Just an info : the version 1.4.1.0 works much better on Vista for importing impulse responses. With the older versions, many times the import crashed (maybe another memory problem ?)
Can you have a time-window and smoothing simultaneously ?
Thanks for this quick evolving software.
Can you have a time-window and smoothing simultaneously ?
Thanks for this quick evolving software.
Last edited:
New version with frequency dependent time-window
Well, FR-smoothing is IR-time windowing with a frequency dependent time-window. I have now implemented complex averaging, which is equivalent to FPPO (Fixed Point Per Octave) See the User Guide for the formulas I use.
The frequency dependent window is shown for each decade (10Hz, 100Hz, 1kHz, 10kHz)
I have attached an example.
Version 1.4.1.2 (2009-10-12)
Features/Changes:
* Complex frequency domain smoothing (FPPO) <-> Frequency dependant time-window
Well, FR-smoothing is IR-time windowing with a frequency dependent time-window. I have now implemented complex averaging, which is equivalent to FPPO (Fixed Point Per Octave) See the User Guide for the formulas I use.
The frequency dependent window is shown for each decade (10Hz, 100Hz, 1kHz, 10kHz)
I have attached an example.
Version 1.4.1.2 (2009-10-12)
Features/Changes:
* Complex frequency domain smoothing (FPPO) <-> Frequency dependant time-window
Attachments
Last edited:
Well it was more just for VST in general and normal DSP filtering. I use SynthEdit a lot for building VSTs in a modular fashion and I believe someone released a convolver module sometime this year but I haven't gotten around to trying it out.
For the most part I have only been using impulses for measuring - using another program and now yours. But foobar 2000 also has an impulse response convolver that should be convinient since it has ASIO out and support for all types of formats.
This is only true in some situations, but not in general. If the system is minimum phase with no multi-path then what you say will work, but if these things are not true then the variable IR-window and the frequency domain smoothing will yield different answers.
And these kinds of situations are more the norm for loudspeakers than the exception. Diffraction, for example, is going to arrive later than the direct wave and so is a multi-path situation that is very much a part of the loudspeaker and its design. If you smooth with a variable IR window then you will be excluding this diffraction more and more at higher and higher frequencies, where, in fact, it is more and more audible. This is NOT an averaging technique that I would recommend and certainly not what I do.
In what frequency range(s) do you notice this effect? Is it more and more noticeable all the way up, or just thru a certain range?
It's not that I "notice this effect", I am talking about the results of psychoacoustic research into group delay and its audibility. Group Delay has been shown to be most audible above 1 kHz, peaking at about 3 kHz and falling off after that. No studies have shown any significant audibility at frequencies below 500 Hz, which makes sense given how we hear differently above and below that frequency point.
Is pure mathematics...
Doing a complex average in FR is the exact same as applying a time-window in IR
Fourier-tranforms tell us that
Doing a complex average in FR is the exact same as applying a time-window in IR
Fourier-tranforms tell us that
I think in theory the larger and more obtrusive you make a speaker the more the imaging will cling to the speakers and give you poor localization. At least that is my understanding. I have to admit I get a bit of this with my box speakers but at times I can still sort of ignore them like late reflections.
Ask, could I request that you show that "mathematically" with a proof, so that I can see what you claim that you are doing. I'm not quite a novice at math and I don't see how you would prove your contention.
And what do you mean by a "complex average"? That could mean a lot of things and they would all be different.
Take for example an impulse followed by another impulse at a lower level some ms behind. At some point the window will exclude the second impulse, but a FR average, complex or not, would never exclude if it were within the original gating window that the FR is based on.
Thanks - I know that it is a lot to Ask (pun intended) but I really do want to understand why you claim this to be true since I cannot see it working in practice. I will look forward to it when you get the time. Maybe I'll do some examples of when I don't think that it will work.
Then let me ask this: Is this common practice?
I do my "averaging" the old fashion way as a true spectral data average across a 1/6 octave bandwidth, using all of the data within the time window. My data always seems to look different than what other people get on the exact same products. Theirs usually looks better. But if we are using two different techniques and they don't give the same results then this explains a lot. Even now I do not use HolmImpulse for the averaging.
This is precisely why I do my own data analysis. I know that what I am doing is correct and I don't know what goes on behind the screen in other software.
So am I
So am I - found a factor of 2 in my smoothing - something to do with FWHM.
What I have had trouble with, is that there is no standard shape for the 1/N octave filtering in the frequency domain.
I am looking forward to present my ideas, formulas, graphs
So am I - found a factor of 2 in my smoothing - something to do with FWHM.
What I have had trouble with, is that there is no standard shape for the 1/N octave filtering in the frequency domain.
I am looking forward to present my ideas, formulas, graphs
Pretty far OT, but I think this is probably largely related to the frequencies involved in the diffraction. A larger baffle will push this more into the midrange. This can be reduced with your normal baffle diffraction treatments- asymmetry, felt, large roundovers and bevels.
The big speaker does also tend to mean bigger drivers, which means a big moving surface closer to the midrange driver. It's one of the many cases for putting a 'subwoofer' in that style of speaker, to the rear or side. Who knows what combining diffraction and intermodulation does for audibility?