Room Measurement Performance metrics

[dmorse6736]

I have been using REW to measure my home cinema and various room of friends and audio club members.

With audio equipment, amps are specified, for example, 20-20KHz +/-0.2 db. A speaker might be 35-20khz +/-3dB. But what metrics are appropriate for a audio system/room (ASR) performance? I have noticed most ASR roll off at about 10KHz. My GIK audio consultant has told me not to worry about that (home audio rooms do that, and I've seen the same in my measurements).

One room I measured that our audio club generally considers a good system (straight wire with gain, no processors, no EQ) had a response range of approximately 7-8 dB (using the REW psychoacoustic filter) measured at the prime listening position. If I used 1) a "flat' objective, 2) used 1kHz as the reference, 3) response range was limited to 10kHz, this system would be described as 40-10kHz -1, +6dB. If you used a different reference you might say the system was 40-10kHz +/- 4dB. When I worked at Cadillac as the Lead Entertainment Systems Engineer we found that typically when we sent a car to "The Mountain" they returned the car with a relatively smooth spectral response that had a say -3dB slope from 40- 10kHz. (I should add we felt the Bose systems had less listening fatigue than our other systems with a more jagged response and no slope or even a rise in high frequencies)

So with this said;
1A) Does REW have a math function that will calculate the +/- dB SPL given a specified frequency range of interest (ie 40-10kHz)?
1B) Given 1A does REW provide the reference frequency(s)?
2A) Is there a generally accepted metric or target for Audio System Room performance?
2B) What are others using REW measuring?
3) Is a slope in ASR common?

A final comment. If anyone can direct me to standards on this topic I would appreciate it. Also, I am referencing to SPL response only. I am not discounting the Spectragram, RT60, and Impulse as important to room performance.

Thanks, Dale

 

[JStewart]

There are no one paragraph answers for 2A, 2B, and 3. Well maybe to 3 for which the answer is yes. Now as to the why …
https://www.amazon.com/Sound-Reprod...074CHY128/ref=mt_other?_encoding=UTF8&me=&qid=

This is one answer I’ve run across for 2A. I’d guess it’s generally accepted, but won’t come with the scientific scrutiny of Dr. Toole. I’m sure there are others.

http://www.acousticfrontiers.com/wp-content/uploads/2011/10/acoustic_measurement_standards.pdf

Copied from page 23:
Conclusion
The purpose of this paper is to offer clear standards in areas that affect the audiophile two channel listening experience. It is important to realize that no single measurement can tell you what a room will sound like. When considered in combination, however, these standards provide powerful insights into your room’s performance. It is our belief that as a result of this article you will be much better placed to determine and prioritize the acoustical areas that need improvement. Further, it is clearly shown that the process of locating two speakers and a single listening position in a room is complex and no speaker in and of itself should be expected to compensate for all possible variables in a given space.
It is not realistic, nor is it crucial, that your room meets all of these targets. But identifying which areas of reproduction are important to you and optimizing your room to meet these goals is an important step towards improving your listening pleasure.

Good luck in your journey!

 

[thothsong]

For 1A, there's graph metrics.

 

[ddude003]

For 1B try the Signal Generator... https://www.roomeqwizard.com/help/help_en-GB/html/siggen.html#top...
As for 2A and 3... There are some standards and some best practices and studies... Mitch at Accuratesound has listed a few references here https://accuratesound.ca/standards.html... The IEEE, AES and SMPTE, to name a few, have libraries that may interest you...

Everyone's room, system, ears and brain are different... And everyone has different preferences in the music they listen to and the way they think their system should sound... Vive la différence...

 

[JStewart]

For 1B try the Signal Generator... https://www.roomeqwizard.com/help/help_en-GB/html/siggen.html#top...
As for 2A and 3... There are some standards and some best practices and studies... Mitch at Accuratesound has listed a few references here https://accuratesound.ca/standards.html... The IEEE, AES and SMPTE, to name a few, have libraries that may interest you...

Everyone's room, system, ears and brain are different... And everyone has different preferences in the music they listen to and the way they think their system should sound... Vive la différence...

Good one. Forgot about Mitch.

 

[kevinzoe]

Dale (the OP):
2A) Is there a generally accepted metric or target for Audio System Room performance?
I second the Acoustic Frontiers white paper and its frequency and time domain "standards." They may not be blessed by the audio Gods, but by God they're good things to shoot for . . .

2B) What are others using REW measuring?
I use the Dayton Audio OmniMic acoustical measurement tool instead of REW. When I measure frequency response, I consider several metrics to gauge if I'm improving my 2channel stereo+room. I put these metrics in a Scorecard - similar to a credit scorecard a bank uses to asses your credit score and lending risk.
> Degree of fit between my actual measured freq response and my target house curve that I'm using as my standard/goal/baseline for comparison purposes. In Microsoft Excel I run an Entropy and Engineering method to calculate the degree of fit.
> Large freq swings of 5dB or greater are captured as I would hear a 5dB change in contiguous notes played by an instrument. I track the number of 5dB+ swings and their largest size.
> Nulls are important too, which I interpret as meaning when a freq response dips below -3dB from the target curve. I want to know how many there are, how deep is deepest one, how deep are they on average, how many instrument's actual notes are within the null.
> How many parametric EQ filters are needed below 500Hz to help smoothen the response by attenuating peaks. The fewer the better.

