A friend recently lent me a Dayton DATS LA to test. Dayton advertises that the DATS LA is capable of performing small- and large-signal analyses. This review focuses exclusively on large-signal analysis, or rather, the part of the program known as " Symmetry Test." Further details and the associated patent can be found on the Dayton website (https://www.daytonaudio.com/product/2090/dats-la-loudspeaker-analyzer). The patent—probably not entirely unintentionally—suggests a similarity to the Klippel analyzer. However, that would set the bar quite high.
Before we get into the measurement results, a little more information about the Dayton DATS LA. I believe the measurement principle was already introduced in 1992 by Blind, Phillips, and Geddes at the 93rd AES Convention, although they used a different measurement principle (DUMAX) in their publication.
Fig. 1
The patent specification, published on the Dayton Audio website, also refers to this source. According to the patent specification, the loudspeaker is to be excited with a very low-frequency alternating voltage and a superimposed 0.7-second sweep.
Fig. 2
The real measurement signal or a measurement run looks like this.
Fig. 3
Fig. 4
The operating point is set using a modified "square-wave signal" of approximately 0.33 Hz, and the impedance at the operating point is then determined using a 0.7-second sweep. Obviously, the measurement is not performed alternating between cone in and cone out as shown in Fig. 6, but rather continuously decreasing in 19 predefined steps, resulting in 19 impedance curves , as shown in the following figure.
Fig. 5
It is inevitable that the voice coil will heat up when exposed to direct current. DATS LA attempts to reduce this undesirable effect during measurements by varying pause or cooling times. For example, a measurement cycle with ± 18V DC takes approximately twice as long as with ± 9V DC due to the longer pause times.
Fig. 6: Wavecor WF182BD10 (9.0V DC)
Fig. 7: Faital 12PR320 (18.0V DC)
The following shows the effect of applying DC voltage to the voice coil on Re and the voice coil temperature. The measurement cycle is clearly visible in the progression.
Fig. 8: Wavecor WF182BD10
Fig. 9: Faital 12PR320
The picture of a measurement run also shows that the cone does not return to the zero position after each application of DC voltage, but rather an offset slowly builds up. Here's an evaluation example for the Faital 12PR320:
Fig. 10
Since only electrical measurements can be performed with the DATS LA, the cone excursion X-DC (red marked) must be calculated from the following data set:
Tab. 1
In the above-mentioned frequency range of 0.33 Hz, the following should apply in the linear range to a good approximation (Beranek & Mellow , Acoustics, page 286):
X(U) = BL(U) * U/Re(U) * Cms ( 0)
However, I still haven't figured out how Dayton DATS LA calculates the excursion in the nonlinear range. Unfortunately, my attempts at correction are only applicable to specific cases.
Below is a first measured example (Visaton AL130). The blue curve shows the DATS LA result for the excursion, while the red dashed curve shows the result calculated using the formula above. In the linear range, it fits quite well, as expected...
Fig. 11
Now let's do it all again with the excursion measured using a triangulation laser (dotted black line). The DATS result isn't quite identical, but it's pretty close.
Fig. 12
Is this the accuracy the system can offer in terms of displacement estimation? More on that later.
What impact do these inaccuracies in the excursion determination have on the parameter analysis? The following analysis is again from the Visaton AL130 (blue = DATS, red = laser-corrected). The linear excursion X(BL) is ± 6.1 mm for the DATS LA and ± 5.4 mm for the laser-corrected version.
Fig. 13
Since a full set of Klippel data is available for the AL130 used here, a direct comparison of the BL(x) analysis between the different measurement methods is a good option.
The following image shows the results of three - or four - measurement methods (red = Klippel, blue long dashed = DUMAX, green short dashed = DATS LA, black dotted = DATS LA laser-corrected). At least for the AL130, the results lie within a fairly narrow scatter band despite the very different measurement methods.
Fig. 14
To verify these results, two more examples are now shown ( Faital 12PR320, Wavecor WF182BD10).
Fig. 15: Faital 12PR320
Fig. 16: Wavecor WF182BD10
Obviously, the quality of the DATS LA excursion estimate in the first example cannot be transferred to other loudspeakers. Especially with the 12PR320, but also with the WF182BD10, there are significant discrepancies between the excursion estimate and the measurement. To illustrate the effects of these differences, the corresponding BL(x) diagrams are shown below (blue line = DATS LA, red line = laser-corrected).
Fig. 17: 12PR320
Fig. 18: WF182BD10
DATS LA shows ± 4.80 mm for the linear excursion X(BL,12PR320) and ± 5.95 mm for the laser-corrected variant, or ± 6.75 mm for X( BL,WF 182BD10) and ± 5.90 mm for the laser-corrected variant.
These deviations in the excursion determination naturally have an impact not only on the BL(x) curves but also on the other results such as Kms (x), Le(x), etc.
To further support the previous results, the X(BL) evaluation for all 22 measured loudspeakers is shown below:
Tab. 2
The difference between the DATS estimate and the laser-corrected measurement averages 16%, with a single standard deviation of ± 20%. This is probably not quite up to the quality standards of the Klippel Analyzer.
