Linkwitz Orions beaten by Behringer.... what!!?

This MTF measurement discussion is quite interesting; I've got some reading and learning to do :D


I found some links that may be useful to others:

PLOS Computational Biology: The Modulation Transfer Function for Speech Intelligibility

This has a seemingly detailed description of the analysis that they performed. Also has attachments of the signals used for assessing MTF<->speech intelligibility link. They modulate a recording of a spoken phrase, which may not be of much interest here other than as a clue for what input signals should be used perhaps?

Since I use Python a bit for my day job, reading code helps me - https://gist.github.com/stefanv/2051954 - this is an implementation of MTF processing of an image


I also found this link that is referenced in the above code interesting for background understanding - Algorithm for MTF Estimation by Histogram Modeling of an Edge

Chris
 

ra7

Member
Joined 2009
Paid Member
Here's some data on MTF as promised.

Again, the two speakers are JBL2445 on a large tractrix horn and B&W DM602.

For making these measurements, the speaker is in the normal position where they would be if I was listening to music and the mic is the listening position pointing straight ahead (between the left and right speakers).

ARTA says the STI should be measured at 68 dbA. Unfortunately, this was not possible because the minimum level in the impulse window is -20 dbFS. This means the hugely efficient horns were pretty loud, about 80 dbA. And so, both speakers were measured this level.

First couple of pics show the two impulse responses measured at the listening position. The horn impulse clearly has reflections at a lower level than the B&W. (The first pic is the horn impulse, even though it says B&W)

Next pics show the MTF for each speaker. Again, ARTA doesn't allow overlay in these charts, so they have to be exported one by one. Not much difference here between the two speakers. Speaker identity is in bottom left corner of charts.

Finally, we have the STI chart. Again, not much difference between the two speakers. Last chart is the B&W.

I can send the pir files if anyone is interested.
 

Attachments

  • horn_imp.jpg
    horn_imp.jpg
    90.8 KB · Views: 263
  • bw_imp.jpg
    bw_imp.jpg
    83.9 KB · Views: 265
  • horn_MTF_500Hz.jpg
    horn_MTF_500Hz.jpg
    80.4 KB · Views: 233
  • bw_MTF_500Hz.jpg
    bw_MTF_500Hz.jpg
    82.1 KB · Views: 232
  • horn_MTF_1000Hz.jpg
    horn_MTF_1000Hz.jpg
    82.2 KB · Views: 229
  • bw_MTF_1000Hz.jpg
    bw_MTF_1000Hz.jpg
    82.7 KB · Views: 83
  • horn_MTF_2000Hz.jpg
    horn_MTF_2000Hz.jpg
    81.2 KB · Views: 81
  • bw_MTF_2000Hz.jpg
    bw_MTF_2000Hz.jpg
    82.2 KB · Views: 80
  • horn_STI.jpg
    horn_STI.jpg
    129.7 KB · Views: 77
  • bw_STI.jpg
    bw_STI.jpg
    131.7 KB · Views: 71
Last edited:

ra7

Member
Joined 2009
Paid Member
RA7, thanks for the measurements. It's easy to see the differences - but which speaker do you think sounds more dynamic? Or are they basically equal? I've often wondered if a dynamic passage were recorded from 2 different speakers would they actually show differences in amplitude? I mean average to peak amplitude.

Pano, the horn sounds infinitely more dynamic than the B&W. And no, I don't mean "loud" by this, as has been suggested. But its hard to find a measurement that correlates with this perception.

At lower volumes, it is easily more intelligible than the B&W. Hmmm.... maybe there's something there.
 

ra7

Member
Joined 2009
Paid Member
The STI can be calculated from an in-room impulse response, though it takes a large one (over one second if I remember correctly). Omnimic uses a long chirp to get the IR and then gets the STI from that.

The process is given at:
A Do-it-Yourselfers Guide to Computing the Speech Transmission Index Synergetic Audio Concepts

Right. The ARTA manual suggests appropriate FFT lengths depending upon sampling rate. No averaging is allowed because this reduces the noise, and we don't want that for the STI.

It's all there in the ARTA manual.
 
Pano, the horn sounds infinitely more dynamic than the B&W. And no, I don't mean "loud" by this, as has been suggested. But its hard to find a measurement that correlates with this perception.

At lower volumes, it is easily more intelligible than the B&W. Hmmm.... maybe there's something there.

Certainly, it may not even be the speakers that cause a problem. Most often I would improve a deficiency based on engineering understanding, then figure out what changed from measurements, and use the measurements to control how the improvement is optimized. Listening is very important because our brain processes lots of information based on experience accumulated throughout our life, no kind of measurement or data processing can do this.
 
Just a quick cursory look at the charts makes me think about a few things. The RTA of the room could play a big part in the intelligibility of the signal. This would have much to do with the different directivity of the two different types of speakers and the other is the impulse response of the horn over the direct radiator. The distortion of the horn just looks much lower than the cone drivers also. All of these things would add up to the commonly acknowledged preference for horns over direct radiators, especially in vocal range coherence.
 
