• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Can an in-room measurement predict what sound is direct vs reflected?

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

Lets assume that anechoic measurements like what is published here are available.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

My Opinion:

Not "predicting", but via in-my-room measurement:

Stretch out an Impulse, or use ETC, is what I think of.

It doesn't tell you frequency, but seems to indicate energy over time.

Here's a wide dispersion (black) vs narrow dispersion (red) impulse response over 40ms, I interpret the display as level relative to the level of the direct sound...

This is vertically a linear display, like voltage...

1586236121403.png



And the same data using the ETC display, using decibels relative to the direct sound, also over 40ms.

1586236389976.png
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Just grabbing a recent example. You have the blue line showing the early reflections which were calculated from the anechoic directivity measurements.
1586238661066.png


You also get the estimated in room response taking into account important room reflections. Like below. The catch being it isn't your room, but a standardized representative room. If you had the needed info you could take the spin-o-rama data and do calculations using your own room. I don't know how to do it myself right off hand, but it certainly would be doable.
1586238812078.png
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,172
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

Lets assume that anechoic measurements like what is published here are available.
Well, what you do is window first the direct sound of your room measurement so usually the first few ms and then make a new window from that time point till the usual 500ms. If you divide both you can have also the ratio of them. This doesn't really work in the bass region though as the first window is usually too short for it.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

Lets assume that anechoic measurements like what is published here are available.

It's an interesting question. If you have the frequency response of the speaker measured under anechoic conditions at a given voltage, a simple way would be to:
  1. Scale this anechoic response to the distance at which you intend to take your in-room measurement (e.g. if the anechoic measurement was taken at 1m and your in-room measurement will be at 2m, scale the anechoic response down by a factor of 2, ie 6dB).
  2. Take an in-room measurement at said location at the same voltage as the anechoic measurement was taken.
  3. Use a "calculator" tool like the one in VituixCAD to divide the measured in-room frequency response by the scaled anechoic frequency response.
The quotient this generates will be the difference in SPL vs frequency.

I haven't attempted the above with in-room vs anechoic FRs, but I've used it to calculate the SPL difference vs frequency for a driver measured at two different drive voltages.

Here's an example of how that looks (NB: disregard data below 100Hz in this partcular case):

1586262617949.png


PS: for an in-room measurement, you'd probably want to smooth the measured in-room FR heavily, or average FRs measured at a number of different locations around the listening position, to mitigate the effects of comb-filtering caused by reflections in the room.
 
Last edited:
OP
S

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
My Opinion:

Not "predicting", but via in-my-room measurement:

Stretch out an Impulse, or use ETC, is what I think of.

It doesn't tell you frequency, but seems to indicate energy over time.

Here's a wide dispersion (black) vs narrow dispersion (red) impulse response over 40ms, I interpret the display as level relative to the level of the direct sound...

This is vertically a linear display, like voltage...

View attachment 57587


And the same data using the ETC display, using decibels relative to the direct sound, also over 40ms.

View attachment 57588
So are you saying that look at the impulse response measured in room and subtract the (known) anechoic response from it. The remainder is room reflection? I may be making some leaps here so sorry if I misinterpret what you are saying.

Well, what you do is window first the direct sound of your room measurement so usually the first few ms and then make a new window from that time point till the usual 500ms. If you divide both you can have also the ratio of them. This doesn't really work in the bass region though as the first window is usually too short for it.
This only works if all boundaries are further away from the speaker than the LP, right?
 
OP
S

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
It's an interesting question. If you have the frequency response of the speaker measured under anechoic conditions at a given voltage, a simple way would be to:
  1. Scale this anechoic response to the distance at which you intend to take your in-room measurement (e.g. if the anechoic measurement was taken at 1m and your in-room measurement will be at 2m, scale the anechoic response down by a factor of 2, ie 6dB).
  2. Take an in-room measurement at said location at the same voltage as the anechoic measurement was taken.
  3. Use a "calculator" tool like the one in VituixCAD to divide the measured in-room frequency response by the scaled anechoic frequency response.
The quotient this generates will be the difference in SPL vs frequency.