How do we evaluate these results?
Regardless of how relevant one considers the deviations to be, if one never knows how close the excursion calculated by DATS LA is to the measured reality without additional laser measuring equipment, there will always remain a feeling of uncertainty regarding the reliability of the measured data and this will inevitably lead to dissatisfaction in the long run.
Regards
Heinrich
Before we get into the measurement results, a little more information about the Dayton DATS LA. I believe the measurement principle was already introduced in 1992 by Blind, Phillips, and Geddes at the 93rd AES Convention, although they used a different measurement principle (DUMAX) in their publication.
The patent specification, published on the Dayton Audio website, also refers to this source. According to the patent specification, the loudspeaker is to be excited with a very low-frequency alternating voltage and a superimposed 0.7-second sweep.
The real measurement signal or a measurement run looks like this.
The operating point is set using a modified "square-wave signal" of approximately 0.33 Hz, and the impedance at the operating point is then determined using a 0.7-second sweep. Obviously, the measurement is not performed alternating between cone in and cone out as shown in Fig. 6, but rather continuously decreasing in 19 predefined steps, resulting in 19 impedance curves , as shown in the following figure.
It is inevitable that the voice coil will heat up when exposed to direct current. DATS LA attempts to reduce this undesirable effect during measurements by varying pause or cooling times. For example, a measurement cycle with ± 18V DC takes approximately twice as long as with ± 9V DC due to the longer pause times.
The following shows the effect of applying DC voltage to the voice coil on Re and the voice coil temperature. The measurement cycle is clearly visible in the progression.
The picture of a measurement run also shows that the cone does not return to the zero position after each application of DC voltage, but rather an offset slowly builds up. Here's an evaluation example for the Faital 12PR320:
Since only electrical measurements can be performed with the DATS LA, the cone excursion X-DC (red marked) must be calculated from the following data set:
Tab. 1
In the above-mentioned frequency range of 0.33 Hz, the following should apply in the linear range to a good approximation (Beranek & Mellow , Acoustics, page 286):
X(U) = BL(U) * U/Re(U) * Cms ( 0)
However, I still haven't figured out how Dayton DATS LA calculates the excursion in the nonlinear range. Unfortunately, my attempts at correction are only applicable to specific cases.
Below is a first measured example (Visaton AL130). The blue curve shows the DATS LA result for the excursion, while the red dashed curve shows the result calculated using the formula above. In the linear range, it fits quite well, as expected...
Now let's do it all again with the excursion measured using a triangulation laser (dotted black line). The DATS result isn't quite identical, but it's pretty close.
Is this the accuracy the system can offer in terms of displacement estimation? More on that later.
What impact do these inaccuracies in the excursion determination have on the parameter analysis? The following analysis is again from the Visaton AL130 (blue = DATS, red = laser-corrected). The linear excursion X(BL) is ± 6.1 mm for the DATS LA and ± 5.4 mm for the laser-corrected version.
Since a full set of Klippel data is available for the AL130 used here, a direct comparison of the BL(x) analysis between the different measurement methods is a good option.
The following image shows the results of three - or four - measurement methods (red = Klippel, blue long dashed = DUMAX, green short dashed = DATS LA, black dotted = DATS LA laser-corrected). At least for the AL130, the results lie within a fairly narrow scatter band despite the very different measurement methods.
To verify these results, two more examples are now shown ( Faital 12PR320, Wavecor WF182BD10).
Obviously, the quality of the DATS LA excursion estimate in the first example cannot be transferred to other loudspeakers. Especially with the 12PR320, but also with the WF182BD10, there are significant discrepancies between the excursion estimate and the measurement. To illustrate the effects of these differences, the corresponding BL(x) diagrams are shown below (blue line = DATS LA, red line = laser-corrected).
DATS LA shows ± 4.80 mm for the linear excursion X(BL,12PR320) and ± 5.95 mm for the laser-corrected variant, or ± 6.75 mm for X( BL,WF 182BD10) and ± 5.90 mm for the laser-corrected variant.
These deviations in the excursion determination naturally have an impact not only on the BL(x) curves but also on the other results such as Kms (x), Le(x), etc.
To further support the previous results, the X(BL) evaluation for all 22 measured loudspeakers is shown below:
Tab. 2
The difference between the DATS estimate and the laser-corrected measurement averages 16%, with a single standard deviation of ± 20%. This is probably not quite up to the quality standards of the Klippel Analyzer.
How do we evaluate these results?
Regardless of how relevant one considers the deviations to be, if one never knows how close the excursion calculated by DATS LA is to the measured reality without additional laser measuring equipment, there will always remain a feeling of uncertainty regarding the reliability of the measured data and this will inevitably lead to dissatisfaction in the long run.
Regards
Heinrich
Last edited:
Heinrich (ente),
Thank you for reviewing the DATS LA system I created for Dayton Audio. It is apparent that you have spent considerable effort to collect data and perform comparisons to the highly acclaimed DUMAX and Klippel measurement systems. In your introduction you stated: "The patent—probably not entirely unintentionally—suggests a similarity to the Klippel analyzer. However, that would set the bar quite high." Yes, our goal was to create a system based on impedance measurement that could compete with loudspeaker measurement legends DUMAX and Klippel while remaining both portable and affordable.
