Single sheet TH challenge

[tb46]

Hi David,

I agree with your statement in Post #1315: "...would be nice to see some actual measured results for two or more identical speakers connected in series...", and I wish I had access to time, speakers and measuring equipment, but at this time I don't.

You already found the major problem, sound power v. sound pressure addition.

Going back to acoustic radiation resistance, it just has to be taken into account. Obviously, you already take it into account for the single driver case, and thus in the multiple case you should only see a shift of the turn-over point where the curve goes from the low end 6dB slope into the constant value high end, or approximately "ka=2". This effect - the shifting down of the ka=2 point - may be small enough to be hard to extract from a measurement, so it may look like a 6dB shift throughout, with possibly a small boost in the low end.

Regards,

[weltersys]

Originally Posted by David McBean

It would be nice to see some actual measured results for two or more identical speakers connected in series, to establish whether or not the SPL actually does stay the same as that for a single speaker - for a constant voltage input.

Kind regards,

David

Below are measurements of a dual Lab 12 BR cabinet, then two in series, same voltage done 2/18/08 (prior to my getting Smaart).
The SPL measurements were averaged up or down to the nearest whole numbers appearing on the dB meter.
The LF are basically the same SPL, the 90 Hz difference of 2 dB may be a clerical error, the 1-2 dB difference above 110 dB probably could be attributed to slight beamwidth narrowing with the larger array.

My stage monitors use a pair of 8” in series, the pair measures the same as a single driver given the same voltage.

For Hornresp purposes, using equal SPL level for 2 speakers in series compared to a single given the same voltage is correct.

As another data point for FLH, the chart below shows one (raised 6 dB) and two of my Chorns (with a BW25/125 Hz filter), a cabinet like a shrunken Labhorn.
It agrees with the slight LF extension and slight HF attenuation or smoothing that is typical of multiple FLH. The 2003 test results show multiples of 2, 4 and 8 cabinets, a 13 dB range from 31.5 Hz to 160 Hz with a pair reduces to only 5 dB range with 8 cabinets, 31.5 having increased by 16 dB.

Thanks to DJM for doing the FLH/sim comparisons, I have included them again below.
I had been thinking Hornresp exaggerated the LF response of TH in multiples, while it appears in DJM’s comparison of the FLH that the LF agrees, while upper addition is in error.
Oddly, it does not appear that the LF corner lowered, while it does in the measured response.

Looking at Ivan Beaver’s 1, 2, & four TH-115 measurements, it is apparent that the TH LF corner does not change in multiples, but the upper response smooths out and does not increase a full 6 dB.

Ivan gave no indication of how the cabinets were arranged, arrangement can make a significant difference at the 10’ measurement distance that test was conducted.
Four cabinets with the mouths on the ground, four stacked 2x2 with mouths together, and four 2x2 with mouths apart will all give different upper frequency response.

The difference in array configuration will account for some differences between measured and Hornresp, the simulation (as far as I know) only considers the horn mouth as the size of the exit.
In actual build FLH generally have a mouth area close to cabinet frontal area, while TH often have a mouth area only half or less of the frontal area.

Large array size and shape make a big difference in response, programs like Meyers MAPP take those features into account.

Hornresp works very well for design, but large scale interaction really needs to be measured, even MAPP only works as a 2 dimensional model, while we live in a 3D world.

Art Welter

[Wayne Parham]

I popped over here to read everyone's replies today and re-read my own. I realized that some of it may have been confusing, so I want to clarify this point:

Originally Posted by Wayne Parham

I don't think using a straight 6dB bump when doubling horns is the best approach. Maybe 3dB is a smidge too much, but then again, maybe not. You've seen several datasets that show the same trend, and it's at least 2dB, even with large horns. Seems like smaller horns show this effect even more, but I haven't scrutinized them. Might be worth lining them all up, side-by-side like Djim did, to look at them closer...

To rephrase that part:

I don't think using a straight 6dB increase when doubling horns is the best approach because it doesn't match measurements. Maybe rolling off the top end by 3dB is a smidge too much, but then again, maybe not. We've seen several datasets that show the same trend, and it's at least a 2dB slope, even with large horns. Seems like smaller horns show this effect even more, but I haven't scrutinized them. They also look like they gain a decibel or two more than 6dB on the bottom end, not just lose a decibel or two on the top. Might be worth lining up all the relevant measurements available, side-by-side like Djim did, to look at them closer.

There's definitely an effect (or set of effects) that cause bass to get more increase (or higher frequencies to get less increase) when grouping basshorns closely together. My thinking is it's one or more of these three things, or perhaps a combination of all of them:

1. Mouth size getting closer to acoustical scale at LF (causing an efficiency increase at LF)
2. Reduced summing at HF because of acoustic distance (causing phase/summing reduction at HF)
3. Shift in the movement of the acoustic center (causing distance to change as a function of frequency)

I suppose the best approach would be to do more analysis to see which of these effects made the most difference, or if they all have to be calculated in the model. It would require the sizes and positions of the sound sources to be entered, and a set of formula that expressed these conditions to be applied. But I think that until and unless this kind of analysis can be done, probably a hard-coded "fudge factor" can be used. I don't think the increase should be 6dB across the board, but I also don't think a fixed rolloff slope always tracks measurements either. I think it might make sense to examine the data that you have, and see if an interpolated slope offset that takes mouth size into account might make sense.

