Loading Profile...
Smoothing and Bark scale
Hi,
It’d be great to add one octave to the list of smoothing resolutions available in the FR graph.
Also, some kind of a psycho-acoustic scale would be even better. I suggest the Bark scale (Zwicker) where the resolutions go like this :
< 150 Hz : 1 oct
150 - 500 : 1/3 oct
500 - 4 kHz : 1/6 oct
> 4 kHz : 1/3 oct
Zwicker’s work is not about JND in frequency, but about the perception of the resulting level from the accumulation of sounds close to one another in frequency (sounds within the same “critical band”). In other words, it doesn’t mean that one won’t hear the difference, in pitch or level, between two notes a minor third apart played by the bass player; nevertheless, on the overall spectrum, when many sounds combine, a frequency response will be perceived as mostly flat when the above resolution reads it as such, even though a coarser resolution would reveal deviations.
What do others think?
Any way to implement this in Fuzz?
For those interested, the paper that is most often referred to about this (the article that started it all I guess) is:
Zwicker, E: Subdivision of the Audible Frequency Range into Critical Bands (Frequenzgruppen), Letters to the editor, JASA, Vol 33, n.2, Feb 1961
It’d be great to add one octave to the list of smoothing resolutions available in the FR graph.
Also, some kind of a psycho-acoustic scale would be even better. I suggest the Bark scale (Zwicker) where the resolutions go like this :
< 150 Hz : 1 oct
150 - 500 : 1/3 oct
500 - 4 kHz : 1/6 oct
> 4 kHz : 1/3 oct
Zwicker’s work is not about JND in frequency, but about the perception of the resulting level from the accumulation of sounds close to one another in frequency (sounds within the same “critical band”). In other words, it doesn’t mean that one won’t hear the difference, in pitch or level, between two notes a minor third apart played by the bass player; nevertheless, on the overall spectrum, when many sounds combine, a frequency response will be perceived as mostly flat when the above resolution reads it as such, even though a coarser resolution would reveal deviations.
What do others think?
Any way to implement this in Fuzz?
For those interested, the paper that is most often referred to about this (the article that started it all I guess) is:
Zwicker, E: Subdivision of the Audible Frequency Range into Critical Bands (Frequenzgruppen), Letters to the editor, JASA, Vol 33, n.2, Feb 1961
5
people like this idea
I like this idea!
Tell me when this idea gets some attention.
The more people who like this idea, the more it gets noticed.
The more people who like this idea, the more it gets noticed.
Create a customer community for your own organization
Plans starting at $19/month
-
Inappropriate?Please excuse the multitopic reply. I'm adding a bunch of related requests in one reply!!!!
ON OCTAVE SMOOTHING:
I DEFINITELY like this idea! For me, 1 octave smoothing is my first request. Bark scale smoothing---possibly, I'd have to investigate it to see which is more accurate for my purposes, which is "voicing a subwoofer" and "determining a subwoofer level" in the listening room. I do like the variable smoothing, it's much more functional than the options in Spectrafoo and produces a much more readable yet accurate trace.
I would like to have 1 octave smoothing in order to help center the trace vertically in the midrange as I think judging the average midrange level between about 250 and 1 kHz in octaves (or barks) matches human perception of speaker voicing pretty well. After I "normalize" the trace vertically this way, I then use it to judge the subwoofer level.
ON NORMALIZING TO A CENTER LINE:
Speaking of this issue, I really would like to be able to manually shift the frequency response trace up or down (on a stationary graph) with the mouse after they have been shot. This would eliminate the cumbersome need to tweak the mike preamplifier gain or similar in order to have the graph arrive and be centered on a given major dB line. Can this please be added to the next version? I know that the Y offset can be entered manually, but it is cumbersome and I usually end up tweaking the interface mike preamp gain to get the centerline I want for the most convenient and repetitive measurement.
Related to that, I find the "normalize" function an interesting idea but fairly useless unless the algorithm can sense the peaks and dips within a certain frequency range and normalize them to a preset dB line.
ON THE GRAPH MARKINGS:
In addition to the major lines every 10 dB, can we please have a minor division (semi-bold) every 5 dB? This would make adjusting gain to center a trace much easier.
ON THE ANALYSIS WINDOW:
Every time I take a new measurement, the analysis window is reset to its maximum. I would like it to remember the last window until the next time I change it. For example, I want the window to start at a particular time which is just before the first arrival, and end 100 ms later so as to have decent low frequency response accuracy and still reject later noise in the room.
BUGS IN THE DB SPL SCALE?:
The dB SPL scale seems to work, but even if you change the sweep level, the dB trace always comes back at the same SPL level. This worries me. I'd like to change the excitation level. Well, probably I can do that externally in the interface, but we must remember not to touch the sweep level once everything has been calibrated.
Also, if you switch to dB SPL scale after previously having the graph in the normal dB scale, I think it produces meaningless results. Perhaps that's an expected weakness and I shouldn't expect it to be correct?
RELATED TO THAT, CAN WE HAVE A "WISH LIST" TOPIC ON THIS FORUM, FOR NEW FEATURES OR IMPROVEMENTS THAT WE WOULD LIKE TO SEE? -
Inappropriate?Bob: It's very difficult to reply to these point-by-point here. At the top level of the group, you can "Share an idea" with each one of these points for further discussion.
All of these are great points, though. The only one I'll address right here & now is:
ON THE ANALYSIS WINDOW:
You can choose Lock Graph Extents in the Impulse menu to cause it to 'stick' where it is. Same goes for the frequency graph.
I’m overwhelmed by ideas!
-
Inappropriate?Dear Chris: Thanks for your nice development of Fuzz Measure. I use "Lock Graph Extents". I do not believe this makes the analysis window come up the same as the last time, it only makes the viewed magnification and time in the impulse window to remain the same. For example, I want the ANALYSIS WINDOW to remain at, say, 5 ms through 105 ms for the next measurement I take, at least until I change it. I hope this clarifies what I'm talking about.
EMPLOYEE
