (which has 10). With 8 or 9 bands it is indeed more, and also on a single core system (tipover point is 5 bands there). If you enable "Flat tops" which some of the newer presets use, the CPU load goes down further. Also, the CPU load of the whole program has gone down a lot in this new version, mostly on new Intels (17%) and - based on what I heard so far - all AMD's.
Author:  hvz [ Thu Nov 13, 2014 2:15 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
Yep, for pre-limit is no go under -3,5dB (for non-fm). I use 16 bands because i like how/where bands are. Still at -3.5 there is some IMD on bursts even if constant sounds looks ok with that settings.
Yup, I will check that in a future version.
Author:  DJ-DOGGY [ Thu Nov 13, 2014 9:06 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Bojcha , Hans .I cut a part of a song and give it to you with the settings to see that -3 db isn`t enough to handle and there are dropouts(still noticeable in my opinion). Also less bands causes other frequencies to be "rediced" too . I mean frequencies around the "target one" .

It`s just around 50 seconds. You can use it for tests also . It`s very good for that and i cut it just good even for loop playing if needed .
I think this on video is .... i call it "unwanted signal"/ringing so it should not present in the sound. I will vote for fixing it if i can :)


http://www62.zippyshare.com/v/33277311/file.html
Author:  mpex2006km [ Fri Nov 14, 2014 2:33 pm ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
I think this on video is .... i call it "unwanted signal"/ringing so it should not present in the sound. I will vote for fixing it if i can :)
+1. I use the pre-limiter for AM use and it sounds fantastic.... If i set it to the highest it can go (-12db and 2048 bands) and let it limit sound as well as tones it make a very loud but still crystal-clear sound. Exception is (as already demonstrated before) that when a tone passes through it get modulated with some other tone and ringing occurs. So if you can fix that it will be perfect....
Author:  Brian [ Sat Nov 15, 2014 3:45 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
I got lots of comments from people that there were too many presets. If someone switches to the new version of ST the preset they have loaded will keep working. I've just done a comparison and the new MB with 7 bands uses the same amount (actually slightly less) of CPU load on the highest loaded core as the old one (which has 10). With 8 or 9 bands it is indeed more, and also on a single core system (tipover point is 5 bands there). If you enable "Flat tops" which some of the newer presets use, the CPU load goes down further. Also, the CPU load of the whole program has gone down a lot in this new version, mostly on new Intels (17%) and - based on what I heard so far - all AMD's.
As far as the number of presets, I agree there were too many. Another person (who is still active) and I had a discussion at least a year, if not almost 2 years ago, about the inclusion of so many, and a concern about the impact on the "reputation" of your product, if I am remembering correctly. I believe around that time I mentioned something about a preset "review board". I am not sure if that is really viable, but I'm just throwing that out there.

Continuing along that same line of thought about the number of presets, I would still say that there are too many. In particular, there are some that are almost identical in their sound profile, specifically in the "Generic" section. The reason for this is that you are downgrading presets designed for FM and Web to "fit" the non-paid segment. In my opinion, this does not always yield a good result, primarily because the person who developed the preset never tested how the preset sounds without the additional audio tweaks that the clipper provides.

As far as the number of bands, that is indeed the "gotcha". The question that needs to be answered for me is if a 5 or 6-band multiband can provide equivalent processing. I have plenty of time again, but I'm less enthusiastic about fiddling. That's my issue, not yours, but it is what it is.

Also, as far as what you've heard from people with AMD systems with respect to the CPU load, it still remains highly L2 cache bound. 1MB per core is not enough L2. Some of the presets do not have certain things enabled, so if I try to enable them, even on 5 or 6 bands, as I have 1MB L2, I will likely quickly overflow the L2 and start having to go to system memory, slowing things down as the CPU load then goes up.

I'll play around some, but I am specifically raising an issue about the "generic" section. As far as Web/FM, I agree with using the newer code.
Author:  Brian [ Sat Nov 15, 2014 4:27 am ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
Quote:
Quote:
Fortunately it locks up here immediately as well. So, I can start to investigate it. Thanks!
Happy to be able to help.
Uhm, how did you manage to get your "L+R to L-R channel mix" setting (AGC Matrix Mode section) to a value below 0? This was causing the software to take the square root of negative numbers in some cases... which apparently didn't cause issues in the older version (?) but it does now.
I enabled Matrix mode in my ini (not tjh's) and changed that setting to -1, and mine still does not crash.

