advertisement


Speaker/Room Measurement Witchcraftery

I have REW and a UMIK-1. I occasionally venture into measuring my room/loudspeakers to learn something. I have read the thread with interest but I wonder if there are simpler but still useful things that one can do.

I have personally found it useful to set up REW to generate noise [1] and then use the RTA (real-time analysis) tool while moving loudspeakers around in real time to observe how the frequency response changes at the listening position. I find I can moderate the peaks and troughs in the low frequency region to a useful extent, and this is noticeable in things like bass-heavy movie scores.

I am never going to get anywhere close to a perfect response this way. However the human ear-brain system adapts and I have found that I can make a useful improvement without aiming for technical perfection.

[1] White PN noise for the RTA tool in "Spectrum" mode (which is what I prefer) or Pink PN noise for the RTA tool in one of the "RTA" modes (EDIT: Pink PN noise is safer for the tweeter).
 
Last edited:
Wouldn’t it be more convenient to ‘virtually’ move your speakers/listening position in the room simulation feature?
Keith
 
I have REW and a UMIK-1. I occasionally venture into measuring my room/loudspeakers to learn something. I have read the thread with interest but I wonder if there are simpler but still useful things that one can do.

I have personally found it useful to set up REW to generate noise [1] and then use the RTA (real-time analysis) tool while moving loudspeakers around in real time to observe how the frequency response changes at the listening position. I find I can moderate the peaks and troughs in the low frequency region to a useful extent, and this is noticeable in things like bass-heavy movie scores.

I am never going to get anywhere close to a perfect response this way. However the human ear-brain system adapts and I have found that I can make a useful improvement without aiming for technical perfection.

[1] White PN noise for the RTA tool in "Spectrum" mode (which is what I prefer) or Pink PN noise for the RTA tool in one of the "RTA" modes.
Indeed, and whilst the room sim is a useful tool it won’t be as accurate as the method you use. The room sim cannot take into account openings, odd shaped rooms, furnishings etc etc. In the last analysis, whilst measuring is a fantastically useful tool, it is only that when used with listening with one’s dear old ears and to one’s own taste!
 
White noise puts more energy in the tweeters, keep the levels down if you use that.
Indeed. That's a very welcome reminder.

In the generator I applied the custom high cut filter at 1 kHz (a couple of octaves below the mid/tweeter crossover frequency) and set a maximum level of 80 dB SPL. Positioning a loudspeaker this way seems primarily to impact frequency response below the room's Schroeder frequency (typically 250 Hz, AIUI) and also a little in the two octaves above. So protecting the tweeter (and my ears) is always a good idea.
 
Indeed, and whilst the room sim is a useful tool it won’t be as accurate as the method you use. The room sim cannot take into account openings, odd shaped rooms, furnishings etc etc. In the last analysis, whilst measuring is a fantastically useful tool, it is only that when used with listening with one’s dear old ears and to one’s own taste!
Yes. The final proof is in what you hear. In that respect I have been interested in discovering how much room equalization is "good enough" for my ears. I am quite sure "ruler flat" is not necessary to enjoy good sound and, more important, good music. My own experience is that simple improvements with the help of excellent tools like REW work very well without needing to buy more "stuff" to equalize the room to within an inch of its life.

And my room is indeed quite asymmetrical acoustically. A fairly plain wall (with a sofa) to the left. Two doors, a fireplace and bookshelves to the right. Playing with REW's room simulator does show some features recognizable in measurements but didn't help me much (I may still have a lot to learn, though). But putting in the final loudspeaker positions from testing with noise and RTA did show a surprising degree of frequency response smoothing, although not quite in the way I was expecting.
 
You could measure, reposition re-measure and overlay the plots, you don’t really want a’flat’ in room response nearly every domestic user prefers the 10dB downward slope.
Keith
 
Wouldn’t it be more convenient to ‘virtually’ move your speakers/listening position in the room simulation feature?
The simulation of my room is quite busy with features, some of which correspond to measured and/or audible features and some that don't (or maybe I just don't understand yet). Plugging numbers into the simulator after positioning the loudspeakers using the noise + RTA method does indicate tentatively what matters in the simulation and what to ignore. I would not understand that from just believing my interpretation the simulation without the practical evidence.
 
I realise that measuring will answer my question, but before I put my back out doing so, for a loudspeaker system that uses a passive bass radiator to extend low frequency response, such as the Celestion Ditton 66, which driver should be used as the position in the REW Simulator, the active woofer or the passive radiator? I'm presuming its the active woofer, since it's the driver that's contributing most to the output?

Also, if I wanted to (very roughly) measure the contribution of the passive radiator to the system's response, can I simply measure the system's response whilst blocking the movement of the passive radiator with my hand, and then compare this against the normal response?
 
The simulation of my room is quite busy with features, some of which correspond to measured and/or audible features and some that don't (or maybe I just don't understand yet). Plugging numbers into the simulator after positioning the loudspeakers using the noise + RTA method does indicate tentatively what matters in the simulation and what to ignore. I would not understand that from just believing my interpretation the simulation without the practical evidence.

The REW simulator is too generic. It doesn't take into account doors, windows, furnishing, has no option for irregularly-shapped rooms and doesn't really recreate the response of your particular speaker model.
I use the same technique as you with RTA/Pink Noise and find it very effective.
 
I've discovered something interesting in the REW Simulator...

When modelling with a single speaker, moving the listening position left or right of central increases the amplitude of the fundamental axial width mode, as you would expect (see Sim#1).

Sim#1:
49584829657_0ab4c90cb3_o.jpg


However, modelling with stereo speakers spaced equidistant from the side walls, there is zero change to amplitude of the fundamental axial width mode when the listening position is moved left or right, but its multiples are affected, as are the tangential modes (see Sim#2). I find this bizarre because it contradicts my own experience (when I move my listening position nearer to a side wall the amplitude of the fundamental axial mode does increase).