My scorecard with metrics, values, and associated points is something I came up with and is in no way any kind of standard . . . just creative thinking on my part I'm afraid. I'm in the process of doing one for the time domain. . . .

Freq Response vs Target Curve Scorecard
Max Scorecard Points =	900​
Metric	Range of Values	Scores	Selected Value	My Score
Goodness of Fit - using math (Entropy method) I measure how well the curve fits/matches the target curve	0.000 - 0.010	100​	0​	0​
	0.011 - 0.020	75​	0​	0​
	0.021 - 0.030	50​	1​	50​
	0.031 - 0.040	25​	0​	0​
Goodness of Fit - using math (Engineering method) I measure how well the curve fits/matches the target curve	0.005 - 0.015	100​	0​	0​
	0.016 - 0.025	75​	0​	0​
	0.026 - 0.035	50​	0​	0​
	0.036 - 0.045	25​	1​	25​
Max Freq. Swing Size in dBs - using 1/12th smoothing, when a frequency swings up or down of 5dB or more, it's considered a big swing and this metric focuses on the largest swing size	5.0 - 7	100​	0​	0​
	8.0 - 10	66​	0​	0​
	11 or more	33​	1​	33​
# of Big Swings - using 1/12th smoothing, this metric focuses on the number of swings that are 5-6dB or larger.	0 to 2	100​	0​	0​
	3 - 4	75​	1​	75​
	5 - 6	50​	0​	0​
	7 or more	25​	0​	0​
Null Width ( # of notes affected) - using 1/24th smoothing, I consider a null as a freq that drops below the -3dB target curve. The number of actual instrument notes affected is shown in this metric.	7 or fewer	100​	0​	0​
	8 - 10	75​	1​	75​
	11 - 13	50​	0​	0​
	14 or more	25​	0​	0​
Null Maximum Depth - using 1/24th smoothing, this metric shows the deepest null amount in dB's.	3 - 5	100​	0​	0​
	5.1 - 7	75​	1​	75​
	7.1 - 9	50​	0​	0​
	9.1 or more	25​	0​	0​
Total Number of Nulls - using 1/24th smoothing, this metric is a count of the number of nulls below 10kHz	2 or less	100​	0​	0​
	3 - 5	75​	0​	0​
	6 - 8	50​	1​	50​
	9 or more	25​	0​	0​
AVG Null Depth - using 1/24th smoothing, the sum of each null's depth is divided by the number of nulls to arrive at the average null depth	1 or less	100​	0​	0​
	1.1 - 3	75​	1​	75​
	3.1 - 5	50​	0​	0​
	5.1 or more	25​	0​	0​
# of PEQ filters needed per channel	1 or less	100​	0​	0​
	2 - 3	66​	1​	66​
	4 or more	33​	0​	0​
			Score	524​
			Score %	58%​

3) Is a slope in ASR common?
By slope I'm assuming you mean a target house curve and if so then YES it is very much a necessity. Audiolense requires it, or at least I use it every time I use Audilense to create a FIR correction filter for me to play in JRiver . . . Mitch's website has a list of various target curve standards.

hope this helps . . .

 

[teyhyrh4r]

Mitch at Accuratesound has listed a few references here https://accuratesound.ca/standards.html.

I always wonder: for which smoothing those are intended?

 

[ddude003]

I always wonder: for which smoothing those are intended?

I'm not sure what you are asking here... Target house curves are just that, a target which you are trying to get the most out of your speakers and how they interface with the room... The big takeaway is that every kit, speakers, room, ears and brain are different and as such there are many ways to "shape" your sound to your preference as well as smooth out the room's response... If you look around you will find many and various ways to do EQ in the analogue and digital domain in both hardware and software...

 

[teyhyrh4r]

I'm not sure what you are asking here... Target house curves are just that, a target which you are trying to get the most out of your speakers and how they interface with the room... The big takeaway is that every kit, speakers, room, ears and brain are different and as such there are many ways to "shape" your sound to your preference as well as smooth out the room's response... If you look around you will find many and various ways to do EQ in the analogue and digital domain in both hardware and software...

the thing is, the more you smooth, the more you ignore the peaks in the room response. If I smooth to Rew's var I will have peaks 6dB higher on the unsmoothed graph.

 

[ddude003]

the thing is, the more you smooth, the more you ignore the peaks in the room response. If I smooth to Rew's var I will have peaks 6dB higher on the unsmoothed graph.

I think I know what you are talking about now, just not sure without seeing a .mdat file to be sure... A Target room curve is usually described as some combination of +db on one side and -db on the other end... However, that does not necessarily mean you want to boost on one side and cut on the other... You may need to set that Target curve a few db below what might be the mid point of your measured frequency response so you are not boosting the low frequencies... You don't change the Target curve itself you move the Target curve level... In REW's EQ window you have control of how much boost you allow with either the generated Target or with your selected House Curve and Filter Tasks...

You might take a look at this thread starting at https://www.avnirvana.com/threads/help-eqing-the-subwoofer.11059/post-83426 which shows several pics of the placement of Target curves... This technique involves no boosting, only cutting...

 

[DanDan]

I use the old B&K Curve. Speaker Controller app Sonarworks included B&K and it's own recently researched actual home responses. They are all remarkably similar, as is the Harman curve. Simply put 6dB slope from 100Hz to 10K..... Harman say room gain is 1dB per octave.