1 - 12 of 12 Posts

#### npdang

·
##### DIYMA Founder
Joined
·
5,213 Posts
Discussion Starter · ·
Let's look at a quick example.

First let's examine the BL (Force Factor) curve in the upper left. This represents the "motor strength" of the speaker, and is essentially the "force" that drives your speaker. The X axis represents the distance your speaker cone or voice coil has travelled in mm. The Y axis represents your motor strength.

This is perhaps the most important curve, since it's responsible for a majority of the distortion produced by a speaker. Ideally, we would want the speaker to produce an equal force in the positive and negative direction. The overall motor strength is not important, but it's linearity... basically how flat the curve is and how symmetrical it is about x = 0. If the curve is non-symmetrical or uneven, distortion will be produced.

Now, we can see that if the BL curve were highly uneven, that would cause quite a bit of distortion since the "force" pushing your speakers would be uneven. Another thing to note is that when BL drops, your motor strength drops, and pretty much it takes MORE power to push your speaker. And when your BL changes, your t/s parameters change as well. So this is a very important curve here.

Now looking at this speaker, we can see that it has pretty good symmetry. The left half of the graph looks like the right half. Also it has a very long and flat plateau around x = 0. However, we can see that the further you move the speaker out or in, the motor strength begins to drop. The point where the motor strength is 71% of the strength at x = 0 (the rest position of the speaker), is what we define as XMag. Beyond XMag, the speaker in theory will generate significant audible distortion.

Also, look at the CMS (mechanical compliance, aka how loose the suspension is) and the LE (inductance) curves. CMS is the inverse of the suspension stiffness, and LE is the self inductance of the driver. KMS is the actual suspension stiffness.

These curves are important as well, and follow the same principles as the BL curve. CMS linearity is generally speaking the second most important factor determining distortion performance, and after that LE. With LE, you must also take into account both frequency and absolute inductance in determining distortion performance, in addition to variance.

We define the limit for the suspension (how far the speaker can move mechanically without producing significant distortion) as XSUS. It is 75% of the CMS value of X = 0.

For this example, we can see that the CMS curve is not exactly centered at x = 0, or the rest position. It's forward biased, meaning that it will allow a little more throw in the forward direction than in the rearward direction.

Now look at the second chart with all the numbers. Find the value of BL and CMS at x = 0. We know that:

Xmag = 71% of rest value
Xsus = 75% of rest value

In this case 71% of rest BL = 2.8897
75% of rest CMS = 1.6575

Take those values and follow them across the graph until you hit the measurement curve, and look directly down to the X axis to find the actual displacement value.

For Xmag we have a range of -5.5mm to +5.5mm.
For Xsus we have a range of -2mm to +3mm.

To calculate Xmax, (the limit of the speaker taking into consideration both suspension and motor linearity)... we take the smallest of both values. In this case the Xmax would be 2mm.

Another useful tip is to look at the dashed line. It's basically a mirror of the original curve about the y axis. For a speaker with good symmetry (like this one), you should not be able to see the dashed line.

#### npdang

·
##### DIYMA Founder
Joined
·
5,213 Posts
Discussion Starter · ·
With some experience now, I'd like to post some comments.

I feel that Klippel measurements can't be directly compared to Dumax. The Klippel system measures non-linearities using large signal excitation, or real power. Dumax uses pnuematic air pressure to move the cone. You'll see temperature and suspension effects with Klippel that you wouldn't with Dumax.

Another issue with Klippel is that without a laser attachment, the x axis scale of displacement is dependent on the imported value of BL. This makes it quite critical to be sure that you're importing the correct value, especially given that some ways of measuring t/s parameters can give a somewhat wide margin of error. I believe this is why some of my tests including the Extremis and XXX 7" mids with the exact same motor have such different x-max figures, which shouldn't be the case.

Also, Klippel is not quite as good as Dumax at measuring driver's extremes. Since it uses "real power" to excite the driver, a driver with limited bl displacement as compared to it's suspension's throw would stop the measurement (due to protection limits) before the limits of the suspension could be measured. Same goes for a situation with a driver that has far more bl displacement than suspension throw. The max. settings for Klippel's protection limits is 25% of bl, and 20% of CMS. Dumax gives xsus at 25% of CMS, oftentimes a difficult figure for Klippel to measure.

#### foley316

·
##### Registered
Joined
·
232 Posts
Well why don't you just invent a testing setup which eliminates all variables? J/K, why not do both, take the most accurate tests of each and compare the results? BTW how much are the testing tools? I think you could make a section with both and charge like an initial 5-10 dollar charge for access. I would pay that even if I don't buy the drivers. I'm a graph junkie.

#### Neil

·
##### Registered
Joined
·
1,295 Posts
npdang said:
Also, Klippel is not quite as good as Dumax at measuring driver's extremes. Since it uses "real power" to excite the driver, a driver with limited bl displacement as compared to it's suspension's throw would stop the measurement (due to protection limits) before the limits of the suspension could be measured.
which is why you it is so important to get a LOT of samples at the + and - extremes of all drivers.

have you considered getting the laser?

#### Excelsior

·
##### Registered
Joined
·
31 Posts
Npdang,

you said "especially given that some ways of measuring t/s parameters can give a somewhat wide margin of error. I believe this is why some of my tests including the Extremis and XXX 7" mids with the exact same motor have such different x-max figures,"

The extremis and XXX 7" mids do not have the same motor at all. Even if the rebate and top plate are the same height, they do not share the same flux fields.