Is VituixCAD smart enough to not just simply divide the dB values? I am thinking we need to subtract instead?

I haven't attempted the above with in-room vs anechoic FRs, but I've used it to calculate the SPL difference vs frequency for a driver measured at two different drive voltages.

Here's an example of how that looks (NB: disregard data below 100Hz in this partcular case):

View attachment 57641

PS: for an in-room measurement, you'd probably want to smooth the measured in-room FR heavily, or average FRs measured at a number of different locations around the listening position, to mitigate the effects of comb-filtering caused by reflections in the room.

So the difference you measured was due to some compression effects? Ideally, if only voltage is changed, the difference would be a straight line (constant bias)?
 

aarons915

Addicted to Fun and Learning
Forum Donor
Joined
Oct 20, 2019
Messages
685
Likes
1,140
Location
Chicago, IL
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

Lets assume that anechoic measurements like what is published here are available.

I've been using PEQ for many years with good results, I use a "less is more" approach and do the bare minimum needed to correct problems in my speakers. I just got a new receiver that has Audyssey XT32 and ran the room correction and not too surprisingly the target flat curve made my speakers sound too bright and harsh, as many people have mentioned. I do think a room correction software could differentiate between direct and reflected sounds and do a much better job than the current ones do. It would basically take the first sounds heard and classify them as the "direct sound" then maybe 15-30 ms later those sounds would be the early reflections. I would rather these room correction softwares just give us more control over exactly what is being corrected so that those of us who know what we want can do it and those who don't care can just take the default.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
So are you saying that look at the impulse response measured in room and subtract the (known) anechoic response from it. The remainder is room reflection? I may be making some leaps here so sorry if I misinterpret what you are saying.

I may have misinterpreted your original question, given the other answers received.

---

"Can an in-room measurement predict what sound is direct vs reflected?"

I don't have anechoic measures with which to compare or predict.

If I measure the in-room, I'll just find ways in REW to "look" at what is direct vs reflected.

Using the output of the preamp as an "anechoic ideal" with which to compare is something I'm looking at.

---

You can easily separate the direct vs reflected when you shout into a large reflective enclosusre "HELLO....HELLO....HELLO...."

REW can assist looking at much smaller enclosures.

---

I'm more of a practical or experimental than theoretical amateur audio scientist (if I may be so bold as to call myself one)

---

Example:

Preamp output vs in-room. I didn't expect the curve for the preamp, but I suppose that shows some limits for what could be seen with a speaker measured anechoically - like comparing the drive signal to the in-room result.

1586284011939.png
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
@spacevector so are you wishing to measure your in room response and separate out the part from reflections or are you wishing to take the data from Amir's reviews and calculate what your room will do to the result? My apologies for not understanding which it is you are wanting to do.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Are there any in-room measurements and post calculations that can predict the ratio of direct vs reflected sound in said-room across the frequency band?

Lets assume that anechoic measurements like what is published here are available.

If I understood yor question correctly you can take a guess looking at PIR spinorama response, as PIR is a mix of direct and reflected sound.

If anechoic measurements are not available than your chances are pretty much none as gating the in-room measurement won't give accurate result for the entire frequency range.
 
OP
S

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
@spacevector so are you wishing to measure your in room response and separate out the part from reflections or are you wishing to take the data from Amir's reviews and calculate what your room will do to the result? My apologies for not understanding which it is you are wanting to do.
Yes, this is correct. My intention is to measure the in-room response of the speaker as placed in the room, at the LP (or MP, measurement position). Then use post-processing to determine whats coming from the speaker (anechoic response) and what is the room.

Don't ask why I want to do this - just a thought that occurred to me last night (I had had a couple nips). Seems an interesting problem though and may inform some things like room-placement and room-treatment optimization.
 