Soongsc,
How can we leave the room out of the equation? If the test was in an anechoic chamber or outside I have not problem with doing that, but not if you are testing in a room unless you are using some gating function to only look at the first response without the room additions. Do you have any other ideas here?
 
So the modulation frequency is sort of simulating the human voice characteristic? What I don't understand is that since the audio speaker normally filters out the subsonic modulation, in effect, the resulting sound had very little modulation, thus MTF is going to be high, would it not? The only problem is that if the driver movement moves to the nonlinear region due to the subsonic signal, then you start creating more harmonics. So what is the true benefit of using the subsonic modulation to drive the test?
 
Last edited:
Soongsc,
How can we leave the room out of the equation? If the test was in an anechoic chamber or outside I have not problem with doing that, but not if you are testing in a room unless you are using some gating function to only look at the first response without the room additions. Do you have any other ideas here?
If we are looking to evaluate the speakers themselves, I would use only the gated section. Now, I may be wrong, but it seems to me there are a few things that is going to effect the playback which should show up in an MTF depending on how it is tested.

1. Delayed release of stored energy of the system.
2. Nonlinear effects of the driver material and the motor.
3. Diffraction from the driver, and enclosure, and any part that is not moving in phase with the voil.

What we really want to do is to evaluate how the combination of these effects the transfer of music signals.

Now, if we look at the Orions and the Behringers, one uses waves from both sides of the driver, and the other has large ports close to the drivers, so it would really be interesting to see how these different design concepts show in the MTF. But I think we should use a gated part of the impulse, and see how a specific sound spectrum is effected by this part of the impulse. Not sure how it would be done mathematically though.
 
So the modulation frequency is sort of simulating the human voice characteristic? What I don't understand is that since the audio speaker normally filters out the subsonic modulation, in effect, the resulting sound had very little modulation, thus MTF is going to be high, would it not? The only problem is that if the driver movement moves to the nonlinear region due to the subsonic signal, then you start creating more harmonics. So what is the true benefit of using the subsonic modulation to drive the test?


In order to understand MTF it may be in order to review what is modulation
Amplitude modulation - Wikipedia, the free encyclopedia
 
What I don't understand is that since the audio speaker normally filters out the subsonic modulation,
No, that's not it. There are no subsonic "tones" to be filtered out . . . what is being looked for is amplitude modulation of higher frequencies (divided into numerous perceptual "bands"). The argument for "speech" is that that's where the actual "information" is . . . and the argument for music is that that's where we (might) find what (some) people call "dynamics".

If that is indeed the case for music then it's not a function of the speaker per se, it's a function of "modulation leveling" by reflected sound . . .
 
Last edited:
Hi All
Elias, music is slow or can be anyway, I think the things which tell you “there it is”, are fast, related to short events. While speech might be centered around 3 Hz (don’t recall) I do know they were measuring the modulation rates up to about 12 or 15 Hz and that like optics, the higher rates tended to suffer first /most. At the semianr, I had asked Sander van Wijngaarden how high a rate they investigated and he said about 30Hz.
I wonder how high a rate is audible though, for example, when you tune a 12 string guitar or piano, you DO NOT tune the groups of strings to the same note, they are ideally slightly different because that “beating” or fast amplitude modulation, sounds nicer than a pure tone.

Soongsc, you hit the nail on the head, the amplitude modulation of each of the frequency bands is a signal which emulates talking or rather how the information in the voice (what words) is transmitted or lost. If you examine a spectrogram (frequency, amplitude, time) of speech, you can see it is largely changes in pitch AND amplitude, we often measure amplitude and time(at one level) but rarely measure anything to do with changes in amplitude, so to me, it looked like an ignored dimension, ripe for investigation.
Best,
Tom
 
I would like to note that based on the techniques in the links, the MTF will not be sensitive to nonlinearity at all. It would be somewhat sensitive to thermal effects. As defined, however, it will depend almost exclusively on the speakers directivity and the rooms reverberation. Horns will win this every time. But this metric could be a good correlate to "dynamics". However, it is just a single factor in that it ignores an awful lot of stuff that we know are factors, like frequency response.

What I do like about it is that it would show how the room and speaker work together in the specific room. But since it is so strongly dependent on RT, comparisons of speakers in two rooms with different RTs would be meaningless. The lower RT will have a significant advantage. This also makes clear that good MTF and good spaciousness would be counter to each other - which may also be very true, just as it is true for spaciousness and imaging being counter to each other.
 
No, that's not it. There are no subsonic "tones" to be filtered out . . . what is being looked for is amplitude modulation of higher frequencies (divided into numerous perceptual "bands"). The argument for "speech" is that that's where the actual "information" is . . . and the argument for music is that that's where we (might) find what (some) people call "dynamics".

If that is indeed the case for music then it's not a function of the speaker per se, it's a function of "modulation leveling" by reflected sound . . .
Well, there are interactive waves in a room, I would not say those are modulating simply because the phenomena is different. What we are really trying to do is how accurately is the signal transferred to the receptor. This is also the intent of it's application in video.