It *DOES* however crash with a -1 for the channel mix and channel max instead of mix turned OFF.
Code:
Matrix mode=1
Matrix mode L+R to L-R channel mix=-1
Matrix mode channel max instead of channel mix=0
Matrix mode stereo reduction=1
...crashes...
Code:
Matrix mode=1
Matrix mode L+R to L-R channel mix=-1
Matrix mode channel max instead of channel mix=1
Matrix mode stereo reduction=1
...does not crash...

This makes sense, because the setting said "Ignore the L+R to L-R mix setting".

Quote:
What I have changed:
1. Made sure that this value is clipped above 0
2. The GUI crashes that would be caused by bad numbers are caught and handled gracefully (exception popup), instead of locking the GUI (in fact an exception was generated and handled correctly, but some GUI exceptions cause the GUI to be switched off so you would never see it). This might fix other issues.
Exception handling is good... ;) The NAN issue a few weeks ago should have been caught too (div by 0, I think).
Author:  hvz [ Sat Nov 15, 2014 3:08 pm ]
Post subject:  Re: Stereo Tool 7.51 BETA
Quote:
Exception handling is good... ;) The NAN issue a few weeks ago should have been caught too (div by 0, I think).
I wish! Unfortunately, divide by 0 does not raise any exception in optimized code - actually I saw it doesn't even always cause one in debug mode... If it did it would be easy to find the cause.

Also, I have seen that fixing a *very* unlikely division by 0 error in the clipper causes the CPU load to increase. So I've opted to instead check the result afterwards and replace any NAN's by 0's. I *have* added logging when this happens, and so far I've been running the version with this logging for 12 hours without a single occurrence.

What also happened was that a NAN value in some filter outputs would cause havoc in the GUI because - annoyingly - functions like min and max don't work (they again output a NAN). And when you convert that to an INT you get a MAX_INT like result, which in turn can cause calculations of memory addresses where to draw something to return weird numbers. I did put in checks for that as well.

What I really don't understand is why I have never noticed these issues with the old compiler. It's good that they occur now and get fixed, but I would really prefer to have some sort of mode in which all NAN's get turned into 0's automatically. Since I'm not interested in NAN's... I'm going to ask a question on the Intel forums about this...
Author:  hvz [ Sat Nov 15, 2014 5:25 pm ]
Post subject:  Re: Stereo Tool 7.51 BETA
BETA751-032

Some people were having some issues with silence and crashes. I think I have fixed (most of?) those now.

I have added a LOT of extra checks and logging when things go wrong. Here are 2 different versions: One normal version and one with extra logging. Warning: It might not be a good idea to use the versions with extra logging for production use. The CPU load is a lot higher due to the extra checks, and if something goes wrong huge amounts of logging (even multiple Gigabytes) can be generated.

If you want to try out the version with logging, make sure that a directory C:\temp\ exists. Logging will be written there in 2 files, one called 'log', the other 'stereo_tool_exceptions.txt'.

Windows 32 bit:
Windows stand alone: http://www.stereotool.com/download/ster ... 51-032.exe
Winamp/DSP plugin: http://www.stereotool.com/download/dsp_ ... 51-032.exe
VST plugin: http://www.stereotool.com/download/vst_ ... 51-032.dll

Windows 32 bit - EXTRA LOGGING:
Windows stand alone: http://www.stereotool.com/download/ster ... 032LOG.exe
Winamp/DSP plugin: http://www.stereotool.com/download/dsp_ ... 032LOG.exe
VST plugin: http://www.stereotool.com/download/vst_ ... 032LOG.dll

CHANGES:
- Check preset CPU loads, and fix those that are very heavy (some have Composite Clipper Strictness set to 20 for example, which causes the CPU load to be extremely high, with very little if any benefit to the sound)
- Highpass filter does not work if DC POST filtering is disabled! Caused in 027, 026 was still ok.
- Fix several NAN issues
- Fix NAN issue causing silence in ACR Stereo

QUESTION: If there are presets here that should be removed, please let me know. Also if I should order them differently. I know people have asked for more info about presets, I cannot add that now but it's still on my todo list...

