Stereo Tool

Phoenix: That's a different type of rounding errors. I don't even know whether there are more or less of them in 031 than 029. But it is most likely a different rounding in a parameter setting (something like a difference in amplification between 5.0000001 or 4.99999). The filters themselves are not changed.

Basically, software that works with floating point numbers will always have rounding errors, and they can differ between versions but that doesn't have to mean that one version is better or worse than another version - they are just different...

@Hans: In my case the quality setting is in the middle and also happened to me the same as Brian.

_________________
by GAP
"Less is More" (Bob Katz)

Sorry but what is the Q-factor of the LOWPass filter Hans ?
Or maybe You could implement a slider for the Q-factor of that filter?!?

Agreed on the floating point numbers and errors. It's just the nature of the beast with IEEE 754.

That said, I think I'd ask if the changes that increase the CPU load are genuinely audible increases in quality as determined by an audience other than yourself. Due to "vested interest", you may be biased to believe that there is an increase (or a decrease) in quality, when there is neither an increase nor a decrease. Further, there needs to be at least two audiences involved, broadcasters and non-broadcasters.

I know you have to cater more to the broadcasters, as that's where most of the money comes from, but I've gotten feedback about my preset development from a broadcast-oriented perspective that I, as a non-broadcaster cannot actually hear the "issue" that the broadcaster has mentioned to me. I've also been offered feedback based on a strict scope view of the waveform, and again something that I cannot hear. Looking at just a waveform and wanting it to match some definition of a "perfect" waveform is all well and good, but no matter how perfect a waveform may be, the acoustics of any waveform, perfect or not, are changed by the environment they are in. The quality of the reproduction device makes a change to the waveform. The THD of an amplifer makes changes to the waveform. Whether a room is carpeted or has hardwood flooring makes changes to the acoustics. Whether you're in an open space with no walls or in a 4-walled room makes a change to the acoustics. Etc, etc, etc...

In my opinion, based on a lot of time trying to work on a preset, a redesigned multiband that allows end-users to control limiting or compression per band, as well as a parametric EQ, and moving to RMS-based leveling is what I'm looking forward to...

I understand thing about more cpu usage... but..
I also heard difference .. and it's HUGE! Ofcourse Hans is still looking (and he will look deep) how to not make more cpu usage.

I appreciate what you're saying, but I absolutely could not hear any sort of "vibrato" in Adele - Set Fire to the Rain. I had asked you to see if RC2 made things better or worse, but I have not received a response.

This is why I am cautioning against using a limited group of people for subjective tests. I flat out don't hear the issue that you say is there. So, which of us is correct, or are neither of us correct? I honestly don't know. What does that mean then when you say that there is a "huge" difference?

What I do know though is that I cannot progress any further with preset development due to the way that multiband works. I have a choice of replicating bass down to the 35-50 Hz range or making the vocals as loud as they are in the original track. I've yet to figure out how to do both. Either the software simply won't let me, or I need to hit my head on the bathroom sink and discover the StereoTool equivalent to the flux capacitor...

http://www.fluxhome.com/products/freewares/stereotool

Ok, that probably does not help Brian at all.
But I had to exploit the opportunity.


@ Brian

Remember that:

viewtopic.php?p=12941#p12941
Ultimately there is only one thing you can trust to decide for yourself which details (expert theories) are relevant and which are irrelevant for your own listening and life:
Your own ears !

We really need to have this level of spec details for system programming / scientific project? No wonder, I am not in to one of these...

_________________
visit website

My high school / early college ("uni" for those in non-US areas) focus was in Math/Physics/Chemistry. I lost the ability to do math higher than "business math" due to not using the math that I learned in classes like Calculus with Linear Algebra and Differential Equations, probably sometime around 1995. I still retain some knowledge of the concepts though...

IEEE 754 is the standards for floating point math as represented in binary. A very, very, very, very basic explanation is that floating point math represented in binary can have rounding errors. Usually though, the rounding errors happen way, way, way out at high decimal places. It will depend on what kind of application as to how significant potential errors may be. If you are designing software handling telemetry for rocket propulsion or nuclear reactor controls, the signficance will generally be more than that of an application that handles, for example, stock market transactions.

What I gained from Hans' post is he was saying that the rounding errors became significant at levels lower than -90 dB. That is a fairly quiet level, and thus it would impact only a very small percentage of tracks.

The bottom line is that there will always be some level of rounding errors. The goal of a developer is to determine when the errors are not statistically important to the functionality of the software. If an application dealing in currency transactions in the United States has a rounding error at the 10th decimal place, it is not worth putting effort into fixing that. On the other hand, if the rounding error is at the 2nd or even 3rd decimal place, then it IS necessary to fix it.

Make sense?

Precisely the intent of my previous post. Honestly, my work domain never necessitated bothering about rounding errors at 10th decimal place and/or precision of digits leveraged by the compiler or run-time. The digits of precision offered for 'double'/'float' was more than enough for most practical calculations. If you were to ask my system architect as to what IEEE 754, I am almost certain he would end up troubling Google to get what the specification is all about as I did a little while back.
Aggravated by the fact that I was never involved in any scientific domain, I find it really difficult to visualize the implication wrt Hans' predicament.

Nonetheless, thanks, I now know what IEEE 754 is.

_________________
visit website