Its hard for me to measure the direct field accurately at 1Khz without moving each speaker onto a stand in the middle of the room to get sufficient window time, but last time I did this they were fairly well balanced through the 1Khz region as I'd applied a little bit of surround dip correction to both.
Certainly there isn't the 4dB difference in response which is showing up in the room response, they're within 1 dB of each other on axis through that region.
The speaker with the large bump at 1Khz is the right speaker, which has a strong ipsilateral reflection from a bare wall that's only 0.7 metres from the centre of the speaker, while the one without the bump is 0.9 metres with both a computer desk and a sofa along its wall providing a lot of diffusion. (The incident wall reflection is almost completely obscured)
(If you remember in the ETC curves from a while back, the ipsilateral reflection of the right speaker was less than 10dB down, while the left one was down below the reverberant field)
I'm hoping its just that side wall reflection that's causing the imbalance, and that the imbalance is resulting in incomplete cancellation in the Left-Right response, but I will be doing a bit more investigating to pin down the cause...
Dave,
it's a fully DSP-EQed active 2-way system (convolution based), nude 15" PA coax, and no other response than the one at l.p is ever used to tune the system, the next level below that is the raw drive units.
Complete magnitude EQ to invert the computed original "weighted psychoacoustic response", a sub-process in tha excellent "Acourate" software package I mentioned earlier, and that's where the specifics are, using wavelets in the analysis, freq.dependant windows and what have you... the software really is a powerfull (semi-automatic) toolbox for most anything you need for elaborated pulse manipulation for whatever purpose, besides a plain measurement/display tool.
I'm using subs occasionally (3way active, then, but with no traditional band splitting XO) when I need more level or low end, they're not shown here as I tend to have them off most of the time.
"Phase EQ" is there as well -- which means pulse compacting and interchannel recorrelation, in the end -- to an overall minphase target (as given per above magnitude), spiced up with a overcompensated low end where phase rolls gently back to even more than flat linphase at LF (see plot), this is my try to partly compensate typical bass roll-offs from highpasses of the recordings. I'm a timing and timbre fanatic, that means I can definitely hear the influence of phase and polarity -- now that I managed to set up speakers that can do it... and one can play with these things so nicely with convolution-based DSP processing, exectionally well suited for realtime blind AB and ABX tests you can conduct yourself, for me there is no way back. That's "scientific DIY" at its best, first hand real experience with most variables pretty well under control, I don't need any other proofs than that.
- Klaus
Attachments
That's massive!! I have a suspicion though that there was another problem... with some of the recent playing around I'd managed to hook up one of the tweeters with the wrong phase. I now get something more like what I'd have expected...though interestingly the in-phase/out of phase difference is now lower (but that could be because I moved the mic before I worked out the problem with the tweeter).
I have to get to work on the new crossover (one channel about 1/2 done).... the current one has poor phase tracking through the crossover region, the new one (in the sim) is very good. I'm very interested to see the difference. The other thing that is a bit confronting with this graph is how sharply the bass is dropping at 300Hz.... in nearfiled they are flat down to 200Hz and about -3db at 100Hz.... I don't yet have my subs running... the active crossover project needs to get some traction for that
Edit: the other very interesting thing is that the 2K suckout I'm always stressing about (in the gated measurements) is completely not there, and is in fact a bit of a peak in the in room measurement!!
Tony.
Attachments
Last edited:
Ah, but how do we describe our subjective impressions in a meaningful way
A proper way would be blind binaural tests with systems of more or less reverberance. At our level I think it is just as meaningful to ask: "for your system would you prefer more clarity or more spaciousness" and compare that to the measured reverberence level.
Of course we mustn't forget that we are taking the "macro" view and that direction and strength of early reflections is still probably more fundamental than simply a view of general reverberance.
Isn't it interesting that after 100+ pages, now we aren't just talking about the directivity of a speaker. We are seeing that we must consider the combination of listener distance, room liveness, and yes, speaker directivity. But the first two may trump the later and it isn't enough to proclaim that a particular speaker has just the right polar pattern. In fact 901s in a dead room might be fairly close in presentation to CD waveguide in a strongly reflective room. It isn't as simple as we thought...or then maybe it is.
