How to judge graphs against each other?

[hilde45]

Having established an optimal LP and Speaker placement, I am actively taking many sweeps of my room with REW to judge the efficacy of different treatments. This is a question about how to compare SPL graphs against each other and how to compare IMPULSE graphs against each other.

Regarding IMPULSE, I understand that a critical range of impulse responses in my room lies between 3-13 ms or so, and that if a peak is below -20db, it likely won't be heard. So, after many sweeps under varying conditions, I'm trying to judge the relevant peaks of one impulse graph against another. In some cases, it's easy to see right away whether one graph is better than another; as I overlay them, I see more peaks, higher peaks, etc.

But in other comparisons between Impulse graphs, it's not clear which is better. QUESTION: Is there a tool in REW for estimating the area-under-peaks for a particular scan? This would help me quantify how one delimited impulse graph stacks up against a (close) competitor.

Regarding SPL graphs, here I have a similar problem. I recognize that in treating a room, the key is to get the bass region correctly, first. 20 Hz - 300 Hz or so. Looking an "all SPL" graph containing many sine waves, I see (for example) that Scan 1 has a big null at 135 and a modest null at 180 and Scan 2 has the reverse. Or Scan 3 has a big peak at 80 but modest nulls at 135 and 180 and Scan 4 has a modest peak at 80 and bigger nulls at 135 and 180. Which scans are better? That's the conundrum.

I realize that I may just have to listen to music at these various, competitive condition to tell. I also realize that some scans may be differentiated by how easy or hard to treat one problem is vs. another. I ask this "how do I judge one scan against another" not because listening or treatment are not relevant tools, too -- I get that -- but because I need to at least use the REW graphs to narrow down my options, if possible.

Thanks for any suggestions or advice.

 

[JLM1948]

Actually you cannot compare graphs and determine which is the best.
Big narrow nuls are impossible to EQ, but you may find that moving the mic shift them or even cancels them.
When tuning a room, you don't want to tune one single response, you want to make the average room response correct. That means you must estimate the average response and evaluate the problems that are dominant, those that do not shift otrdisappear when you move the mic. It's those you must treat or EQ.
That's what I did when I treated rooms for a living, a few decades ago, using an audio wobbulator, the size of a small TV.
I don't use REW for acoustics, but I believe there is an averaging option in it. You must consider putting more weight on the measurements made close to the preferred listening area.

 

[hilde45]

Thank you, JLM. One suggestion I got recently about comparing nulls or peaks within a graph is to try to correlate them with which note they correspond to. If a null, for example, is at a place where there is no note in our tuning scale, it is less important to fix compared to another similarly sized null which does correspond to a null. This is, of course, after the maximum number of adjustments to reduce them.

 

[JLM1948]

Please allow me to think it's ridiculous. The actual pitch of many pieces of music is not exact, because of deliberate or accidental changes in playback speed. Also the concert pitch has changed considerably over the years. Check

Concert pitch

en.wikipedia.org

.

 

[hilde45]

Permission to think that: granted.

Good point.

 

[Wayne A. Pflughaupt]

Adding to what JLM noted, it should be noted that frequency response plots generated with sine wave sweeps merely map the fundamentals. But fundamentals are only one part of the music reproduction picture. Notes from musical instruments are extremely complex tones comprised of not just the fundamental, but also multiple-order harmonics, sub-harmonics, overtones, etc. That’s why an F# on a tenor saxophone sounds nothing like the same note on a piano, even though both have the same fundamental.

Thus, just because you happen to have a null, that doesn’t necessarily mean the note will totally vanish.

Regards,
Wayne

 

[JLM1948]

Indeed. The numerous attemps at trying to establish a bi-univocal relation between frequency and musical perception have all failed.

http://acousticslab.org/RECA220/PMFiles/PMImages/frequencyranges.jpg