TO DO:
- Forget last value in FFT copy - fixed for declipper, check rest of code!
- Fix HW version crash: Exception in: Modify_samples call internal (Exception 0xc0000005 at address 0x72200fe4, read/write=0, r/w location=0x1fdaf020). Offset-info: 0x8b33098 (NSA). 72200FE4 movzx ebx,byte ptr [ecx] - CRASH ONLY OCCURS IN WINAMP BUILD OF HARDWARE VERSION, IN THE SOUND CARD HANDLING. IGNORE FOR NOW...
- Check with Valgrind
- Try what happens when initializing with NAN's in claim_ss function.
- Exception in: GUI Event Handler (Exception 0xc0000005 at address 0xf5f5b8, read/write=1, r/w location=0x0). Offset-info: 0xeb0c84. Reported by Fario and 2Sense. Unfortunately the crash occurs in a function called __intel_memset and I cannot see where it's called from. (Well, the GUI obviously...) Has not occurred recently anymore (last few beta's), fixed?
- Improve web interface further. -> Postpone, can be done together with hardware GUI development. (very soon, among others preset selection)

OLDER:
Reorganized preset order.
Added separate FM and AM lowpass frequency setting; stream output uses a different frequency than FM/AM.
Added separate volume for 1st and 2nd MPX output.
Added Multiband Pre-Limiter. Protects clipper against volume effects (pumping) and distortion.
Added Bass Boost "Drive" slider for better control.
CPU load: Added faster processing paths for SSE3, SSE4.1/2, AXV and AVX2.
CPU load: Optimized Delossifier.
CPU load: Optimized Stokkemask (CPU load reduced by around 70%).
Delossifier: Disabled at low latencies and monitoring output (adds latency, makes no sense in lower latency modes).
Compressor/Limiters: Moved limiter before compressor. Slows down the attack speed when limiting, giving a more stable sound.
Limiters: Solved distortion when limiting bass sounds.
Limiters: Default "Limiter Distortion" values changed. Range increased downwards from 1.0% to 0.2%.
Advanced Clipper: Stokkemask: Redesigned to have less impact on the audio.
Advanced Clipper: Made highs clipping level configurable.
Advanced Clipper: FM Multipath protection: Improved the behavior, effect is much bigger now.
Web interface improved (scales to screen size, many widgets added).
GUI: Simplified tilt correction. Only one slider left.
GUI: Added CPU load display (as text) to the CPU load sliders.
GUI: FM: Add volume meters which show MPX and Left/Right above 100% modulation
GUI: Added warning that it's better to use ASIO and setting the sample rate correctly in Windows.
GUI: Changed band steepness minimum values for Multiband/Natural Dynamics from 6 to 3 dB/octave.
Code improvement: Assigned unique ID's to built-in presets, to make it easier to add, remove and shuffle presets.
Bug fix: Potential audio drop in FM output with Tilt correction enabled solved.
Bug fix: Single-channel file in plugin version caused crashes when using the composite clipper.
Bug fix: GUI: Tilt calibration was reset when loading a preset, removed.
Bug fix: RDS ClockTime summer time mode on 64 bit OS's fixed.
Bug fix: Stokkemask "Skip Smoothing" mode didn't work properly.
Bug fix: Bug that caused sliders and top menu to stop working fixed.
Bug fix: Calculation of FM MPX left/right peaks was wrong when using Composite Limiter.
Bug fix: MPX Power calculation was wrong after switching to a different sample rate.
Bug fix: Fixed broken GUI display in AM mode when in full-screen
Added new and removed old presets.
Author:  DJ-DOGGY [ Sat Nov 15, 2014 6:58 pm ]
Post subject:  Re: Stereo Tool 7.51 BETA
Is it the time for cleaning out the old stuff ?
Like : Old bass AGC , Old MB , And all the old stuf . Then to be cleaned the presets that using them ?
Page 42 of 44 All times are UTC+02:00
Powered by phpBB® Forum Software © phpBB Limited
https://www.phpbb.com/