- Joined
- Oct 25, 2019
- Messages
- 12,390
- Likes
- 17,770
Pretty sure there is an existing thread with almost exactly this title. Suggest browsing for it.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Room correction, speaker correction anything above Schroeder a mistake?
- Thread starter Blumlein 88
- Start date
After looking harder I found this.
Room correction, speaker correction anything above Schroeder a mistake?
Discuss. The Schroeder frequency is the boundary between reverberant sound and a region dominated by widely space room modes. Obviously not a fine exact line. Below the Schroeder frequency the room responds with areas of resonance and nulling which will depend upon the location of the source...
www.audiosciencereview.com
I'll ready though it.
Since our hearing gets poorer as the frequency increases would you recommend changing a REW frequency graph reading to variable so that the peaks and troughs become more of a psychoacoustic estimation in the highest frequencies?
ernestcarl
Major Contributor
I think the more apt descriptor is that it's "tricky" to EQ above Schroeder -- not counting the more common broad type HF shelving filters nearly all studio professional monitors have built-in.
syn08
Senior Member
At the first glance, I read "Schroedinger frequency" and I could not stop a shiver 
.
.
EQing above Schroder is useful to get L/R speakers to match better, in the presence of room interactions. This directly impacts imaging performance; the better the speakers are matched to one another, the more precise the imaging. This is quite obvious, IMO; if L/R is 100% identical and a tone is played, you get a perfectly centered image. If one channel is louder than the other evenly, the image shifts towards the louder speaker. However, if one channel is unevenly (across the freq range) louder/quieter than the other, then the image smears; some portions of the sound go towards one speaker, and others go to the opposing speaker. Room interactions/reflections can affect this behavior for the different speakers, and room correction could be used to reduce this effect.
As to meeting a particular target curve or correcting some phase response, I'm not quite as sold on that part, other than for some basic tone control/correction (i.e. tame a bright speaker).
Below is some analysis I had done when I was evaluating Audyssey XT32 and Dirac Live in my system. Both DRCs improve L/R matching vs. no correction for SPL. Each line is the average of 6 measurement points around the MLP.
Additionally I looked at L/R phase error, which is also something that is audible (i.e. flip the phase of one of your speakers, you'll hear a difference). DRC improves upon it, although I found Dirac doing something weird here and increasing the error in the mid range. Subjectively I did not like the way Dirac sounded above the Schroder frequency; the imaging got more precise but the soundstage sounded like it reduced. Audyssey, on the other hand, sounded better on than off above the Schroder frequency. The average error numbers are calculated on linear scale, but when viewed in log scale, it looks like Dirac is worse than No Correction.
As to meeting a particular target curve or correcting some phase response, I'm not quite as sold on that part, other than for some basic tone control/correction (i.e. tame a bright speaker).
Below is some analysis I had done when I was evaluating Audyssey XT32 and Dirac Live in my system. Both DRCs improve L/R matching vs. no correction for SPL. Each line is the average of 6 measurement points around the MLP.
Additionally I looked at L/R phase error, which is also something that is audible (i.e. flip the phase of one of your speakers, you'll hear a difference). DRC improves upon it, although I found Dirac doing something weird here and increasing the error in the mid range. Subjectively I did not like the way Dirac sounded above the Schroder frequency; the imaging got more precise but the soundstage sounded like it reduced. Audyssey, on the other hand, sounded better on than off above the Schroder frequency. The average error numbers are calculated on linear scale, but when viewed in log scale, it looks like Dirac is worse than No Correction.
Last edited:
The bolded part is triggering a thought...I am not sure we are talking about the same thing. The topic to search for is Critical Bands. This is the bandwidth of auditory filters. See https://en.wikipedia.org/wiki/Critical_band
What we are discussing here is the difference between F2 and F1. As the center frequency goes up (Fc), that differential gets wider. That relationship is indeed linear with frequency. See this for example from Zwicker & Fastl, "Psychoacoustics, Facts and Models" book:
View attachment 9846
Above 500 Hz the relationship is almost linear.
In lay terms then what this says is that we cannot discriminate sharp drops and peaks at high frequencies as we do on low frequencies. So we better filter our measurements to reflect the same.