To provide readers with a broader perspective on these products here are images of each system:
First is the DUMAX system created by David Clark of DLC Design circa 1991:
This is a physically large system employing pneumatic pressure to displace the cone for measurements.
Next we have the Klippel system (circa 1997 created by Prof. Dr. Wolfgang Klippel), or a portion of the system (the rack mounted gear is not shown):
Like the DUMAX system the Klippel system is large and bulky. In addition, the Klippel system consists of a number of separate elements that are purchased together as a system.
In contrast to the DUMAX and Klippel systems the DATS LA system from Dayton Audio is compact and inexpensive:
DATS LA is portable and works via USB interface to your PC.
As described in the review, DATS LA collects data by performing a number of impedance measurements. After measurements are complete the data is used to create plots of various speaker parameters as a function of X (cone displacement). The displacement itself is given as a function of DC voltage.
Heinrick gave us Beranek's equation: X(U) = BL(U) * U/Re(U) * Cms ( 0)
Heinrich wrote: "… I still haven't figured out how Dayton DATS LA calculates the excursion in the nonlinear range."
Well, I began by calculating X much as you show, by applying Hooke's Law to force and compliance to get a linear displacement. And then introducing the dependence on the applied voltage U. (I'm using V here)
Starting with Hooke's Law:
F = K*X where: F = force; K = Kms = spring constant ; X = displacement, we solve for X and get:
X = F/Kms substituting loudspeaker compliance for K (Cms = 1/Kms) we get:
X = F*Cms next we note that: F = BL * I, where I is the current through the voice coil, and substitute for F
X = BL * I * Cms finally we can substitute: I = V/Re (Ohms law) and add dependence on V to get Beranek's equation:
X = BL(V) * V/Re(V) * Cms(V) where V is the voltage across the voice coil. NON-LINEAR!
Before the last line X it was easy to calculate X as the terms on the right side were all constant. The last line is where the complexity of nature emerges as we note that each of the parameters on the right side is changing with the applied voltage. The solution for X is no longer trivial. I pursued solutions to the non-linear equation for quite some time because it could enable measurements of cone excursion from a series of complex impedance measurements made at a series of cone displacements. Yes, "complex" because impedance is a complex number having both Real and Imaginary parts. As I worked toward a solution the math swamp got deep fast. But after several failed attempts I finally emerged with a precise solution for calculating X from a series of Z (impedance) measurements. The details of the complex non-linear algorithm I developed to calculate excursion (no estimates!) remain proprietary at this time. This is the method DATS LA uses to determine speaker cone displacement.
As Heinrick showed in the displacement measurement of the Visaton AL130 reproduced below, DATS LA provides a very close match to the laser measured displacement . Had he aligned the two curves to both have zero offset at 0 VDC the match would have appeared to be even better. The only discrepancy seen is at the far left where the driver is approaching its suspension limit and parameters can vary widely. I note that the AL130 has an Xmax (82% BL) of 6 mm based on the BL plot given in the review. Ignoring the cone displacement beyond -6 mm (Xmax) the small difference between the reviewer's laser measurement and the DATS LA calculated displacement is reduced further.
The rather large discrepancy seen in your next comparison with the Faital 12PR320 appears to be the result of a gain difference as indicated by the different slopes at zero offset in the center of the linear region. I also note that the zero offsets are again not fully aligned with the laser data indicating a small negative offset at zero volts. So this measurement is bogus and should be ignored.
The plot (below) showing all 3 systems measurements for the Visaton AL130 again shows DATS LA in excellent agreement with DUMAX and Klippel. Starting at the left side we see DATS LA showing BL outside the range between the DUMAX and Klippel responses. For the rest of the plot the DATS LA data splits the difference between the DUMAX and Klippel data (DATS LA is the green dashed plot). I consider this to be a nice validation of my excursion algorithm. Note that at the Xmax limit of -6mm DATS LA's BL is closer to the Klippel measurement than the DUMAX measurement.
The DATS measurement (green dashed line) rises above the Klippel curve at the far left because DATS LA provides higher resolution than either DUMAX or Klippel systems and shows details that are missed by the other systems.
The detail seen in the DATS LA BL measurement (above) would not be resolved by the other systems. Curve fitting sparse data results in the ultra-smooth lines plotted by the DUMAX and Klippel systems. By preserving this detail DATS provides the user with a higher level of information about the driver.
Now, regarding the error analysis. Based on the measured data on multiple drivers that was tabulated you concluded that DATS LA had an average deviation of 16% from the laser data. This data included measurements on 2 drivers that appear to be "outliers"
and ideally would be excluded from the error analysis. My own spreadsheet (recreated from the data in the review) reveals an average deviation of 15.18% when the outlier points are included but falls to 12.64% if the two outlier measurements are excluded.
Here is my recreated spreadsheet with the corrected averages:
Questions and Omissions:
Why did the reviewer choose to increase his 15.18% average up to 16%?
Why did he include outlier data in the error analysis? Other analysts would have tossed this data out.