OP
S

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
I may have misinterpreted your original question, given the other answers received.

---

"Can an in-room measurement predict what sound is direct vs reflected?"

I don't have anechoic measures with which to compare or predict.

If I measure the in-room, I'll just find ways in REW to "look" at what is direct vs reflected.

Using the output of the preamp as an "anechoic ideal" with which to compare is something I'm looking at.

---

You can easily separate the direct vs reflected when you shout into a large reflective enclosusre "HELLO....HELLO....HELLO...."

REW can assist looking at much smaller enclosures.

---

I'm more of a practical or experimental than theoretical amateur audio scientist (if I may be so bold as to call myself one)

---

Example:

Preamp output vs in-room. I didn't expect the curve for the preamp, but I suppose that shows some limits for what could be seen with a speaker measured anechoically - like comparing the drive signal to the in-room result.

View attachment 57673
In that example you posted, what is the signal that leads to the blue (pre-amp) ETC like that? Is it an impulse? Also, I am not fully familiar with what ETC signifies - cribbing now.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
In that example you posted, what is the signal that leads to the blue (pre-amp) ETC like that? Is it an impulse? Also, I am not fully familiar with what ETC signifies - cribbing now.

https://www.roomeqwizard.com/help/help_en-GB/html/impulseresponse.html

https://www.roomeqwizard.com/help/help_en-GB/html/graph_impulse.html

I will guess it has something to do with Fourier Transform results near a frequency of 0Hz.

@JohnPM will know, it's his software.

---

Expanding the ETC trace shown above to the left:

red - speakers
black - preamp out via 25ft cheap cable and PC on-board ADC


1586298568956.png
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
Is VituixCAD smart enough to not just simply divide the dB values? I am thinking we need to subtract instead?

The decibel is a ratio, hence it must be divided ;)

So the difference you measured was due to some compression effects? Ideally, if only voltage is changed, the difference would be a straight line (constant bias)?

Compression and other causes of nonlinearity, yes.
 
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
In that example you posted, what is the signal that leads to the blue (pre-amp) ETC like that? Is it an impulse? Also, I am not fully familiar with what ETC signifies - cribbing now.

I'm sorry, misunderstood part of your question yet again...

Both speakers and preamp are excited using the REW Swept Sine measurement signal.

REW "listens" to the results of the sweep and calculates all the displays it has available.

The preamp measures are with no EQ - flat 2Hz to 24kHz.

The speaker measures include EQ, and the sweep was from 10Hz to 24kHz, but the idea of using EQ is to correct their output and room contribution to approximate a flat signal.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Ok, so here is the answer though not a convenient one. Take your speaker outdoors, and measure where there are no nearby reflections. Even then, not perfect and there will be noise, but close to anechoic. Take it back inside and remeasure. Subtract one from the other and the difference is the room reflections.

The Klippel unit essentially does what you are asking, and you see how much it costs. So its possible, it isn't easy or cheap at this time.

Next best is to measure your speaker rather close say at 1.5 meters maybe you can even get closer. Then measure at the listening position. Room reflections are the difference even though there are reduced reflections included in the closer measure.
 
OP
S

spacevector

Addicted to Fun and Learning
Forum Donor
Joined
Dec 3, 2019
Messages
553
Likes
1,003
Location
Bayrea
So if anechoic data is already available - like from speaker reviews here, then all I need to do is take in-room measurement, scale it so that its at similar distance and voltage level like @andreasmaaan outlined and subtract one from other. The difference is room.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
The difference is room.

Wouldn't the difference would also include the speaker and microphone locations in the room?

I guess I'm not sure what you really seek.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
Wouldn't the difference would also include the speaker and microphone locations in the room?

I guess I'm not sure what you really seek.

I assumed that's what @spacevector was interested in. After all, how could an in-room measurement be taken that wasn't mic/speaker location-dependent?
 
Top Bottom