'LGT' Construction Diary

[uli.brueggemann]

I like to show what is possible today. The pictures are not smoothed. Measurements at the listening position.

Shown is the behaviour of a B&W 801 D speaker before and after optimization by FIR filter with minphase and excessphase correction (filter size 65536 taps).

So first: Amplitude response shown


Uli

[uli.brueggemann]

Second the step response

Uli

[Salas]

Nice step response and room modes correction. But why aiming for flat at listening position? Gentle slope is the natural sound in the far-field.

[Wingfeather]

Hi Uli,

Those measurements look impressive. Could you comment a little on how they were taken?

e.g.
swept sine/MLS?
What was the room like? Anechoic? 'Normal' domestic room? Size?
Speaker placement?
Speaker distance from listening position?

Thanks,
Matt

[uli.brueggemann]

Nice step response and room modes correction. But why aiming for flat at listening position? Gentle slope is the natural sound in the far-field

There is a slope given in the target curve. So why do you not see it

If you study room correction you will see that all the corrections are typically done by smoothing or averaging the frequency response. Quite often you only see only this response but not the unsmoothed FFT result.

You have to know that of course the inversion of an unsmoothed curve is simply not possible (you also do not wnat to get nailed at the listening position).

Typical smoothings like e.g. 1/n-octave analysis creates errors, the curve is pulled down by the dips of the unsmoothed response. Thus e.g. with TacT you find even targets with quite steep slopes.

In this case a psychoacoustic analysis has been applied. And the target follows the natural behaviour of the result of the analysis which is typically a slope of -3 dB falling at the right side.
So the correction as the result of target minus the interpreted measurement is more or less at level zero.

At the end this results in the shown curve. And there is nothing wrong when listening. No bright or harsh sound, no nasty digital sound.

The unsmoothed response looks more or less flat though. But no problem.

Uli

[uli.brueggemann]

swept sine/MLS?

logsweep signal

What was the room like? Anechoic? 'Normal' domestic room? Size?

Normal room, not anechoic.

Speaker placement?

in this case a triangle (60° angles) of about 2.5 m size

Speaker distance from listening position?

about 3 m


Uli

[Salas]

Originally posted by uli.brueggemann




There is a slope given in the target curve. So why do you not see it

If you study room correction you will see that all the corrections are typically done by smoothing or averaging the frequency response. Quite often you only see only this response but not the unsmoothed FFT result.

You have to know that of course the inversion of an unsmoothed curve is simply not possible (you also do not wnat to get nailed at the listening position).

Typical smoothings like e.g. 1/n-octave analysis creates errors, the curve is pulled down by the dips of the unsmoothed response. Thus e.g. with TacT you find even targets with quite steep slopes.

In this case a psychoacoustic analysis has been applied. And the target follows the natural behaviour of the result of the analysis which is typically a slope of -3 dB falling at the right side.
So the correction as the result of target minus the interpreted measurement is more or less at level zero.

At the end this results in the shown curve. And there is nothing wrong when listening. No bright or harsh sound, no nasty digital sound.

The unsmoothed response looks more or less flat though. But no problem.

Uli

Do you have a 6th octave smoothed one to post? Good you follow a slope. Thanks for the explanation, because I could not possibly know if it pulls down when you smooth. I am no CPU.

[Wingfeather]

Hi Uli,

in this case a triangle (60° angles) of about 2.5 m size

Surely at this kind of distance (unless your room is pretty huge), the first reflection won't be long behind the direct sound. Yet your uncorrected impulse response doesn't seem to show any room effects at all - it looks very close to the electrical output you'd measure from a three-way minimum-phase crossover! (I'm referring to the three separate peaks - one for each driver).
That would be a good impulse response to get if you'd taken it in a very large anechoic chamber, let alone an average untreated room.

Did you take any special precautions to minimise room effects?

[uli.brueggemann]

Do you have a 6th octave smoothed one to post? Good you follow a slope.

No, I do not have a 6th octave smoothed curve as I do not use this kind of smoothing. Appended a picture of the unsmoothed amplitude response (red), treated by a psychoacoustic calculation and then by a frequency depandant windowing (in this case the width is 15 cycles, this means e.g. 150 ms at 100 Hz, 15 ms at 1 kHz and 1.5 ms at 10 kHz) [green curve]
Anyway you see that the curve does is not influenced by the dips.

Also shown is the target (blue).

Hope this helps

Uli

[uli.brueggemann]

That would be a good impulse response to get if you'd taken it in a very large anechoic chamber, let alone an average untreated room.

Please note that the shown picture is a step response which shows the integral of the pulse response.
Appended a comparison of pulse response (red) and step response (green) in one picture.

You can see that reflections in the pulse response are in a way smoothed by the integration.

So it is always necessary to clearly tell what a curve shows.

Uli