Some comments from Floyd Toole about room curve targets, room EQ and more

[dlaloum]

I don’t fully grasp the meaning of “minimum -phase” behavior of rooms and why transducers are intrinsic minimum-phase devices. All in the light of the ability of room EQ to improve the room response.
Are there any good articles or books on this topic?

Most phenomena in nature are minimum phase.

Standard resonances, both mechanical/acoustic and electrical, are all minimum phase.

Hence correction should also use minimum phase, which has a corollary of a symmetric relationship to frequency response.

In a fully minimum phase environment, correcting f/r using minimum phase filters will also correct phase, and vice versa.

Hence for most things audio, it is best to stick to minimum phase filters.

 

[René - Acculution.com]

You can have for example mechanical or electrical systems that are NMP, but to put it very non-stringent, they typically take more effort to make.

 

[Tim Link]

Harman Target doesn't apply to my room, I thought.

With a DIY speaker system like my own and no way to test it anechoically or with a Klippel thing, I'm shooting in the dark somewhat. Lately it's been sounding pretty darn good to me so decided to see where I was at in terms of a room curve. I've been shooting for flat with a 7 cycle window at the listening position, with some rise in the bass, ear lobing it, knowing that at 7 cycles more and more bass is leaking in to the room as it goes down.

I thought I didn't have any use for that Harman Target, but to my surprise, other than the lack of a bass drama hump, I'm pretty darn close at 500ms!

The earlier direct sound rules. Something was bothering me about my setup for quite a while and I didn't understand it until I started taking 7 cycle windowed response at the listening position. The crossover zone measured good with 500ms, but when narrowed down there was a sizeable dip around 1k. My drivers are spaced far apart so I had to measure at the listening position to really understand what was happening.

 

[thewas]

Harman Target doesn't apply to my room, I thought.

As written in the first posts of this thread there is no real Harman target for loudspeakers except to have flat direct sound which is what you also preferred. The "preference curves" that exist are results for specific rooms and depend on the loudspeakers directivity, listening room absorption, listening distance etc.

 

[krabapple]

Harman Target doesn't apply to my room, I thought.

With a DIY speaker system like my own and no way to test it anechoically or with a Klippel thing, I'm shooting in the dark somewhat. Lately it's been sounding pretty darn good to me so decided to see where I was at in terms of a room curve. I've been shooting for flat with a 7 cycle window at the listening position, with some rise in the bass, ear lobing it, knowing that at 7 cycles more and more bass is leaking in to the room as it goes down.

I have no idea what this means, can you explain? And how is low bass leaking *in* to the room?

 

[Tim Link]

I have no idea what this means, can you explain? And how is low bass leaking *in* to the room?

I should have said "bass leaking in to the room measurement."

I'm setting an impulse response window on my measurement at the listening position to make it more like an anechoic measurement. 7 cycles means the window of time that's allowed into the measurement is 7 wave cycles, so at 1000 Hz that's 7 milliseconds, so I only get in the measurement what's reached the microphone in the first 7 milliseconds at that frequency. By 500 Hz it's 14 milliseconds, which means more reflections are getting included because my room is small. By 100 Hz it's 70 ms, which means a lot more reflections are getting included. So, more sound power reaches the microphone as the frequency goes down, and the measurement becomes less anechoic. The goal is to get an anechoic response that's flat on axis. But since I don't have an anechoic environment to test in, reducing the timing window to 7 cycles gives me a closer idea to the anechoic response, but becomes less accurate as the frequency goes down. If I turn the timing window down to below 7 cycles the analysis becomes increasingly less accurate, so without some more sophisticated trickery of multiple measuring distances like a Klippel machine can do it's not easy to get an anechoic response in the bass. I know more energy from reflections is making it into the measurement in the bass, so I have to shoot for a rising response of some kind as measured to get close to what a true anechoic flat response would be.

One way to attempt it is to measure right up close to the woofer. If you just have one fairly small woofer this can be of some use. In my case, I've got stacks of woofer horns in each corner, which produces surprisingly different listening position response than when measuring up close. If I EQ the woofer horns flat up close, the response at the listening position has a surprising upward tilt towards their upper cutoff frequency, I think due to the increasing directivity of the signal as the frequency goes up with such a large radiating surface that's 2' wide and floor to ceiling height. at 500Hz that's nearly a full wavelength width. I run them up to about 300 Hz or 262 Hz more precisely seems to be where the phase and amplitude integrates most neatly with the midrange horns.

Honestly, I'm not sure how to make sense of it, because the increasing directivity should cause the total sound power in the room to go down, not up. I'll have to think more about that. It's probably more complex than my explanation. It could be that my room is leaking lower bass faster than it is upper bass. I've looked at decay times and that doesn't seem to be the case. The upper bass is actually decaying faster, around 300ms RT60, while the lower bass is much longer. Also clarity is higher in the upper bass, which means the ratio of early to late energy is higher.