I'm a bit surprised that the majority are finding they are well within the critical distance. My understanding was that it took a fairly dead room for that, but apparently not.
Regards,
David
Er.... impressive measurements you guys have, but I have a question.
The wavelengths of HF sounds are very short, so, large phase shifts should be easily seen by only small differences of path lengths. And, reflected sounds of multiple paths must be in quite large a proportion in the total energy received by a listener. So, to get uniform out of phase cancellation between 2 channels, except for good matching, I guess several other conditions are also needed:
1. very dead room
2. highly directional HF
3. pin point precise of listening position
....
Otherwise there're too many HF phase differences to be cancelled. No way to catch them all.
Maybe other conditions are also required, I don't know.
Pushing to the extreme (no offense, I'm not saying you guys are listening this way), using 2 long pipes from speakers to listener, the cancellation should be excellent. No?
OTOH, is that enough to show 'the ideal directivity pattern' of stereo speakers? Have I missed something? Sorry if that's the case. I haven't followed this very closely.
The wavelengths of HF sounds are very short, so, large phase shifts should be easily seen by only small differences of path lengths. And, reflected sounds of multiple paths must be in quite large a proportion in the total energy received by a listener. So, to get uniform out of phase cancellation between 2 channels, except for good matching, I guess several other conditions are also needed:
1. very dead room
2. highly directional HF
3. pin point precise of listening position
....
Otherwise there're too many HF phase differences to be cancelled. No way to catch them all.
Maybe other conditions are also required, I don't know.
Pushing to the extreme (no offense, I'm not saying you guys are listening this way), using 2 long pipes from speakers to listener, the cancellation should be excellent. No?
OTOH, is that enough to show 'the ideal directivity pattern' of stereo speakers? Have I missed something? Sorry if that's the case. I haven't followed this very closely.
Ok, here's my measurement. Note that the scale is 5db/div. Measured just in front of the listening position, approximately 8 feet from the speakers. Room is 17 feet deep by 22 feet wide. However, speakers are in the left half of the room (11 feet wide).
Looks pretty similar to most of the curves being shown here. Speaker is a vented 15" crossed (LR4) to a 2 inch compression driver on a 90x40 horn at 1 kHz.
Looks pretty similar to most of the curves being shown here. Speaker is a vented 15" crossed (LR4) to a 2 inch compression driver on a 90x40 horn at 1 kHz.
Attachments
Last edited:
Yes Pano. I built the boxes recently and the idea is to cross them to these paper mâché horns: http://www.diyaudio.com/forums/multi-way/190326-building-johns-paper-mache-horns.html
The paper horns are not quite done yet and the 90x40 horn is a temporary solution.
The paper horns are not quite done yet and the 90x40 horn is a temporary solution.
That's the whole point of the measurement - when we measure out of phase only the direct signal from the speakers will cancel completely, all reflections, except perhaps perfectly symmetrical early reflections (like the floor bounce) will arrive at essentially random phase angles and time delay and therefore will not cancel.
What we're left with is a good approximation of the reverberant field.
When we measure in phase we get the direct signals summed together (6dB addition) plus the reverberant field.
Last edited:
The intention isn't to cancel everything in the room, but to simply cancel the direct sound to reveal the level of the reverberent sound. A deader room or more directional speaker would make the direct to reflected ratio higher yielding more cancellation (a lower difference curve) but that is not necessarily the intention, rather just to see what the ratio is.
Pushing to the extreme (no offense, I'm not saying you guys are listening this way), using 2 long pipes from speakers to listener, the cancellation should be excellent. No?
OTOH, is that enough to show 'the ideal directivity pattern' of stereo speakers? Have I missed something? Sorry if that's the case. I haven't followed this very closely.
