If the cancelling is working well at the optimum point, then chances are good that the comb filter patterns will be symmetrical left and right from that (read: identical), where our ears are located during actual listening.
Yes, exactly.
Sorry Simon, it didn't occur to me that both could be mixed up. I just wanted to show that my mic position was sufficiently "time-aligned" between both speakers to give meaningful results.
"Dipolisti" like Klaus (?) and me always expect our systems to react opposed to the rest of the (monopole) world.
Thanks for leading us to a common and simple explanation.Rudolf
Just so that everybody is clear on what this measurement is about, you should dig up some references on the Hopkins-Stryker equation.
For example: http://www.google.ca/url?sa=t&sourc...GZjNVM&usg=AFQjCNGg0TP3QRufTIuKFm-NZkE8nz2GjQ
and Architectural acoustics - Marshall Long - Google Books
from page 306 or so.
This equation is commonly used by PA system designers to calculate what level of direct sound they will have at a point in the audience based on room absorption and loudspeaker Q (directivity). PA designers worry primarily about speech inteligibility: can you understand the words? As a rule of thumb it has been found that inteligibility is fair up to a -12 dB direct to reverberent ratio. (Yes, the reverberation can be 12dB louder than the direct sound and you can still have speech inteligibility, such is the brains ability to focus on the early sound.) This would occur within a hall at 4 times the critical distance.
We would like to get more from our sound systems than simply being able to understand the words! High clarity and an ability to hear nuances of the original signal requires a much greater direct to reflected ratio.
A second comment on our measurements is that we are using a trick to reveal the system's reverberent level for a given (primarily) direct level. The trick only works if our cancelation is sufficient. In the end, we are really measuring the reverberent field level, or the level of left right imbalance, whichever is greater.
Regards,
David S.
For example: http://www.google.ca/url?sa=t&sourc...GZjNVM&usg=AFQjCNGg0TP3QRufTIuKFm-NZkE8nz2GjQ
and Architectural acoustics - Marshall Long - Google Books
from page 306 or so.
This equation is commonly used by PA system designers to calculate what level of direct sound they will have at a point in the audience based on room absorption and loudspeaker Q (directivity). PA designers worry primarily about speech inteligibility: can you understand the words? As a rule of thumb it has been found that inteligibility is fair up to a -12 dB direct to reverberent ratio. (Yes, the reverberation can be 12dB louder than the direct sound and you can still have speech inteligibility, such is the brains ability to focus on the early sound.) This would occur within a hall at 4 times the critical distance.
We would like to get more from our sound systems than simply being able to understand the words! High clarity and an ability to hear nuances of the original signal requires a much greater direct to reflected ratio.
A second comment on our measurements is that we are using a trick to reveal the system's reverberent level for a given (primarily) direct level. The trick only works if our cancelation is sufficient. In the end, we are really measuring the reverberent field level, or the level of left right imbalance, whichever is greater.
Regards,
David S.
You can get an idea by directly comparing Left and Right to see how well they track. (Like I did with my first graph showing Left and Right overlaid on each other)
If they track well within a dB or two you know what's left is probably the reverberant field. If Left and Right diverge more than this then accuracy will be lost at frequencies where they diverge, as there will be incomplete cancellation.
So to do a really thorough job you want to measure noise floor to make sure the noise floor is far enough below the measurement, left and right separately to check how well they track, (adjusting channel balance for best match if needed) and then Left+Right and Left-Right to compare with each other to determine the ratio.
Last edited:
Problem is that you would need to match the direct component only. How to do this within a small enclosed space like a living room or home theater?
Gating?
Thanks Simon, that makes good sense. Mine track well and noise is low, so I can be pretty confident that the direct to reflected ratio is about 12-15dB. Otherwise said, the direct sound is more than twice as loud the ambient. Right?
I hear your point, but I wonder if that's over thinking the problem ?
If we were actually measuring the direct to reverberant ratio you would be right, but we're measuring the direct+reverberant to reverberant ratio.
Whatever we do to adjust the left-right channel balance to maximize cancellation between the two speakers in out of phase mode should be a good thing - regardless of whether the imbalance is due to the speakers or asymmetrical room effects like one wall being more heavily damped.
The object of the exercise is to reveal everything left over which is not phase correlated with the speakers, with as much of the phase correlated stuff cancelled out.
The whole basis of the test is only an approximation anyway - we aren't really comparing the true direct response vs the true reverberant field, we're really comparing the ratio of phase correlated+ phase uncorrelated material (in phase connection) versus phase uncorrelated material. (out of phase connection)
If you were were worried about it you could try setting the channel balance by comparing high frequencies with a gated measurement, but I'm not sure whether we'd see much difference in the results. Perhaps try both ways and compare the results ?
Yes, but how close is the back of your couch for example? This determines the longest gating time. The shorter the gate the higher the frequency from which on data would be valid.
Last edited:
Do we need to match the direct response from measurements at the final speaker position? Can't we measure both speakers in the middle of the room and compare from that?
Rudolf
Yes, few things are more sobering than the slow sine sweep. You can hear just how bad the room reflections and resonances are. I'm left thinking that it's amazing that it sounds as good as it does with music. Very good, but lots of coloration in both frequency response and spacial cues, constantly varying with frequency...
Don't place too much importance on the result of a slow sine sweep at high frequencies in a room though.
Because it has an extremely narrow bandwidth (depending on sweep speed) it reveals the very dense modes and interference patterns caused by the room, but which are far narrower than the critical bandwidth of our ears.