The larger 5" OD ferrite magnet on the XXX 7" should yield MORE excursion even if the rebate height, top plate thickness, and pole rebate height are identical. The reason for this is simple and speaker designers tend to use this often. The "fingers" of the top plate (the parts above and below the rebate) will have far wider flux fields due to approaching or actually being in magnetic saturation. This would extend the flux fields in a larger magnitude past the top plate height and it will act nearly like an extended pole overhung design above and below the top plate, past where the coil would normally drop off drastically in BL.

Also I am not sure that the XXX and Extremis share the same top plate thickness nor rebate heights.

I do not know all of how klippel works, maybe you can enlighten some questions. Will klippel read BL and account for it in the BL graph? If so the inputed BL specification does not matter, as Klippel seems to have no problems graphing ripples and other BL annomolies accurately, it would not show a flat BL graph if the inputed BL were too low or too high, there would be a dip/peak at rest correct?

I'm curious a bit more on how Klippel actually reads BL, it tends to show accurate Xmax figures and is often used as the industrial standard in driver testing. Dumax is definitely not considered the standard currently.

Your BL measurements tend to be spot on with the few selected companies' motors and coils I have simulated in a FEA program. I would not worry deeply if your tested BLs differ from manufacturer's specifications, especially when they tend to be much lower than spec'd . Companies tend to use the measuring rule with overhung speakers to get Xmax figures, but that is often not accurate and the actual parameters are much lower. That is the beauty of a site like your's.

#### npdang

·
##### DIYMA Founder
Joined
·
5,213 Posts
Discussion Starter · ·
The laser isn't all that cheap, and it was suggested to me that I should be importing BL in any case. It's been awhile so I can't recall the details exactly. I've also had the thing running for hours, with little or no change to the curves compared to a fast 15 minute test.

As far as the XXX/Extremis I was told that they have identical gap topologies... I'm sorry I didn't mean to say they had identical motors.

Unfortunately, I'm not a loudspeaker engineer, nor do I have a background in physics, math, or EE. I can only read the works of more knowledgeable people and interpret them as best I can.

That being said, I believe Klippel models loudspeakers as an electro-mechanical equivalent circuit, reads the current, voltage, temp, etc. at the terminals, and calculates the remaining parameters using some model. Unlike Dumax, it doesn't measure actual BL.

The input value of BL also doesn't affect the shape of the curve, just the X displacement scale in mm. Change it a bit, and suddenly you have more xmax, or less. If you recall the CA&E tests on the Lotus ref 6.5" driver, it was read as 10mm 1 way. Anyone with enough experience with the driver will realize that that figure is obviously not right, and probably due to too low of an imported BL throwing off the scale.

#### Excelsior

·
##### Registered
Joined
·
31 Posts
npdang said:
That being said, I believe Klippel models loudspeakers as an electro-mechanical equivalent circuit, reads the current, voltage, temp, etc. at the terminals, and calculates the remaining parameters using some model. Unlike Dumax, it doesn't measure actual BL.

The input value of BL also doesn't affect the shape of the curve, just the X displacement scale in mm. Change it a bit, and suddenly you have more xmax, or less. If you recall the CA&E tests on the Lotus ref 6.5" driver, it was read as 10mm 1 way. Anyone with enough experience with the driver will realize that that figure is obviously not right, and probably due to too low of an imported BL throwing off the scale.
Ah ok, they could possibly have identical top plates and rebate heights, I do not know.

Dumax tends to be a tad more accurate, of course, but at \$100 a test it is not worth it for most drivers (unless the companies themselves front the money, which they all should).

Generally the longer/multiple BL runs will give you more accurate data. What programs do you use to get T/S parameters for drivers?

If you would like to increase your resolution try using two programs like woofertest, praxis and/or speakerworkshop to calculate the T/S parameters (of course that's assuming you don't already do this).

With the right wiring rig (1% resistors and a nice 0.001 accurate ohm meter) I have seen T/S parameters be accurate to a thousandth of a decimal place. If you are getting the same T/S (to within a hundredth of a place) with two programs I would say that your BL graphs are incredibly accurate.

It would be interesting to see a Dumax test of a speaker you tested as a comparison tool

·
Joined
·
325 Posts
Thank you!

#### amungal

·
##### Registered
Joined
·
92 Posts
Well why don't you just invent a testing setup which eliminates all variables? J/K, why not do both, take the most accurate tests of each and compare the results? BTW how much are the testing tools? I think you could make a section with both and charge like an initial 5-10 dollar charge for access. I would pay that even if I don't buy the drivers. I'm a graph junkie.
I don't know the technical difficulty or the cost factors involved, but otherwise not really a bad idea.

#### 454Casull

·
##### Registered
Joined
·
28 Posts
If 75% of resting Cms(x) (i.e. Cms(0)) is taken as a displacement limit, does it matter how linear Cms(x) is then? The shape of Cms(x) or Kms(x) sometimes varies wildly depending on the driver but "75%" seems to indicate that it doesn't matter whether Kms(x) resembles a linear, parabolic, cubic, etc. function.

#### ErinH

·
##### Moderator
Joined
·
24,186 Posts
Draw a flat line along the x-axis and take 75% of that value in the y-axis. The point where the y axis value intersects the x-axis is how you get your Xmax for this case. Now do the same, but with a parabola. The difference is the flat line value extends further out along the x-axis (more Xmax), where the parabola crosses the x-axis closer in (less Xmax).

The same logic applies then to offsets and assymetry. If the line has a skew or tilt so that it is not symmetrical along the y-axis, this means the 75% value will cross the indented side sooner (unless it's a really odd shape).

This is easier with a photo but I can't access that right now. Hopefully that helps, though.

·