[Djim]

I have also worked out Ivan Beaver's measurements of the DSL-TH115 in stacks to an higher resolution to get a more precise picture.

First the original:



Second with +12dB for 1 TH115 and +6dB for dual stack:

[tb46]

Hi Djim,

The one in Post #1324 (and in Art's Post #1322) would be a great example if it didn't have obvious measurement problems in the red curve. The 80-120Hz hump appears compressed, and the 150(?)Hz dip has almost disappeared, that looks more like a measurement problem than proper combining of SPL values. So what else is wrong with that particular data set?

This stuff is not just difficult to analyze, it's also really difficult to measure that accurately.

Regards,

[Djim]

Originally Posted by tb46

Hi Djim,

The one in Post #1324 (and in Art's Post #1322) would be a great example if it didn't have obvious measurement problems in the red curve. The 80-120Hz hump appears compressed, and the 150(?)Hz dip has almost disappeared, that looks more like a measurement problem than proper combining of SPL values. So what else is wrong with that particular data set?

This stuff is not just difficult to analyse, it's also really difficult to measure that accurately.

Regards,

Hi Oliver,

I agree, the red line looks way too flatten out and limited above 80Hz. Probably a mic and/or mic amp problem of the measuring system. 'Normally' you would expect an downward motion is followed by upraise motion (you always see such balance). The downwards motion is way bigger than the upward motion in the dip which is a prove of compression in the measuring system. However below 80Hz I do believe the red line is accurate and follows the principles of horn size theory. Also the dual stack seems to keep it's level below the compression threshold of the red line.

[weltersys]

Originally Posted by tb46

Hi Djim,
The one in Post #1324 (and in Art's Post #1322) would be a great example if it didn't have obvious measurement problems in the red curve. The 80-120Hz hump appears compressed, and the 150(?)Hz dip has almost disappeared, that looks more like a measurement problem than proper combining of SPL values. So what else is wrong with that particular data set?

This stuff is not just difficult to analyze, it's also really difficult to measure that accurately.

Regards,

Nothing wrong with Ivan's data set, note how it results in a similar response to my 2003 FLH Chorn tests shown in post #1322 in the upper response.
Upper response becomes flat as a pancake, because the upper humps are attenuated, likely because of interference patterns.
However, the FLH low corner obviously goes down in multiples, but the TH obviously does not.

As I said in that post, and will repeat, the difference in array configuration will account for some differences between measured and Hornresp.

Hornresp can't be expected to model array shapes.

[Djim]

Oliver.... or do you think I'm just trying to safe something of my 'blood and tears' time investment ? (I can deal with your truth )

Edit: Yes there is Art and it's more than obvious... At 116dB it looks like a brick limiter not like an stack of 4...

[David McBean]

Hi Everyone,

Many thanks for all your recent inputs. While there are some understandable differences in the sets of data that have been provided, on balance the measured results would appear to be closer to the predictions made by the proposed new Hornresp multiple speakers model, rather than the existing model.

To summarise the differences between the two models:

Two speakers connected in parallel compared to single speaker - same voltage:

Existing model:
+6dB at low frequencies and +3dB at high frequencies (see Old_2P.png).
Proposed model:
+6dB at low frequencies and +6dB at high frequencies (see New_2P.png).

Two speakers in connected in series compared to single speaker - same voltage:

Existing model:
+0dB at low frequencies and -3dB at high frequencies (see Old_2S.png).
Proposed model:
+0dB at low frequencies and +0dB at high frequencies (see New_2S.png).

The results generated by the proposed new model are consistent with AkAbak predictions and the standard SPL summing rules for coherent / correlated signals. The new model does not however, allow for bass extension due to mutual coupling of multiple horn mouths - it seems from measurements that this is not really an issue anyway, in that the real-world coupling between adjacent speakers is not as tight as perhaps first thought.

Please note that these are the only two model options under consideration - I not want to deviate from the standard models in an attempt to better fit measured data. Also, I don’t want to include both options in Hornresp - it could be confusing for some users, and make comparison of results more difficult.

If anyone believes that it makes more sense to retain the old model rather than change to the proposed new model, could they please let me know as soon as possible, either by way of this thread or by private e-mail to the address given on the Hornresp web page. Thanks.

Kind regards,

David

[NEO Dan]

I don't see how one can invalidate the coupling calculations without also claiming there would be a problem with the validity of radiation angle calculations?

Seems like a lack of understanding problem to me WRT what is acoustically sizable/correct. IMO it is entirely possible to build an array of acoustically small devices and create an acoustically large result. Having the ability to predict the change in driver loading is important to me.

Thanks for all the hard work David.