Good work Tony and very interesting. You seem to maintain about 5dB across the band. Similar to my measurements, but less of a null. I worked harder at the mic position, I think.
It it coincidence or that fact that we both took a very similar, methodical approach to our passive crossover designs? Certainly the speakers themselves are very different.
It it coincidence or that fact that we both took a very similar, methodical approach to our passive crossover designs? Certainly the speakers themselves are very different.
One of them is on my website Polar Map App (which you probably can't run because it requires Windows). The other I have not publsihed.
That one? Doesn't look too bad if one is looking for a very wide dispersion chassis. What driver is it?
An externally hosted image should be here but it was not working when we last tested it.
I carefully removed the bug screens. Mind you, out of sheer curiosity, not for any specific acoustic reasons (there was only miniscule difference in the LF and HF response with and without). Close-up pic and full view attached.
The brown-ish color change comes from the light used (flash and regular bulb), in reality it is rather gray-ish.
Attachments
Last edited:
Hi Pano, I'm going to have to have another go tonight. apart from just sticking the mic randomly where I thought might be the centre, after I went to return the speaker cables to normal I noticed that the negative wire on the left channel had snapped off (not a mean feat for stranded cable) and was hanging down not connected. I think it must have still been connected when I did the measurement but it is possible that it came off when I was changing the phase of the right channel, and that my second measurement is just one speaker running. )or would that only result in 3db difference?)....
oops gotta run will miss the train!
Tony.
Both signals are coherent (they must be to make this measurement meaningful at all) and therefore add up to +6 dB. But then we have a problem with the situation at the cursor position ...
Hi Rudolf, Yes I thought that was a little odd too. why just at one frequency? around 200Hz which is coincidentally (or perhaps not) the frequency that the raw drivers start rolling off in these enclosures.
I've also realized this morning that the hump at 560 Hz explains why I wasn't seeing a normal smooth baffle step drop with the first crossover. It was being filled in by that hump and then dropped like a stone at 300Hz. I'm wondering if the 560 Hz bump is actually due to the TV screen which is between the speakers. It is an 81cm CRT and has a slightly convex screen (ie not dead flat) I'm wondering if it is extending the baffle and acting like a bit of a lens at that frequency. (raw driver nearfiled for the mw144's is ruler flat from 200 - 1Khz)
and pano, yes I think that you might be onto something (assuming my measurement was valid) and perhaps the crossover itself plays just as big (or bigger role) than the type of drivers or enclosures?? BTW are your horns time aligned? My MTM's I believe are.
Tony.
I've also realized this morning that the hump at 560 Hz explains why I wasn't seeing a normal smooth baffle step drop with the first crossover. It was being filled in by that hump and then dropped like a stone at 300Hz. I'm wondering if the 560 Hz bump is actually due to the TV screen which is between the speakers. It is an 81cm CRT and has a slightly convex screen (ie not dead flat) I'm wondering if it is extending the baffle and acting like a bit of a lens at that frequency. (raw driver nearfiled for the mw144's is ruler flat from 200 - 1Khz)
and pano, yes I think that you might be onto something (assuming my measurement was valid) and perhaps the crossover itself plays just as big (or bigger role) than the type of drivers or enclosures?? BTW are your horns time aligned? My MTM's I believe are.
Tony.
You know, in all honesty I forgot the brand on that driver. I did it for someone else who had done the design and I just don't recall what the drivers were, just that it was a coax horn on a woofer.
I've seen worse responses thats for sure, but I have seen better as well. In a pinch, where space is at a premium, coax is very attractive, but the belief that they somehow "solve some serious problems" is clearly not correct. They create more problems than they correct.
There is an appeal to having a unit with no off-axis crossover holes, but most of them are poorly implemented.
I was around when the KEF UniQ line came out. They did a fair job but the cone and edge obstructions gave a messy top octave.
My best results were with a waveguide/flare on a tweeter (PSB CW180). You could get good tweeter response and use directivity to kill the reflection off the woofer cone behind. (see pic.) The BBC approach of mounting the tweeters on a perforated plate in front of the woofer seemed to work well. Also, JBL has some "pendent" speakers that mount a waveguide with LF slots in front of the woofer.
Regards,
David S.
Attachments
Hi KSTR,
Maybe you could try to change the profile of this horn. It's possible to make a kind of half donut (or half tore) because you have enough space between the horn exhaust and the coil. It would be relatively easy to do it with wood, and with a better profile than the half round of my scheme. Only the back oblique part will require some adjustment skills. Or maybe it's disposable and then you can made an easy flat base ?
You will not harm the driver, and if it's not positive, you will at least be sure that the designers did a good job.
Maybe you could try to change the profile of this horn. It's possible to make a kind of half donut (or half tore) because you have enough space between the horn exhaust and the coil. It would be relatively easy to do it with wood, and with a better profile than the half round of my scheme. Only the back oblique part will require some adjustment skills. Or maybe it's disposable and then you can made an easy flat base ?
You will not harm the driver, and if it's not positive, you will at least be sure that the designers did a good job.