The "2 long pipes" would be a case of extreme direct-to-reflected ratio. (Same as near field monitoring or even headphones.) We shouldn't presuppose that that is the desired case. The idea is more to have a simple measurement of direct-to-reflected ratio and then see if there is a consensus that some range is best.
David S.
KEF had a fairly large room (I think 10 x 10 x 10m cube) for transient testing. We found that a 2nd order at 200(?) system would have a transient tail that settled quickly enough to have little or no truncation error before the first wall reflection. If you could get a system to look like a 2nd order highpass at 200 Hz you could have a perfect measurement.
All you had to do was: Measure impedance, best fit an elaborate model to the impedance curve, predict LF response from the model, best fit a 2nd order highpass to the model, calculate a pole canceling filter to replace real poles and Q with 200 Hz (BDC or bass diminishing circuit), measure and truncate, multiply times the inverse of BDC and....perfect measurements as simple as that.
I think Laurie Fincham and Mike Berman wrote up the procedure in one of their papers.
David S.
Very true. RT60 or the difference measurement we've taken don't answer what is really important and that's level, time, angle, spectrum and number of first reflections. We must also not forget that all of this depends on absolute SPL and not just on relative values.
But the overall level of the reverberant field is important, it's just a separate factor to level, delay and direction of early reflections.
If your listening position is beyond the critical distance and the direct field and early reflections are all below the reverberant field level, it doesn't really matter too much what the early reflection patterns are - you won't get precise imaging or good clarity.
On the other hand if the direct field and early reflections are significantly above the reverberant field then the characteristics of the early reflections will dominate the listening experience in terms of clarity and imaging.
As far as the connection to absolute SPL values - obviously the ratio we're measuring doesn't change with absolute SPL, but our perception of the ratio between direct and early reflections and the ratio between direct and reverberant field does change with level.
If you're only just at or slightly in front of the critical distance, at low SPL the reverberant field may not be too noticeable however at high SPL levels I find it becomes a lot more prominent.
In other words maintaining good clarity at high SPL requires a higher direct/reverberant ratio than low SPL playback. That's certainly been my subjective impression in many different listening environments.
Last edited:
Looks like the inverse method to what JohnK describes here : Matched Filters
Both methods aim to have the initial pulse decay low enough before the first reflection shows up and might have some s/n-ratio issues that must be controlled.
I have never tried these tricks so far, though, as I've been doing OK with nearfield measurements right at the cone (and port, if any), added time coherent after matched the slopes, and then spliced with the 1m (or whatever) measurements (which needs linear phase filters and manual aligning of the pulses, too. The latter is a home-run for HolmI, thanks to the realtime shift and filter functions).
That's why I'm trying to use virtual rooms at will, using my pretty dry and damped listening setup (too dry for me, RT60 at 0.3s broadband, with most music and the levels I use). Except for angle which is limited in virtual room emulation to what can be safely done with HRTF-based re-localization, I have a quite big freedom of choice in how much and what kind of soundstaging and envelopment I can possibly dial in. Plus it makes tons of fun and is educational too.
What we perceive is all that counts, everything else is meaningless. There is no diffuse sound field in acoustically small rooms, simple D/R measurements aren't very meaningful.
Toole ("Loudspeakers and Rooms for Sound Reproduction—A Scientific Review"): "In a small listening room we are in a transitional sound
field, consisting of the direct sound, several strong early
reflections, and a much diminished late reflected sound
field. What we hear is dominated by the directional characteristics
of the loudspeakers and the acoustic behavior of
the room boundaries at the locations of the strong early
reflections. RT reveals nothing of this. As a measure, it is
not incorrect; it is just not useful as an indicator of how
reproduced music or films will sound. Nevertheless, excessive
reflected sound is undesirable, and an RT measurement
can tell us that we are “in the ballpark,” but so
can our ears, or an acoustically aware visual inspection."
I think there is a lot of truth in that - polar response is a tool to help you achieve the desired early reflection pattern and ratio, and direct to reverberant ratio, in situations where other parameters such as listening distance or room characteristics may be limited in the amount they can be optimized.