Why did he not see that the differences seen between DATS and Klippel data are smaller than differences between DUMAS and Klippel?
Why was there no mention of the preceding generations of DATS which has become the world-wide standard for small-signal loudspeaker parameter measurement?
There was no mention of the essentially perfect match between DATS, Klippel and DUMAS through the critical central region of the BL plots.
Not reported is the reliability with which DATS LA reveals asymmetry in the BL curve as well as the other parameters.
There was no mention of the portability or ease of use of the DATS LA system compared to the others.
There was no mention of the enormous cost difference between the systems. (DATS LA < $500, Klippel > 10's of k$).
On the other hand, the review over-emphasized the significance of the differences in the end range (even beyond Xmax!) while ignoring the overall excellent performance of DATS LA.
Future development of DATS LA:
I am always looking for ways to improve the measurements of DATS V3 and DATS LA and when I can verify reported bugs or errors I take action to improve the system. We do have plans for a Pro system with a much higher power output capable of driving the largest woofers to their mechanical limits. Meanwhile DATS LA provides an excellent way to measure the large-signal performance of our speakers up to the 100 Watt level.
I appreciate the effort Heinrich expended to collect data with the DATS LA system, but based on the questions and omissions I listed above, "there will always remain a feeling of uncertainty" concerning the validity and integrity of this anonymous review.
Best regards,
John
John L. Murphy
True Audio®
Thank you for reviewing the DATS LA system I created for Dayton Audio. It is apparent that you have spent considerable effort to collect data and perform comparisons to the highly acclaimed DUMAX and Klippel measurement systems. In your introduction you stated: "The patent—probably not entirely unintentionally—suggests a similarity to the Klippel analyzer. However, that would set the bar quite high." Yes, our goal was to create a system based on impedance measurement that could compete with loudspeaker measurement legends DUMAX and Klippel while remaining both portable and affordable.
To provide readers with a broader perspective on these products here are images of each system:
First is the DUMAX system created by David Clark of DLC Design circa 1991:
Next we have the Klippel system (circa 1997 created by Prof. Dr. Wolfgang Klippel), or a portion of the system (the rack mounted gear is not shown):
In contrast to the DUMAX and Klippel systems the DATS LA system from Dayton Audio is compact and inexpensive:
As described in the review, DATS LA collects data by performing a number of impedance measurements. After measurements are complete the data is used to create plots of various speaker parameters as a function of X (cone displacement). The displacement itself is given as a function of DC voltage.
Heinrick gave us Beranek's equation: X(U) = BL(U) * U/Re(U) * Cms ( 0)
Heinrich wrote: "… I still haven't figured out how Dayton DATS LA calculates the excursion in the nonlinear range."
Well, I began by calculating X much as you show, by applying Hooke's Law to force and compliance to get a linear displacement. And then introducing the dependence on the applied voltage U. (I'm using V here)
Starting with Hooke's Law:
F = K*X where: F = force; K = Kms = spring constant ; X = displacement, we solve for X and get:
X = F/Kms substituting loudspeaker compliance for K (Cms = 1/Kms) we get:
X = F*Cms next we note that: F = BL * I, where I is the current through the voice coil, and substitute for F
X = BL * I * Cms finally we can substitute: I = V/Re (Ohms law) and add dependence on V to get Beranek's equation:
X = BL(V) * V/Re(V) * Cms(V) where V is the voltage across the voice coil. NON-LINEAR!
Before the last line X it was easy to calculate X as the terms on the right side were all constant. The last line is where the complexity of nature emerges as we note that each of the parameters on the right side is changing with the applied voltage. The solution for X is no longer trivial. I pursued solutions to the non-linear equation for quite some time because it could enable measurements of cone excursion from a series of complex impedance measurements made at a series of cone displacements. Yes, "complex" because impedance is a complex number having both Real and Imaginary parts. As I worked toward a solution the math swamp got deep fast. But after several failed attempts I finally emerged with a precise solution for calculating X from a series of Z (impedance) measurements. The details of the complex non-linear algorithm I developed to calculate excursion (no estimates!) remain proprietary at this time. This is the method DATS LA uses to determine speaker cone displacement.
As Heinrick showed in the displacement measurement of the Visaton AL130 reproduced below, DATS LA provides a very close match to the laser measured displacement . Had he aligned the two curves to both have zero offset at 0 VDC the match would have appeared to be even better. The only discrepancy seen is at the far left where the driver is approaching its suspension limit and parameters can vary widely. I note that the AL130 has an Xmax (82% BL) of 6 mm based on the BL plot given in the review. Ignoring the cone displacement beyond -6 mm (Xmax) the small difference between the reviewer's laser measurement and the DATS LA calculated displacement is reduced further.
The plot (below) showing all 3 systems measurements for the Visaton AL130 again shows DATS LA in excellent agreement with DUMAX and Klippel. Starting at the left side we see DATS LA showing BL outside the range between the DUMAX and Klippel responses. For the rest of the plot the DATS LA data splits the difference between the DUMAX and Klippel data (DATS LA is the green dashed plot). I consider this to be a nice validation of my excursion algorithm. Note that at the Xmax limit of -6mm DATS LA's BL is closer to the Klippel measurement than the DUMAX measurement.