Play a 10Khz tone and move your head a few inches and our perception of the amplitude will vary wildly by +/- 10dB or more - when only one frequency is present critical bandwidth smoothing can't take effect.
Replace that with a broadband spectrum like pink noise (or many kinds of music) and you'll notice no dramatic variations in perceived high frequencies with a few inches of head movement (at least with a single speaker instead of stereo) because as some spot frequencies are increasing in amplitude others are reducing - within 1/3rd octave bands the overall power level stays roughly the same, and we perceive this.
For transient sounds you also have the ability of the brain to isolate and separate first arrivals working in your favour as well - there might be comb filtering present, but if the brain can isolate the initial arrival separately it can perceive it without the effect of comb filtering.
Narrow band steady-state measurements of rooms without 1/n octave averaging look pretty horrific at high frequencies but fortunately for us its not representative of how we will perceive music in such a room.
Which suggests the slow sine sweep is not highly correlated with how good a room sounds when playing music, and hence not the most useful of tests for that goal?
Simon, your comments about the bass frequencies in my measurement are dead on. The left speaker is near a wall, whereas the right one is next to the kitchen, i.e, no wall. I've been fiddling with different EQ settings but haven't quite settled on something I like yet. There is also a big bump in the midbass region, which has confounded me so far. But, in my defence, these are new cabinets, a new design and I haven't been able to devote enough time to fine tune them yet.
I think I understand what you're saying about the high frequencies, where the wavelengths are very small and our perception above 6kHZ changes. I was more surprised by what I heard in my room in the midrange area, 200HZ - 2kHZ. At times it sounded like the room was putting out more sound than the speakers. With dipoles and parallel walls (3 sets of parallel walls even though the floor is carpeted) I guess you'd expect some of that.
On the contrary, any test method, the results of which I am fairly adept at interpreting, that shows errors in a bigger fashion, seems good to me. What I really want to do is test my total system with tone bursts in a gaussian or "Blackman" envelop, so I can separate areas of frequency that are resonant from ones that are about reflection cancellations. Those waterfall graphs sound like a great way to do this, but I keep hearing that they aren't accurate in the lower frequencies, and I'm not clear on what to do with that. Anyway, there's not a lot you can do electronically with cancellations, but resonance, because of its elongation effect, might make certain frequencies sound psycho-acoustically louder than a calibrated mic and pink noise might show. Not everything in music is impulse. Perhaps a compromise between calibration for impulse vs. steady state would be the best bet.
It's unlikely to be your speakers, do a near-field measurement of their bass response, and I'll bet there is no big peak in the mid-bass region, unless something went horribly wrong in the design.
It'll be your asymmetric room, which can play havoc with bass frequencies. I recognised the room was asymmetric from the measurements because I've been in similar situations before.
I had a room years ago that was 8 metres long by 4 metres wide (living room and kitchen connected) with speakers on the short wall at one end, and listening position about 3 metres from the speakers. This would have been fine except due to the kitchen layout at the back of the room the right hand 2/3 of the room was the full 8 metres long while the left hand 1/3 was only 5.5 metres long.
To compound the problem there was a 1 metre wide opening in the left wall to a long corridor positioned just in front of the 5.5 metre protrusion from the kitchen, and not far from the listening position. (Yet more trouble was the floor was carpet over concrete, so bass reflectivity was very high from the floor, and two of the four walls were exterior walls with brick on the exterior and not much low frequency absorption between the brick and the interior - probably just jib board and glass wool)
Because the room was a vastly different length on one side to the other, the standing wave pattern was very unusual and asymmetric between left and right speakers, despite the speakers and listener being symmetrically located within the speaker end of the room. (Same side wall, front wall distance etc)
Bass response from 20 to 50Hz was excellent and very smooth in that room but 60-200 Hz was a real mess, with a deep notch in the mid bass at the listening position which was at different frequencies for each channel due to room asymmetry - 65Hz for one and 75Hz for the other. This resulted in weird out of phase characteristics for certain critical bass notes in this frequency range.
The response from 100-200Hz was pretty ugly as well, despite near floor mounted woofers, with a deep notch at 120Hz and a big peak at 150Hz which I had to tame with EQ.
Despite living there for several years I was never able to solve the issues in the 60-200Hz region with only two speakers no matter where they were positioned. In hindsight with what I know now the only solution for that room was a distributed multi-sub approach, but I wasn't fully aware of the approach at the time or its benefits of modal smoothing, in fact I was fairly anti-subwoofer in those days.
(I still don't like subwoofers as a crutch for main speakers which are too small to play low enough and with sufficient SPL in the bass, eg sub/satellite approach, but I'm now sold on using them in conjunction with full size main speakers, distributed around the room for optimal modal smoothing)
Don't be too hard on your speakers or set-up, a hugely asymmetric room like that is going to be a nightmare to get any sort of smooth bass from two speakers alone, your only realistic option is probably a multi-sub approach with 3 or 4 subs positioned optimally around the room...done properly you should be able to get smooth bass response from just about any oddball shaped room.
Last edited:
Which kind of waterfall graphs ? Most are cumulative spectral decay, which are not accurate or particularly useful at low frequencies, no.
But ARTA (and others) also implement "burst decay" measurement which is providing essentially the same information as gaussian tone bursts, (but actually derived from the impulse response) displayed in a waterfall output format, and these are useful at bass frequencies. (Although I still find them a bit difficult to interpret since rooms have so many resonances)
Have you looked at this ? There is a good theoretical discussion about the hows and whys of the burst decay mode in the built in help menu in ARTA.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.