What is the ideal spinorama "estimated in-room" average line tilt?

[73hadd]

I am describing "tilt" as the angle of the averaged line for the estimated in-room response.

I have read that it is a design goal for the in-room treble response to "not exceed the midrange" and we see that result in some speaker measurements.

Revel M16 "tilt" 6db

KEF R3 "tilt" 8db.

KEF Reference 5 "tilt" 10db


1. Since complete flatness is not a goal, how much tilt is too much? Thoughts?
2. Many speakers show a dip of some sort between 1k-3k. Is that better or worse (subjectively) than a smoother but more aggressive tilt?

Thank you.

 

[Blumlein 88]

Someone else can give you the best or ideal tilt. It is probably near 1 db/octave as this would be near 3db per decade. (.9 db/octave is 3 db/decade)

Do note the speaker response ideally in an anechoic chamber will be flat on axis. The in room measured response of such a perfect speaker would result in a tilt.

The dip in the 1-3khz range is usually due to the crossover and directional changes between woofer and tweeter or midrange and tweeter. Ideally you wouldn't have that. Smoother would be better.

 

[73hadd]

Thanks! I see that some have the "tilt" on the anechoic on axis result, and I think that is strange, as I agree with you that the in-room result will have the desired tilt. Anechoic tilt may result in maybe too much tilt later. (i.e. subjective "not enough openness")

 

[thewas]

It probably depends on the room and listening distance, I used to use the classic 1 dB/octave target but nowadays I rather prefer a 0,8 dB one which is also closer to what Toole plots at his AES paper as the predicted steady-state room curve for highly rated loudspeakers in a typically-reflective room and the Dirac default target curve (both around 0,6 dB):

From https://secure.aes.org/forum/pubs/journal/?ID=524 (free download available)

 

[thewas]

2. Many speakers show a dip of some sort between 1k-3k. Is that better or worse (subjectively) than a smoother but more aggressive tilt?

A so called presence dip (often called also BBC dip as it happened often to 2-way loudspeakers without a waveguide) can make mids sound less shouty, especially when they are mixed with such a loudspeaker and thus too much EQ was added by the mastering engineer in this region to compensate that dip, but this way we also continue the unfortunate Audio's Circle of Confusion.

 

[GelbeMusik]

1. Since complete flatness is not a goal, how much tilt is too much? Thoughts?
2. Many speakers show a dip of some sort between 1k-3k. Is that better or worse (subjectively) than a smoother but more aggressive tilt?

There is none. Except it only applies to an average over many control rooms and listening rooms combined, and loudspeaker directivity, combined again. So, consequentially Tool himself advertises the use of tone control. Even with each single recording that could be adjusted for scrutinizing listening

The dip is from the crossover, from constriction by interference in the vertical direction. It's addressed by coaxials, but these trade it against other imperfections then. As far as I understood it right, reading between the lines, Toole shrugs his shoulders about it.

A hint may be in order. Spinorama evaluates a single speaker. With stereo some other peculiarities come up. So, I personally appreciate the approach of Toole to ease not-so-critical listening by adding balanced reveberation with less directive speakers. But don't exaggerate in becoming pedantic about it again.

 

[phoenixdogfan]

I am describing "tilt" as the angle of the averaged line for the estimated in-room response.

I have read that it is a design goal for the in-room treble response to "not exceed the midrange" and we see that result in some speaker measurements.

Revel M16 "tilt" 6db

KEF R3 "tilt" 8db.

KEF Reference 5 "tilt" 10db


1. Since complete flatness is not a goal, how much tilt is too much? Thoughts?
2. Many speakers show a dip of some sort between 1k-3k. Is that better or worse (subjectively) than a smoother but more aggressive tilt?

Thank you.

Good question. I think what you're referring to is the slope of the line for an ideal respsonse. It would be interesting to know, I agree. It might be possible to fit a line with very small deviation. We would say such a line has a good fit which statisticians measure by it's R-squared or correlation coefficient, but that may not mean anything very good if the line has a 120 degree slope! Obviously, the quality of the in room response rests on how close it comes to the optimum slope and how little divergence exists at the indivudual points from the line that has been fitted to describe the curve. (Its linearity).

If there is an agreement on that optimum line is, it might even be useful to put it into the measurement of in room response to see how much the fitted line deviates from it..

 

[richard12511]

It probably depends on your room and speakers, so not sure there's one "correct" target curve. I would experiment with different curves and see what sounds best to you. Mine measures about 8db, which I find sounds very good. I've also found that I enjoy a much steeper slope starting just above 100hz(maybe 120-150ish?), and a much more shallow slope above that. Not sure if that makes sense.

20 0
40 0
70 -1
100 -3
140 -5.75
200 -6.75
300 -7
400 -7
700 -7.25
2500 -8
6000 -8.5
16000 -9
19000 -9

^^^ kinda like that. It's still an 9db slope overall, but it's not just a diagonal line. I attached my right speaker MM measurement. You can see it kinda follows that curve. Currently I'm only correcting up to 400hz with Dirac, so it doesn't slope off as much in the high end as the ideal curve, but oh well.

*Edit - Added my center channel measurement (the blue curve) to better illustrate what I'm going for. The center channel has what I guess would be called the more traditional 9db straight diagonal line "Harman Curve", but I find that I like the other curve better.

 

[edechamps]

It might be possible to fit a line with very small deviation. We would say such a line has a good fit which statisticians measure by it's R-squared or correlation coefficient

This is exactly what SM_PIR in the Olive preference score model is: it is quite literally the r² of a linear least-squares regression line on the Predicted In-Room response curve. Complete with all the headaches that this metric entails.

Olive also states some "target slopes" in the paper. They are defined as "mean slope values of speakers that fall into the top 90 percentile based on preference ratings". For PIR that is b=-1.75 which translates to about -1.2 dB/octave. The paper doesn't explain, however, what effect the slope has on preference, if any.

 

[73hadd]

In the last 24 days of reading here, I have a new context for the question. I had read reviews where they said certain speakers seemed to have a loss of treble (brightness etc.) in non-reflective rooms. My conclusion at this point:

1. A deliberate attempt to engineer a steep slope but smooth PIR doesn't seem to result in actual "preference"
2. Rather, the preference is a result of flat on axis, with good directivity, resulting in a sloped PIR

I have also seen what looks to be the "slope" in anechoic measurements, which I assume to be an attempt to engineer the PIR in advance.

Basically, let the PIR "happen"? Correlation vs. causation.

 

[Jon AA]

Basically, let the PIR "happen"? Correlation vs. causation.

Yes, I definitely fall into the "let it happen" camp and I've seen no research to indicate the opposite will be more preferred. In other words, good on axis, good directivity control results in a good PIR. Using a PIR as a target, especially if it results in less smooth or sloping upward or steeply downward on axis/LW in order to hit an arbitrary PIR target slope is not likely to be preferred.