The DATS measurement (green dashed line) rises above the Klippel curve at the far left because DATS LA provides higher resolution than either DUMAX or Klippel systems and shows details that are missed by the other systems.
The detail seen in the DATS LA BL measurement (above) would not be resolved by the other systems. Curve fitting sparse data results in the ultra-smooth lines plotted by the DUMAX and Klippel systems. By preserving this detail DATS provides the user with a higher level of information about the driver.
Now, regarding the error analysis. Based on the measured data on multiple drivers that was tabulated you concluded that DATS LA had an average deviation of 16% from the laser data. This data included measurements on 2 drivers that appear to be "outliers"
and ideally would be excluded from the error analysis. My own spreadsheet (recreated from the data in the review) reveals an average deviation of 15.18% when the outlier points are included but falls to 12.64% if the two outlier measurements are excluded.
Here is my recreated spreadsheet with the corrected averages:
Questions and Omissions:
Why did the reviewer choose to increase his 15.18% average up to 16%?
Why did he include outlier data in the error analysis? Other analysts would have tossed this data out.
Why did he not see that the differences seen between DATS and Klippel data are smaller than differences between DUMAS and Klippel?
Why was there no mention of the preceding generations of DATS which has become the world-wide standard for small-signal loudspeaker parameter measurement?
There was no mention of the essentially perfect match between DATS, Klippel and DUMAS through the critical central region of the BL plots.
Not reported is the reliability with which DATS LA reveals asymmetry in the BL curve as well as the other parameters.
There was no mention of the portability or ease of use of the DATS LA system compared to the others.
There was no mention of the enormous cost difference between the systems. (DATS LA < $500, Klippel > 10's of k$).
On the other hand, the review over-emphasized the significance of the differences in the end range (even beyond Xmax!) while ignoring the overall excellent performance of DATS LA.
Future development of DATS LA:
I am always looking for ways to improve the measurements of DATS V3 and DATS LA and when I can verify reported bugs or errors I take action to improve the system. We do have plans for a Pro system with a much higher power output capable of driving the largest woofers to their mechanical limits. Meanwhile DATS LA provides an excellent way to measure the large-signal performance of our speakers up to the 100 Watt level.
I appreciate the effort Heinrich expended to collect data with the DATS LA system, but based on the questions and omissions I listed above, "there will always remain a feeling of uncertainty" concerning the validity and integrity of this anonymous review.
Best regards,
John
John L. Murphy
True Audio®
John,
Thank you for your response to my review of DATS LA. For the sake of clarity, I would like to emphasize that I only looked at one part of DATS LA's range of features, namely the " Symmetry Test." Furthermore, I have to agree with you that a direct comparison with the Klippel Analyzer or DUMAX is unfair, especially since both systems are in a different league in terms of price. This is precisely why your approach of creating something affordable with a similar range of features made me curious to explore DATS LA.
To begin with, perhaps a comment on your concluding remark: I am simply a privately interested loudspeaker builder with no commercial interests and am not affiliated with any commercial provider. Therefore, I don't understand why my review calls my integrity into question. If there are any doubts about this statement, I'm happy to answer any questions here.
Now to your comments:
I'm aware that David Clark developed DUMAX. The AES publications [1] and [2] are mentioned only because both principles (electrical and pneumatic membrane displacement) are already described and evaluated there in the context of nonlinear parameter determination. Figure 1 of the "Experimental Setup" is taken from [2].
The two principles are also normatively defined in "IEC 62458:2010 - Sound system equipment - Electroacoustical transducers - Measurement of large signals parameters ". It says:
“IEC 62458: 4.2 DC large signal
A constant DC voltage or DC current of a specified magnitude and sufficient duration is applied to the electrical terminals to measure the steady-state response curve of the transducer. If the transducer is installed in a sealed enclosure, a difference between the air pressure inside and outside the enclosure may be used as a static excitation.”
“IEC 62458: 4.3 DC large-signal and AC small-signal
A constant DC signal of specified magnitude and sufficient duration (see 4.2), linearly superimposed with a small AC signal, is used as the excitation. The AC signal (e.g., noise, sinusoidal sweep signals, impulse test signals) should have sufficient bandwidth to detect all parameters of the loudspeaker model.”
That calculating the excursion from the electrically determined data is not trivial became clear to me when I was dealing with DUMAX. As stated in [1]: "The determination of the nonlinear parameters will require a means for varying the dependent variable, primarily the cone displacement , and measuring the parameters under this modified condition . There are several ways to displace the cone , electrically and pneumatically . The electrical method will cause problems since , as stated before , the electrical current is one of the variables which we expect the parameters to vary . Thus if current is used to displace the cone then the measure of the BL product will be complicated by the interaction of two effects , its variation with both displacement and current ."
In addition, I believe that time-dependent creep effects in the membrane suspension are difficult to capture using only electrical measurements. It's probably not without reason that both DUMAX and Klippel have integrated triangulation lasers for excursion measurement.