So from that I think its fair to say that if there is any conclusion from 100+ pages of discussion, there probably is no individual ideal polar pattern that is best for all circumstances, and that it is these other factors (along with preference for imaging precision vs ASW) which are guiding our selection of directivity, even if subconsciously.
However I think it would also be fair to say that although there is no one ideal, unless you are going for a diffuse field effect some directivity is desirable - eg omni's are out. On the other extreme a speaker with extreme beaming is not going to interact well with a room or provide a good listening experience over a reasonable range of seating positions.
So there is a "range of useful directivity" somewhere in the middle, and the debate is more about what the upper and lower limits of that useful range might be.
One practical advantage of directivity beyond the interaction with the room though is helping to control cabinet diffraction. Unless you have a carefully designed and fairly radically sculpted cabinet (B&W Natilus 802, KEF Blade etc) diffraction is always going to be a problem for extremely wide dispersion drivers, with typical dome tweeters on a flat baffle being probably the worst offenders of all, and yet most commonly used.
Drivers with enough directivity control to keep the polar response down below 120 degrees or so will give a significant reduction in baffle diffraction without much loss of power response, and if the polar response is fairly uniform with frequency the reduction in power response will be fairly uniform, and thus beneficial.
This was brought home to me recently while doing some diffraction measurements and experiments with my ribbon tweeters. They have a round 110mm faceplate and the wave-guide depth to ribbon element is a relatively small 25mm or so.
Despite being such a small wave-guide the diffraction effects of an un-baffled tweeter are rather small. Mounted in free air with no baffle there is a broad smooth on axis dip of about 1dB at 6Khz and about half a dB rise from 3-5Khz, compared to being mounted on a large flat baffle.
This is a very minimal amount of diffraction for an unmounted driver, try the same test with a dome tweeter and the on axis cancellation from its circular un-baffled faceplate will be horrendous. Although I can hear the difference, the ribbon sounds just fine un-baffled.
Of course you would still want to deal with what diffraction there is with an appropriate baffle and/or some absorption, but the narrower polar response makes this job so much easier, especially for DIY designs that don't look like art sculptures...
Keeping baffle diffraction down to a bare minimum especially at high frequencies could be a valid reason to use some directivity control even if you're not particularly looking for a high directivity design from a room perspective.
I think there is probably a lot of self-selection bias in our small survey. The majority of people who will be actively participating in a thread on "What is the ideal directivity pattern for stereo speakers" will be those that probably believe some directivity is beneficial, so they will already be aware of and thinking along the lines of direct to reflected ratios, etc, when designing and setting up their systems and rooms.
Also at least two of the respondents have large format horn systems, which are naturally going to perform well in this test. Your Snell Reference speakers are not typical of the type of polar response that the average joe's speakers will have either, and neither are mine for that matter.
If you were to take a survey of the speaker/room setup of the general population then yes I'd expect to see pretty poor results, with many people probably listening beyond the critical distance with perhaps an average hovering around the critical distance.
Even a lot of "audiophile" systems belonging to non-speaker building audiophiles with limited measurement experience or acoustical knowledge are likely to have a less than ideal ratio.
Last edited:
This was Earls position a few pages ago when he attacked RT60 of small rooms, and I think it was reasonably well rebutted by Dave S. The same Toole quote was also used at the time, perhaps taken slightly out of context of its original meaning.
How can there not be a diffuse field in a small room ? Above the schroeder frequency a small room behaves the same as a large room, its just the decay is a bit quicker because of the shorter path lengths.
A typical 400-500ms decay time is still a long time compared to the modulation speeds in music, so of course its meaningful. I can hear the difference between a high direct/reverberant ratio, and a low ratio, can't you ?
If we can hear it, we need to characterise it objectively some way. Both the measurements we've just been taking and RT60 give some indication of it although they are not identical, as RT60 only includes the decay of the room, whilst our reverse phase measurement also includes the effects of speaker directivity in feeding that reverberant field, and the overall ratio.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.