Attachments
Last edited:
Only a quick test tonight, but after it I suspect my test last night had only one driver running for the second test. I was able to reproduce a very similar result tonight only by running both drivers then disconecting one. When I reverse the phase of one it is all over the place!! interestingly tonight I don't have the large 560hz bump, It may have been the bicycle helmet I was using as a mic holder
I was running with a much lower spl tonight as well, but almost deafened myself when I quit REW and started getting massive feedback!
ugly response follows, more testing required!!
1st pic clearly not right. 2nd pic with left channel disconnected.
Attachments
A temporary respite, but no less busy down the road. I wish that I had time to do more measurements of my room, but I just never seem to get that.
I would like to point out that in the limit, your approach does not work. For example, in a perfectly symmetrical situation, superposition requires that the entire field cancel on the centerline, not just the direct field. Doesn't an impulse response show the result just as well, if not better? A few years ago this may have been more difficult, but today its so easy (well easy enough once you get the time).
I would like to develop an impulse response test that did not take all reflections in equally. A sort of baffle (actually a piece of foam might be better), with the mic on one side to minimize contralateral reflections. Then an average of this done for left and right ears. I am not trying to simulate head diffraction just the ipsi/contra-lateral reflection issues that I believe are a strong influence.
This is probably true in theory but in practice it isn't the case. I can't imagine any room with symmetry so good that a 2nd or 3rd bounce of any left/right rays would be close enough to cancel. A bit of a "butterfly effect" that any difference, ever so slight, would get rays off on diverging paths and prevent cancelation. It is true that first bounces of high symmetry (say the floor bounce) would be cancelled in the test. You would have to decide philosophically whether you think the floor bounce is reverberent field or direct sound, as it can be argued both ways.
No problem with measuring and gating the impulse response other than you will have overlap between the LF part of the direct response and the onset of room reflections, so simple gating won't split them cleanly.
Remember, this came about as a byproduct of seeking an easy method of 2 speaker time allignment and developed into a way to guage the direct to reflected level that anybody, with an RTA, could measure. If nothing else, it is very good at illustrating the different fields, direct and reverberent, and revealing approximate levels of each.
David S.
Last edited:
It's a chaotic system isn't it ? Like weather. The slightest imperfection in the symmetry in the room, (room shape, objects, speaker positions) will cause a massive snowball effect that means the rays are unpredictable by the time they've reflected around enough times to be considered part of the reverberant field.
Any furnished room is going to have enough imperfection and asymmetry for this to happen.
What I find interesting (and a bit puzzling) is that even at low midrange frequencies a similar degree of randomization seems to be present - as seen by good ratios at those frequencies by some like Pano, ratios not hugely different to the higher frequencies.
The reason I find that puzzling is that high frequencies should become phase randomized much more easily due to their very short wavelength relative to path lengths, and require less room asymmetry to "mix up" their phase, yet the results at lower frequencies (at least above 200Hz or so) seem to be similar, suggesting that there is far more "randomization potential" present in the room than required to sufficiently randomize the phase. (Think of it like the quantity of "entropy" available when generating good random numbers - once you have "more than enough", adding more entropy doesn't make the random numbers any better)
I would argue that the initial floor bounce is not part of the reverberant field because -
a) It's apparent source location is very close to the actual speaker - in fact it forms a mirror image, and could be considered part of the speaker. Reverberant field shouldn't have an apparent source location.
b) It has a predictable phase relationship to the direct signal, at least at lower frequencies before the path differential becomes too great in number of wavelengths. Reverberant field should have a completely unpredictable phase relationship to the original signal above the Schroeder frequency.
For these reasons I think this test does the right thing by cancelling out the floor bounce effect from the measurement.
I agree. It's not a perfect measurement, and in theory it requires some imperfection in the room to actually work, but in practice it seems to be a pretty good approximation that shows useful information.
Sometimes perfect is the enemy of good.
Last edited:
Yes, an easy and effective test for time alignnment of two channels. But I would not see much use in further development beyond that role.
I was glancing at Zwickers book the other days and it dawned on me that critical bandwidths change with frequency and the perceptual gating times should be interrelated through the Fourier transform. The narrower critical bands at LFs indicating that a much longer time resolution aspect is required - consistant with results that you keep quoting. The gating time is constant up to about 500 Hz and gets gets shorter and shorter with frequency above this, a relationship that I keep quoting. The nice thing about the Zwicker data is that it puts actual numbers to this discussion, 10 ms up to 500 Hz rising to about 1 ms at 5 kHz, etc. It seems to me that one could take the impulse response and "smooth it" via the critical band filter widths and then inverse FFT it to get back the time signal for the room that is perceptually correct, i.e. the reflections would be weighted by an estimate of the effect that they would have on the perceived frequency response.
I think that with the vastly improved data taking capabilities that we have today (HolmImpulse) that the time has come to look at better measurement techniques. Balancing two speakers by nulling the frequency response just seems to me to be pretty simplistic albeit necessary.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.