I have tried to demonstrate that these effects play a role in my 12PR320 example (Figure 10). The DATS LA measurement cycle begins with the highest selected negative excursion (Figure 7). The pause times are obviously not long enough to return the membrane to its zero position before the next measurement sweep starts. In this respect, it is not surprising that at 0 volts DC – which is in the middle of the DATS LA measurement cycle – there is not 0 mm excursion, but rather an offset. This can be clearly seen in Figures 4, 6 and 7. Furthermore, it is unavoidable that heating up occurs when the voice coil is subjected to DC voltage, which is also clearly visible in Figures 8 and 9.
I deliberately did not compensate for the aforementioned offset in the evaluation, as I would not have known about it without accompanying displacement measurements. I also deliberately did not exclude any "outliers," as this is a direct comparison of the DATS LA calculation with the laser measurement. I would like to emphasize again that the laser was running simultaneously with the DATS LA measurement, thus the signal was identical but the results were different. And since I still believe that the complex mechanisms of a loudspeaker, including creepage effects, are difficult to derive mathematically from electrical measurement data alone, I still trust my laser more than any model calculation.
At this point, a note regarding the difference in the average deviation. This wasn't malicious, but rather due to the fact that I added a good measurement result and forgot to correct it. I also do not want to comment on the conclusions you drew from Figure 14 regarding the resolution of DATS LA; owners/users of Klippel analyzers are certainly more competent in this respect than I am.
Of the 22 loudspeakers I measured (Table 2), there were several examples with comparably good agreement, such as the AL130 (Figure 13), but unfortunately, there were also several that were similar to the Faital 12PR320 (Figures 15 to 18). The loudspeakers I measured cover a wide range, from vintage loudspeakers (SPH130, GDN13-40) to current hi-fi drivers (Scan Speak, SEAS, Wavecor, SB) to modern PA loudspeakers (Faital). Despite all my efforts, I have not been able to discover any systematic explanation as to the possible reasons for agreement or discrepancies between calculation and measurement (Good agreement: SLS65S25, 6RS140, 10RS350, AL130, L16RN-SL, SW223BD03, W170S4, SB15NBAC30-8, 18W8434G00; Bad agreement: FE103, SPH130, SB12NRX25-8, SB13PFCR25-4, 12PR320, 15PR400, WF182BD10, SPH100C, GDN13-40, 15W8434G00, 15W8531K00, 22W4534G00, WF120BD03).
And that is precisely the reason why I came to my final evaluation result.
Regards
Heinrich
[1] Efficient Loudspeaker linear and nonlinear parameter estimation by Earl Geddes and Alan Phillips, AES 1991
[2] Efficient Loudspeaker Parameter Estimation -An Extension by Henry Blind, Alan Phillips, and Earl Geddes, AES 1992
Thank you for your response to my review of DATS LA. For the sake of clarity, I would like to emphasize that I only looked at one part of DATS LA's range of features, namely the " Symmetry Test." Furthermore, I have to agree with you that a direct comparison with the Klippel Analyzer or DUMAX is unfair, especially since both systems are in a different league in terms of price. This is precisely why your approach of creating something affordable with a similar range of features made me curious to explore DATS LA.
To begin with, perhaps a comment on your concluding remark: I am simply a privately interested loudspeaker builder with no commercial interests and am not affiliated with any commercial provider. Therefore, I don't understand why my review calls my integrity into question. If there are any doubts about this statement, I'm happy to answer any questions here.
Now to your comments:
I'm aware that David Clark developed DUMAX. The AES publications [1] and [2] are mentioned only because both principles (electrical and pneumatic membrane displacement) are already described and evaluated there in the context of nonlinear parameter determination. Figure 1 of the "Experimental Setup" is taken from [2].
The two principles are also normatively defined in "IEC 62458:2010 - Sound system equipment - Electroacoustical transducers - Measurement of large signals parameters ". It says:
“IEC 62458: 4.2 DC large signal
A constant DC voltage or DC current of a specified magnitude and sufficient duration is applied to the electrical terminals to measure the steady-state response curve of the transducer. If the transducer is installed in a sealed enclosure, a difference between the air pressure inside and outside the enclosure may be used as a static excitation.”
“IEC 62458: 4.3 DC large-signal and AC small-signal
A constant DC signal of specified magnitude and sufficient duration (see 4.2), linearly superimposed with a small AC signal, is used as the excitation. The AC signal (e.g., noise, sinusoidal sweep signals, impulse test signals) should have sufficient bandwidth to detect all parameters of the loudspeaker model.”
That calculating the excursion from the electrically determined data is not trivial became clear to me when I was dealing with DUMAX. As stated in [1]: "The determination of the nonlinear parameters will require a means for varying the dependent variable, primarily the cone displacement , and measuring the parameters under this modified condition . There are several ways to displace the cone , electrically and pneumatically . The electrical method will cause problems since , as stated before , the electrical current is one of the variables which we expect the parameters to vary . Thus if current is used to displace the cone then the measure of the BL product will be complicated by the interaction of two effects , its variation with both displacement and current ."
In addition, I believe that time-dependent creep effects in the membrane suspension are difficult to capture using only electrical measurements. It's probably not without reason that both DUMAX and Klippel have integrated triangulation lasers for excursion measurement.
