Originally Posted by jj_0001
I'll let Amir translate that for you
since he volunteered the other day.
I did? I don't recall but if I said that I can translate your English to my English, it must have been a moment of insanity! I will try below.
Flat response about 200Hz can also be a big problem, because in fact room contribution falls off with frequency, and you can wind up, depending on the window you measure with, a system that is way, way too bright at high frequencies.
JJ, I assume you mean "above 200 Hz" as opposed to "about 200 Hz." If so, indeed the room contributions drops off substantially as you get above the transition frequencies of 200 to 300 Hz. You can see this (indirectly) from this measurement of the same room in different places in the room from my article on bass frequency optimization
As the graph indicates, above transition frequencies the speaker response dominates. The room has little effect there. So if you are not happy with that response, it is an issue with your speaker, not the room.
The second part of JJ's comment is that if you have a flat response from low to high frequencies, subjectively your system will sound bright, i.e. too much high frequencies. So you want a "target response curve" that is sloping down like this implemented by JBL Synthesis ARCOS automatic EQ system (this is for our theater):
Now, do we know what the talent heard when he/she approved the content? No. Maybe they heard the bright version and that is what they liked. Maybe not. Audio is "broken" in that we have no reference as to what is right. We have to go by what sounds good and research shows that the above or some variation of it is. I can provide references if there is interest.
Unfortunately automatic EQ in consumer AVRs rarely if ever lets you select the target response. A crude way around that is to boost the level of your sub (assuming you have one) a few dbs post calibration.
What you want to measure above about 500Hz is the DIRECT COMPONENT of the speaker, as determined not by the room, but by the ear, i.e. you need to use a frequency-varying window with different time lengths at different frequencies.
I hope this part is self-explanatory. If not, JJ is saying to vary the time window over which you measure the frequency response to match what we hear, devoid of the contributions from reflections. Alas, no standard measurement system does this.
I use an alternate method which is to apply increasingly higher amount of averaging as frequencies go up. This is because the frequency selectivity/bandwidth of the ear goes up as frequencies go up. I explain this in my article on perception of room reflections
with this graph related to this topic:
This means that if you are measuring bass frequencies, you want to have high precision down to 1 Hz detail so that you can find the room contributions/modes. But if you are trying to determine what is going on say, above 1 Khz, you want to use averaging as to better match our perception there.
If you are using REW for example, you want to use 1/24 or 1/12 smoothing for bass frequencies but go down to 1/6 or even 1/3 for higher frequencies.
At 200Hz, you still want to look at direct, but you want to actively consider room bounces, because you can't be rid of them. Still a window no longer than 1/30th of a second is appropriate.
200 Hz is in the transition frequency meaning you have not gotten rid of room modes but they are starting to overlap good bit as to randomize the response. So you want to look at the peaks and valleys but don't go nuts analyzing them at high detail.
Below that, the room is primary, and you need to look at a much wider window, maybe up to .125 seconds at 90Hz, and up to a second at 20Hz.
As I showed in the first graph, the room determines the response, not the speaker below transition frequencies. You can put any speaker in your room and the response more or less will be the same. Your measurements need to be accurate down to 1 Hz or even lower. JJ is using the time version of the same metric which goes up, as frequency selectivity narrows.
But then you also want to know the room power response, as opposed to the pressure response at one point, below about 100 to 200Hz anyhow.
Same thing I explained above