Sim#2:
49584598011_cdfe53d511_o.jpg


To illustrate the same contradiction, but this time changing the absorption coefficients, reducing or increasing the absorption of the side walls with a single speaker increases or decreases the amplitude of the fundamental axial width mode (see Sim#3 and Sim#4). When there are stereo speakers, the fundamental is unchanged, only the multiples and tangential modes are affected (see Sim#5 and Sim#6).

Sim#3:
49584099653_5c5f6c47c9_o.jpg


Sim#4:
49584829597_123392bfc9_o.jpg


Sim#5:
49584099608_6998044715_o.jpg


Sim#6:
49584829547_acd5e0f3b3_o.jpg


However, if any one variable is changed so that the stereo speakers are no longer operating in perfect unison, e.g. if the LF rolloff patterns of the two speakers are slightly different, or if the gains are slightly different, or if there is a slight delay between the two speakers, then all of the above stereo speaker simulations no longer hold true, and we do finally see a change in the amplitude of the fundamental axial width mode when the listening position is moved left or right and the amount of absorption is increased or decreased (see Sim#7 and Sim#8).

Sim#7:
49584157843_f5814b714c_o.jpg


Sim#8:
49584157793_1b3499a7a8_o.jpg


Is it fair to assume that no room or loudspeaker setup will be so perfectly aligned as to provide the circumstances shown in Sim#2, where moving left or right of the central listening position produces zero increase in the amplitude of the fundamental axial width mode, and that Sim#7/Sim#8 is a more realistic outcome?
 
If you have perfect left/right symmetry of sources along the length the odd order width modes cannot be excited as they are driven in equal and opposite phases by the sources either side of the mid line. A pair of speakers doesn't quite achieve that, unless they are halfway down the room, but still helps. A sub in each corner of a room does, or subs on the side wall mid points. All need to be driven at the same amplitude and with the same phase, however. A common suggestion for speaker placement is symmetric left and right and at the 1/4 width points, to also reduce excitation of the first even width mode.
 
I believe I've made this point before but it's probably worth reiterating, for those using sine wave sweeps or pink noise RTA to measure the in-room response of their speakers at their listening seat with RoomEQWizard.

IME measuring both speakers at the same time is far more likely to produce spurious results in the middle and high frequencies than measuring each speaker individually.

I took listening seat measurements on two consecutive days (yesterday and the day before), keeping the mic in the same position as much as possible (I have the mic on a boom-stand so can effectively 'swing it' into my listening position without moving the stand so am able to maintain location accuracy to within a few mm's).

On the first day, the stereo measurement showed a downward sloping HF response, but on the second day it showed a flat HF response. The individual measurements of left and right speakers showed a flat HF response on both days, which strongly suggests the first stereo measurement was spurious.

I have in the past experienced other anomalies with stereo measurements, such as broad peaks and dips in the HF response that are not audible to the ear and do not exist when the left and right channels are measured individually. Since moving the position of the mic a couple of mm's often makes the anomalies disappear, I can only conclude they are a spurious artefact caused by using a single mic to measure sound patterns from two sources.

I therefore strongly advise against stereo measurements. IME they're fine for measuring bass response but are too flaky for high frequencies. Better to take individual measurements of the left and right speakers and then use the Average function in REW to approximate a stereo measurement.

Or even better, if you can be bothered, take several individual measurements of the left and right speakers covering a small area around your listening position and then use the Average function in REW to produce an Averaged Left and Averaged Right measurement. You can then average the Left and Right Averages if you want to see the overall "house curve" your speakers are presenting in your room.

AFAIK room correction software such as Dirac does not include a stereo measurement as part of the measurement process, so one can only assume they too have misgivings about the reliability of stereo measurements.

PS - For those posting frequency response graphs in the future, if you do include a combined measurement it would be helpful if you could clearly indicate whether the measurement is Stereo or an Average of the Left and Right channels, so that we're able to take the former with an even larger pinch of salt! ;)
 
IME measuring both speakers at the same time is far more likely to produce spurious results in the middle and high frequencies than measuring each speaker individually.

I would go a step further and suggest than for in-room response measuremnts each speaker must always be measured separately.
 
Thank goodness all the plots of my horrific in-room responses have expired. My system sounds so much better now :D:D
 
I don’t really understand this. Surely most of us are using measurements in order to double-check what we are hearing from the listening seat in our rooms? As such that will always be the sound of two speakers playing in that space. To my ears the sound of one speaker alone is way less than half the sound of a well setup stereo system, it just doesn’t drive the room in anything like the same way. Surely we want to be measuring what we are actually hearing, not some theoretical abstraction. FWIW I’ve similarly never understood the point of measuring a speaker at 1m unless you are trying to track a very specific fault in pair-matching or whatever. Without an anechoic chamber any such measurement won’t tell you much about the speaker design, so really all that interests me is the sound of my full system measured at the listening seat in the way I listen to it. I’d never cite any of my measurements as being representative of a Tannoy Monitor Gold, a JR149 or whatever as they are just an REW measurement from the listening seat in my room and as such valueless to anyone other than me (assuming they actually have any value to me!).

I guess that there could well be phase/time errors that a single mic placement would suffer from that a pair of ears wouldn’t that could give spurious results, but in that case surely a crossed pair and simultaneous stereo measurements should be done. The whole point is to measure what we are experiencing, and we are not hearing two entirely isolated mono signals. That is not happening as both speakers are driving the room and any reflections, time and phase delays etc are happening in real time between both speakers and that listening space. That is what needs to be measured.
 


advertisement


Back
Top