I have tried to demonstrate that these effects play a role in my 12PR320 example (Figure 10). The DATS LA measurement cycle begins with the highest selected negative excursion (Figure 7). The pause times are obviously not long enough to return the membrane to its zero position before the next measurement sweep starts. In this respect, it is not surprising that at 0 volts DC – which is in the middle of the DATS LA measurement cycle – there is not 0 mm excursion, but rather an offset. This can be clearly seen in Figures 4, 6 and 7. Furthermore, it is unavoidable that heating up occurs when the voice coil is subjected to DC voltage, which is also clearly visible in Figures 8 and 9.
I deliberately did not compensate for the aforementioned offset in the evaluation, as I would not have known about it without accompanying displacement measurements. I also deliberately did not exclude any "outliers," as this is a direct comparison of the DATS LA calculation with the laser measurement. I would like to emphasize again that the laser was running simultaneously with the DATS LA measurement, thus the signal was identical but the results were different. And since I still believe that the complex mechanisms of a loudspeaker, including creepage effects, are difficult to derive mathematically from electrical measurement data alone, I still trust my laser more than any model calculation.
At this point, a note regarding the difference in the average deviation. This wasn't malicious, but rather due to the fact that I added a good measurement result and forgot to correct it. I also do not want to comment on the conclusions you drew from Figure 14 regarding the resolution of DATS LA; owners/users of Klippel analyzers are certainly more competent in this respect than I am.
Of the 22 loudspeakers I measured (Table 2), there were several examples with comparably good agreement, such as the AL130 (Figure 13), but unfortunately, there were also several that were similar to the Faital 12PR320 (Figures 15 to 18). The loudspeakers I measured cover a wide range, from vintage loudspeakers (SPH130, GDN13-40) to current hi-fi drivers (Scan Speak, SEAS, Wavecor, SB) to modern PA loudspeakers (Faital). Despite all my efforts, I have not been able to discover any systematic explanation as to the possible reasons for agreement or discrepancies between calculation and measurement (Good agreement: SLS65S25, 6RS140, 10RS350, AL130, L16RN-SL, SW223BD03, W170S4, SB15NBAC30-8, 18W8434G00; Bad agreement: FE103, SPH130, SB12NRX25-8, SB13PFCR25-4, 12PR320, 15PR400, WF182BD10, SPH100C, GDN13-40, 15W8434G00, 15W8531K00, 22W4534G00, WF120BD03).
And that is precisely the reason why I came to my final evaluation result.
Regards
Heinrich
[1] Efficient Loudspeaker linear and nonlinear parameter estimation by Earl Geddes and Alan Phillips, AES 1991
[2] Efficient Loudspeaker Parameter Estimation -An Extension by Henry Blind, Alan Phillips, and Earl Geddes, AES 1992
Last edited:
Why are you omitting the bigger picture of the review and nitpick small things that don't and won't change the end results?
Arguing about 15% vs 16% is totally missing the point obviously, it most definetely doesn't change the order of magnitude we're talking about
The same goes for 13% vs 15%.
Btw, in error analysis we always round UP to the highest digit, never down!
Cost differences here are entirely invalid, since both systems serve entirely different markets.
As a professional you should know that there is extremely little correlation about the technology and price something is sold for.
In that case, why don't you share some other 3rd party results with a Klippel vs DATS LA system comparison?
Because if we want to talk about certainly, @ente provides us wit hat least SOME data.
Which is an awful lot more significant than no data at all.
Based on my own 20 years of experience of experimental physics, with quite some knowledge in uncertainty and error analysis, it's kinda hard to beat a direct excursion measurement with a laser.
Well, there is, it's called a velocity sensor, and I only know of one company in the Netherlands that have products to do this.
Some very good arguments have been giving above, which on an higher level are very valid and sound.
Besides, I assume that if you did all the very long partial differential equations already, anyone with some math background would come to the same conclusion.
Since Dayton doesn't have any proven record of being a super highly respected and almost de facto standard measurement company like Klippel is, it's kinda hard to argue about Klippel's results.
That doesn't mean they don't make mistakes or have other quirks.
Some of those have been already raised by recent researchers like Novak.
But again, at that point we're nitpicking results.
Last edited:
My calculations are as following.
The average is 15.2±11.6% standard deviation
Which we normally would round up and write as 16±12%. (68%)
16±23% standard deviation based on 2 sigma (95%)
16±35% standard deviation based on 3 sigma. (99.8%)
Following the normal distribution (don't see a reason to assume something else yet)
Without very compelling arguments, I don't see any reason to remove the "outliers".
But just for sake of completeness, the results are:
13±9% Standard deviation
13±17% for 2 sigma
13±30% for 3 sigma
They are in the same order of magnitude anyway.
So any further conclusion will be basically the same.
Obviously we can only base the results on the samples used in the experiment here.
That being said, 22 samples is already quite a substantial amount of samples!
What's most striking, is just how uneven and widespread the differences are, which is clearly reflected in the standard deviation.
Which is FAR more important (and alarming) than the actual average itself.