 

[Beershaun]

I should have said "bass leaking in to the room measurement."

I'm setting an impulse response window on my measurement at the listening position to make it more like an anechoic measurement. 7 cycles means the window of time that's allowed into the measurement is 7 wave cycles, so at 1000 Hz that's 7 milliseconds, so I only get in the measurement what's reached the microphone in the first 7 milliseconds at that frequency. By 500 Hz it's 14 milliseconds, which means more reflections are getting included because my room is small. By 100 Hz it's 70 ms, which means a lot more reflections are getting included. So, more sound power reaches the microphone as the frequency goes down, and the measurement becomes less anechoic. The goal is to get an anechoic response that's flat on axis. But since I don't have an anechoic environment to test in, reducing the timing window to 7 cycles gives me a closer idea to the anechoic response, but becomes less accurate as the frequency goes down. If I turn the timing window down to below 7 cycles the analysis becomes increasingly less accurate, so without some more sophisticated trickery of multiple measuring distances like a Klippel machine can do it's not easy to get an anechoic response in the bass. I know more energy from reflections is making it into the measurement in the bass, so I have to shoot for a rising response of some kind as measured to get close to what a true anechoic flat response would be.

One way to attempt it is to measure right up close to the woofer. If you just have one fairly small woofer this can be of some use. In my case, I've got stacks of woofer horns in each corner, which produces surprisingly different listening position response than when measuring up close. If I EQ the woofer horns flat up close, the response at the listening position has a surprising upward tilt towards their upper cutoff frequency, I think due to the increasing directivity of the signal as the frequency goes up with such a large radiating surface that's 2' wide and floor to ceiling height. at 500Hz that's nearly a full wavelength width. I run them up to about 300 Hz or 262 Hz more precisely seems to be where the phase and amplitude integrates most neatly with the midrange horns.

Honestly, I'm not sure how to make sense of it, because the increasing directivity should cause the total sound power in the room to go down, not up. I'll have to think more about that. It's probably more complex than my explanation. It could be that my room is leaking lower bass faster than it is upper bass. I've looked at decay times and that doesn't seem to be the case. The upper bass is actually decaying faster, around 300ms RT60, while the lower bass is much longer. Also clarity is higher in the upper bass, which means the ratio of early to late energy is higher.

Curious. What are you attempting to accomplish?

 

[Tim Link]

Curious. What are you attempting to accomplish?

I'm hoping to make my system sound as good as possible, assuming that flat anechoic response at the listening position is a good thing, and that smooth off axis response is also a good thing. Windowing down to 7 cycles works pretty good above 1000Hz. So if you've got flat response down to 1000Hz at 7 cycles, then compare that to the slope with the window wide open you can get an idea of what the slope of your room should be. I guess I shouldn't be too surprised that my typical domestic listening environment has a fairly typical slope. I'm using a BL-409 horn I got off Parts Express and I've never seen any directivity data on it. My test suggests it's pretty decent, without me having to make hundreds of measurements. If the open window slope were really wacky it'd suggest some directivity problem, or stange absorption issues in my room.

If we're going to EQ towards a target curve for a room we should be able to verify that the anechoic response is flat when that whole room curve is applied. My speaker system isn't something that's ever been properly tested so I have to try to infer what's going on with it based on whatever methods I have at my disposal. My previoius way of doing it was to just EQ flat up close and assume that the anechoic response held at distance. It would, assuming a uniform directivity and fall of rate. No speaker seems to have that, especially not large horns, so the bass power falls off faster with distance than the upper frequencies. This shouldn't have to be the case for even a directional speaker if all the frequencies emerged with a close to spherical wave front. With enough nearfield measurements it should be possible to determine the frequency response at any distance and any angle off the speaker, I suppose.
Wouldn't that be neat for a speaker review, a web widget that lets you pick any distance and angle off axis from the tweeter and gives you an anechoic response at that distance? Does the Klippel machine measurement have all that data in it? I suppose if it can calculate one position it can calculate any other.

 

[Beershaun]

Why try to separate and eq the direct and reflected responses? What you hear is the sum correct? Are you trying to tune your speakers themselves mechanically? Or the final in room response?

 

[Tim Link]

Why try to separate and eq the direct and reflected responses? What you hear is the sum correct? Are you trying to tune your speakers themselves mechanically? Or the final in room response?

First direct sound dominates. If you have a speaker that's already been tested anechoically then you don't want to EQ for it's direct response. My DIY speaker has not been through that process. I have to get the crossovers set and the driver/horn combinations EQd, and then EQ for the room modes and reflection issues.