For this reason, when looking at low frequency response of the room, we want to use no filtering or 1/12 octave or better. At the extreme of our hearing, we probably want to be at 1/3 to 1/6 octave to roughly match the discrimination of our hearing.
Why do we care so much how a speaker measures over 500 Hz if our hearing cannot perceive the "drops and peaks" above that frequency? I might be missing something, but is the linearity of the speaker above the Shroder frequency going to be 'lost' on my hearing? If so, shouldn't we worry more about the linearity of the speaker below said frequency (and I don't mean worry with DSP... I mean in general). I am not suggesting the curve is meaningless... just that small variations that we would have trouble discerning might also be a similarly small variation of different speaker measurements over the same frequency range.
And I also can't hear over 14,500 Hz... so add that into the mix.
Thanks.
The bolded part is triggering a thought...
Why do we care so much how a speaker measures over 500 Hz if our hearing cannot perceive the "drops and peaks" above that frequency? I might be missing something, but is the linearity of the speaker above the Shroder frequency going to be 'lost' on my hearing? If so, shouldn't we worry more about the linearity of the speaker below said frequency (and I don't mean worry with DSP... I mean in general). I am not suggesting the curve is meaningless... just that small variations that we would have trouble discerning might also be a similarly small variation of different speaker measurements over the same frequency range.
And I also can't hear over 14,500 Hz... so add that into the mix.
Thanks.
Well, the 'drops and peaks' aren't audible with broadband signals,but if I sweep a sine slowly over the peak, you'll notice! There really are many more constraints than show up in most discussion. The issue is signal dependent to an extreme.
I'm giving that talk on room design this Wednesday for the AES local section https://pnw.aessections.org/ I won't talk about the correction part in great detail,but you can read about that in much older talks by myself and Serge Smirnov (which you can find in "meeting recaps" in the old site for now (www.aes.org/sections/pnw) although that will eventually vanish.
This one has not disappeared in my archives .
http://www.aes-media.org/sections/pnw/pnwrecaps/2008/jj_jan08/
http://www.aes-media.org/sections/pnw/pnwrecaps/2008/jj_jan08/
It's URL is very likely to eventually change. The AES is changing everything in their web setup.
More like 2kHz. It's a hearing question more than an acoustic question, and there are some very notable issues about the "schroeder frequency" idea, as well.
Keith_W
Major Contributor
Room correction and speaker correction are not the same thing. If you can get a quasi-anechoic measurement of the speaker, measure the listening window, etc. then you can go for it all you like all the way up to 20kHz. But once the speaker is in a listening room and the microphone at the listening position, only the bass can be corrected and the top end adjusted with a gentle tilt.
Again... sorry if I am beating a dead horse, but this goes back to my original question. Why is there no correction over 500 Hz? If we can't perceive the nuanced differences that the mic can hear, why do we care about the linearity of a measurement (within reason)? If we are close and we can't notice the slight variation, what does it matter how linear our result is if we can't really distinguish minor variances? Now... if the answer is, it is signal-dependent and you can adjust minorly above the Schroeder frequency, then ok. It is just very veiled and difficult to understand.
I'm afraid we must disagree, unless your "gentle tilt" is more than 1 or 2 dB per ERB.
Digital_Thor
Addicted to Fun and Learning
Isn't the tilt just a product of the easier absorption of higher frequencies in normal rooms, and the narrowing of dispersion in most speaker designs as frequency goes up?
Because when I measure my speaker in the middle of the room - quasi anechoic - and EQ/filter it flat. I get a gentle tilt, as I put the speaker back in place and measure in my listening position.
I just read JJ's presentation. Makes mostly good sense to me, even though I only used IIR filters in my own system.
The room curve with the tilt is simply what listeners tend to prefer at the main listening position.
How you get to it, is up to you.
If the speaker already measures 'flat' and otherwise well in a Klippel/chamber, it's what you get at the MLP if the acoustics of your room at the MLP are optimal.
That's my simple understanding.
How you get to it, is up to you.
If the speaker already measures 'flat' and otherwise well in a Klippel/chamber, it's what you get at the MLP if the acoustics of your room at the MLP are optimal.
That's my simple understanding.
Similar threads