When looking at the samples and values, there doesn't seem a very obvious and clear correlation.
Since we would at least want SOME kind of sense of how this will result in the bigger picture ("even more samples"), we will use the 2 sigma/95% results*.
In that case the standard deviation is larger than the actual average value itself, which means that based on this particular experiment compared to a Klippel system, the end-results are not reliable.
They are not even valid as a whole because of the fact that the standard deviation is bigger than the average value itself.
*Which is already pretty easy on the conclusion, for most test equipment at least a minimum of 3 to 5 sigma is being used.
Last edited:
I work with people all over the world that I get to remotely measure their drivers. Many have the WT3 or the DATS system. To get the same measurement repeatedly is not a simple ask. Personally I use two systems. Smith & Larson Speaker Tester Pro, ( The most consistent measurements that I have ever done. Sad they quit producing them. ) and I'm learning to use the Speaker Bench method. ( https://speakerbench.com ) I use the QuantAsylum QA403 to do the measurements. For Impedance it is rock solid. The measurement results are consistent and with a few percentage points of the same each time. For the products from Parts Express I need to ask for as many measurements spaced 5 minutes apart as the client is willing to do and I average them. Many are all over the place.
Are you talking about variations between production samples or measurements?
I have seen some variation with Klippel, but never this much never in the range of 40%.
Especially not with BL(x), except for a offset, caused by the compliance.
In that case it's mostly variation in the environment.
Temperature obviously changes some things as well, so measurements should always be done around the same room temperature.
As well as having the same duration and level of the excitation = energy.
My bit of critique and feedback for @ente would be just to show a sort of base null measurement with a couple of drivers, remounted and measured again.
Just to rule that part out.
Hopefully it wasn't done this way
I don't understand why this picture has been used for marketing.
That is dead wrong to begin with.
@ b-force
Unfortunately, I can't take any more measurements with the DATS LA at this time, as I had to return the device to the owner. However, I can ask him if I can borrow it again.
You can read how I perform my parameter measurements here: https://artalabs.hr/AppNotes/LIMP-HB-D2.53.pdf (unfortunately, the English version is no longer up to date).
Chapter 3.1.3
My clamping for the measurement looks like this (approximately 15 kg of steel is housed in the base).
@ Mark
We conducted a round robin test with LIMP a few years ago. The boundary conditions were specified, and the equipment (sound card, amplifier, etc.) was freely selectable. Nevertheless, the results were very close.
Regards,
Heinrich
Note: Before anyone starts to interpret this again: No, I have never had any commercial affiliation with Ivo Mateljan. Not to mention that Ivo deregistered his business.
Unfortunately, I can't take any more measurements with the DATS LA at this time, as I had to return the device to the owner. However, I can ask him if I can borrow it again.
You can read how I perform my parameter measurements here: https://artalabs.hr/AppNotes/LIMP-HB-D2.53.pdf (unfortunately, the English version is no longer up to date).
My clamping for the measurement looks like this (approximately 15 kg of steel is housed in the base).
@ Mark
We conducted a round robin test with LIMP a few years ago. The boundary conditions were specified, and the equipment (sound card, amplifier, etc.) was freely selectable. Nevertheless, the results were very close.
Regards,
Heinrich
Note: Before anyone starts to interpret this again: No, I have never had any commercial affiliation with Ivo Mateljan. Not to mention that Ivo deregistered his business.
Last edited:
I hate those clamps. I first bought one about 35 years ago. I grew up in a Cabinet shop. You really can't generate much pressure with them. But, that is why I would say they are perfect for this kind of application.
That manual deserves to be translated. I will take a stab at doing it with DeepL Translate. My German is very rusty, but I can read enough of it. Perhaps if I manage this I will send it to you Heinrich and you can see if it is accurate.
I need to update my test stand. Maybe I will make a post about that on social media so I have something to share 😉
That manual deserves to be translated. I will take a stab at doing it with DeepL Translate. My German is very rusty, but I can read enough of it. Perhaps if I manage this I will send it to you Heinrich and you can see if it is accurate.
I need to update my test stand. Maybe I will make a post about that on social media so I have something to share 😉
Mark,
The English version is also available, as mentioned, just not in the latest revision: https://artalabs.hr/support.htm
Regards
Heinrich
The English version is also available, as mentioned, just not in the latest revision: https://artalabs.hr/support.htm
Regards
Heinrich
I have it! When it was first released. There is some great information available on Ivo's website.
As far as I remember, I wrote the first version in 2006. It was always based on Ivo's original manuals, supplemented by examples and practical information and tools.
Regards
Heinrich
Regards
Heinrich
@John L. Murphy Hello. This is a query regarding the older DATS V3.
Presently, the impedance scale has an "Upper Z Limit" of "10k Ohm".
Would it be possible to increase this to "200k Ohm", or even "500k Ohm"? The change would potentially involve adding some extra values to the list of values that are presently being used.
The reason for this change is that some operators are using DATS V3 to measure the impedance curves of Hi-Z guitar pickups.
I'm hopeful that my request can make it into the improvements pipeline for DATS V3 (and maybe DATS LA, if applicable). Cheers!
Last edited: