Hi Dave,
See graph for the best I can get with a sub-sample shift of ~0.85 Samples@44.1K (~7mm, 20uS) between the individual L and R channel recordings. So it was an indirect measurement using post-processing, not realtime diff, probably making the graph overly optimistic, notably at LF, using the data I took when I dialed in the system (with the Acourate software toolbox which does not do sub-sample pulse alignment by default)
I might try realtime test with out-of phase noise + RTA as well as out-of-phase logsweep, late tonight when it's as "quiet" as it gets, downtown where I live...
- Klaus
Attachments
You're off center by less than an inch, from what your graph looks like, that's all.
See below what a pulse offset as low as 6cm (8samples@44.1K) does to the sum and difference, using the optimized data of my last post.
Attachments
I'm not sure if I know that one. Is this the approach of applying a high pass filter function to the noise / MLSS spectrum before being sent to a speaker so that the impulse response settles within a typical short window period, and then applying a reverse transformation to the derived frequency response ?
If so, I was under the impression that the main purpose of it is to avoid truncation errors when windowing which otherwise cause spurious ripple in the response below the time-bandwidth limits, but that it still can't create low frequency resolution that doesn't exist in a short time window.
So for example it would show a reasonably correct bass roll-off of a speaker even with a short window time, but there is very little frequency resolution at those frequencies, so resonances would go unnoticed.
Still seems like an interesting approach to minimize truncation errors, and it's a pity that ARTA doesn't have an implementation of it.
Looking at some of the data I'm wondering whether those using Holm Impulse and similar are using a sufficiently long (and single, not scaling) window time to capture all of the room reverberation. The FFT length needs to be at least as long as the RT60 of the room, preferably twice as long, so 500ms minimum, 1000ms preferable.
My measurements with ARTA were done with a full FFT length of 2700ms, which will be giving a true steady state response. A couple of the measurements I've seen look like they could be using a fairly short gate time of around 300ms, which would give an unfair bias as reverberant field would not have had enough time to build up in a room with an RT60 of more than 300ms. (EG just about any room)
This may be why some people are seeing significantly higher ratio than I am, and the fact that many people seem to be seeing around 10dB and I'm the only one using ARTA makes me think we're not comparing apples to apples due to different measurement technique. I know my room is bad, but maybe not that much worse than everyone else's
Can those people using impulse based measurements confirm what gate time they are using ? And if its less than 500ms please try again with a much longer windowing period, preferably 1000ms or more. Remember that we're trying to measure the steady-state room response, and this can only be done with a very long FFT/window time or a RTA.
Last edited:
I started to set up for measurements and realized I hadn't brought my probe mic home. In the end the only thing I had was a cheap and cheerful Sony lavilear stereo mic, so ignore the curve shapes and concentrate on the differences.
I also took a few different curves in my living room (main stereo room) and in the upstairs home theater.
I started with the living room system at about half distance at 5.5 ft, then moved to the normal distance of 10 ft from the speakers.
The whole wall behind the speakers is glass and sliding doors, but we also have very heavy lined curtains that can be closed. I repeated the differences with the drapes shut. A second curve has all 4 curves on one plot (sums and differences for drapes open and drapes shut). Notice two things: drapes shut loses a dB at HF (there is a rear firing tweeter that would be well absorbed plus probably some miscellaneous energy coming from the front, and drapes shut knocks the difference curve down by a full 2dB. Differences are mostly in the 7-8dB range.
This speaker is a Snell XA reference with symmetrical arrays over most of the range up to the tweeter. The tweeter is in a waveguide to help match directivity. Still, it looks like directivity maxes out from 1 to 5 kHz. the dual driver arrays should have an added 3dB d.i. (2 units add on axis 6dB louder but only 3dB in power response, hence d.i. +3dB).
I went up to the theater room. This is a large bedroom with some damping panels and thin rugs over hardwood. There is also a lot of glass (1 window, 4 closet doors), so I consider this a moderately lively room. I also had a problem getting a good null here, something was going on around 3k, either some unbalanced strong reflections or maybe the speaker balance wasn't as good as it should be. I did try face to face cancelation, primarily to balance the subwoofer level (built in powered subs), and the nulling seemed pretty good. Perhaps I'll redo this with a better mic and also use impulses to see what the early activity is.
Finally there is a difference curve for all cases. Based on what frequencyies you look at the range is from 9dB to about 4 dB over most of the range.
So where is Earl? I'm wondering if his high waveguide directivity is enough to offset that marble palace he lives in?
David S.
I also took a few different curves in my living room (main stereo room) and in the upstairs home theater.
I started with the living room system at about half distance at 5.5 ft, then moved to the normal distance of 10 ft from the speakers.
The whole wall behind the speakers is glass and sliding doors, but we also have very heavy lined curtains that can be closed. I repeated the differences with the drapes shut. A second curve has all 4 curves on one plot (sums and differences for drapes open and drapes shut). Notice two things: drapes shut loses a dB at HF (there is a rear firing tweeter that would be well absorbed plus probably some miscellaneous energy coming from the front, and drapes shut knocks the difference curve down by a full 2dB. Differences are mostly in the 7-8dB range.
This speaker is a Snell XA reference with symmetrical arrays over most of the range up to the tweeter. The tweeter is in a waveguide to help match directivity. Still, it looks like directivity maxes out from 1 to 5 kHz. the dual driver arrays should have an added 3dB d.i. (2 units add on axis 6dB louder but only 3dB in power response, hence d.i. +3dB).
I went up to the theater room. This is a large bedroom with some damping panels and thin rugs over hardwood. There is also a lot of glass (1 window, 4 closet doors), so I consider this a moderately lively room. I also had a problem getting a good null here, something was going on around 3k, either some unbalanced strong reflections or maybe the speaker balance wasn't as good as it should be. I did try face to face cancelation, primarily to balance the subwoofer level (built in powered subs), and the nulling seemed pretty good. Perhaps I'll redo this with a better mic and also use impulses to see what the early activity is.
Finally there is a difference curve for all cases. Based on what frequencyies you look at the range is from 9dB to about 4 dB over most of the range.
So where is Earl? I'm wondering if his high waveguide directivity is enough to offset that marble palace he lives in?
David S.
Attachments
That looks like excellent cancelation, an even 10dB across the board. Is your speaker really that flat or did you just draw it in with a ruler? EQed system?
David
Guys, I hope you realize there is some real science here. All we need is to define the correlation between power response difference level and the subjective impression of where a system lies on the spaciousness/clarity continuum, and we will have defined something that nobody else has really explored. At least not in this particular way.
David S.
David S.
I have one of those little gadgets, bought it to use with a Sony minidisc recorder.
David, looks like mostly 5-6dB across the board for you. The top end of the out of phase is smoother than in phase, like mine. I think that's a good thing as it tends to smooth out what we hear in the room, right?
Yes, where is Earl? He may be off in China.
Hi Simon,
I was using 128K@44.1 pulses, generated from Farina log-sweeps recorded with Acourate and which were 60(!) seconds long. 1/3rd global smoothing of that 3 seconds worth of room response was used for the ungated plots. HolmI was used for display only, but you can set up big FFT sizes (64K++) with it, too. Only use log-sine sweep with it when recording the pulse with HolmI
Attached find early (<7ms) and late gatings (7ms...50ms) of the same data. The ragged top end in the direct sound comes from the coax HF driver, a diffraction device.
BTW, A very nice experience is to listen to the pulses at 1/4th speed or slower.
- Klaus
Attachments
Last edited:
Ok that's fine. I guess its just my room set-up. Having the speakers on the short wall and sitting against the far short wall of a small room is definitely not conducive to a high direct/reverberant ratio
Although I didn't measure it, perceptually I had a far better direct/reverberant ratio when I had the speakers and listening position on the long walls - which results in a much greater distance from speakers to side-walls, and a closer listening distance.
Quite interesting how you're seeing sort of the opposite to me - you're seeing maximum ratio around 500Hz to 3Khz, with quite a reduction above 4Khz. A good ratio through the midrange due to the dual driver array perhaps, as you suggested ? Do you think the reduction at high frequencies could be the rear tweeter ? Does covering it up drop the reverberant field at the listening position significantly ?
On the other hand I'm getting the best ratio from 1.5Khz through the treble with a further increase at the high end of the treble. This is consistent with an 8" full range driver crossed at 4Khz with a wave-guide loaded ribbon.
The power response of the FR driver will be starting to drop already at 1.5Khz, and the 5Khz bloom in the power response of the tweeter is quite evident in the reverberant field. (I'm not sure what its cause is, perhaps some sort of slot diffraction effect, as it shows up at that frequency as a significant broadening of the horizontal polar response)
I'm still baffled by the large hump in the my reverberant field around 1Khz, I think it needs further investigation to see if its due to early reflections or not.
What sort of design are these upstairs speakers ? There doesn't seem to be much in the way of ratio at all in the treble with the two almost equal.
Has Earl ever provided any measurements of his listening room in any of these type of group measurements though ? I haven't seen any...So where is Earl? I'm wondering if his high waveguide directivity is enough to offset that marble palace he lives in?
I'd call it 7 in the living room and 4 in the home theater, at least above 1k. It probably should be much higher in the theater, especially now that I have 5 channels going! The living room sounds good to me, drapes open or shut (Actually I only notice the drapes when we talk. The speakers face away from them but you face towards them.)
Yep, got mine for minidisc also. Still have it but now I'm using a Tascam DR-07, really cute gadget and easy to use. Cheap at $100 or so.
David
I'd have to set up again to test that. Notice that the drapes behind made a dB difference in the summed response. Some of that has to be the rear tweeter.
As mentioned the XA Reference has symmetrical arrays of mids and lower mids. Up stairs units are Snell XA75 with MTM upper range, so they would have a similar d.i., probably a little more directional below the tweeter's range. I'm not totally sure the theater measurements are good. There were some big imbalances around 3k. I need to try again with Holm.
I did notice that your 2 direct curves had significant imbalance there. Couldn't guess whether it is the speakers or some nearby reflection.
David
Ah, but how do we describe our subjective impressions in a meaningful way
For my part subjectively I find my current listening set-up just barely within the critical distance - it seems to vary depending on small changes to the room like open or closed blinds or door, large cushions moved to different locations etc.
I do find the clarity is good at high frequencies with precise imaging, but not as good in the lower midrange where I perceive significant "room sound" and less imaging precision. This is where the ratio is worst on my measurement too.
As I mentioned in a previous post, with my room in the other orientation the subjective direct/reverberant ratio was considerably better, although I didn't attempt to measure it in this fashion at the time. If I had to guess I would say it would have been maybe 3dB better across the board.
Of all the listening rooms I've had this one is definitely the worst in regards to trying to achiving a good direct/reverberant ratio at the listening distance, at least in its current orientation.
Again if I was to guess I think I would classify 3dB average ratio as poor (3dB representing the critical distance) 6dB as ok to good, and 10dB as excellent.
More than 10-15dB may be venturing into the region where only a very low RT60 will achieve this, so it could be too much of a good thing. I would expect that anyone getting a result of 10dB across a wide range is most probably getting pretty good results for imaging precision.
One interesting question that could be difficult to answer is how much ratio is needed at different frequencies, and which frequencies are more important ? Is it more important to have a good ratio through the midrange, through the treble, or both ?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.