Diagnosing Design Issues Using the Klippel Distortion Analyzer
April 17 2024, 07:10
The founder and principal of Redrock Acoustics, renowned for offering engineering and design services, Patrick Turnmire has tested thousands of driver samples during his career, including countless for Voice Coil's Test Bench. Having recently reopened access to its Klippel testing and transducer analysis services, Turnmire shares his vast experience, advising on how to make sure the tests are robust enough to clearly see the loudspeaker performance.
I have tested thousands of driver samples during my career as a loudspeaker transducer engineer, for many brand names and factories around the world as well as Test Bench testing for Voice Coil magazine. The Klippel Distortion Analyzer (DA) has become the standard test instrument for analyzing all aspects of performance and optimizing designs. The Large Signal Identification (LSI) module is the primary tool for this with other modules helping to identify specific issues that are first seen in the LSI tests. Think of this like getting an overall health checkup followed by deeper tests that help identify the cause of underlying issues.
The key to getting good data is making sure the tests are robust enough to clearly see the performance. I have evaluated many hundreds of tests done by factories and brands that own a DA2 but are clearly not operated in a way that shows the true performance and more importantly issues that may be missed. The following is a test case with multiple issues that were not caught prior to the product making it to production.
The test driver is a 6.25” woofer with carbon fiber cone, large diameter CCAW coil and spider with a woven lead wire. Datasheets show two curves from DA2 testing Bl(x) and Kms(x). These are shown in Figure 1 and Figure 2, respectively. The datasheet shows a linear excursion of ±12mm and maximum excursion of ±15mm.
Figure 1: The 6.25” test driver with the manufacturer’s Klippel Bl(x) graph.
Figure 2: The 6.25” test driver with the manufacturer’s Klippel Kms(x) graph.
The Bl(x) curve (Figure 1) shows a very slight reduction at the ± extremes of excursion, with a pronounced inverted bell shape on the Kms(x) curve (Figure 2). From my experience, these curves do not show the real curve data and are limited by the protection limits in the LSI test. By default, these are set to 50% Bl limit and 50% Kms limits. This means that the test is limited to a 50% change in Bl and Kms. These protection limits should be set to a range that can identify the real limits of excursion.
For most of my tests I set these limits at 25% with a high temperature limit of 150°C. These are a more realistic approximation of what the driver will see in (aggressive) use. Note that I “ride” these values constantly during the test to prevent damage. The updated Bl(x) and Kms(x) curves depicted in Figure 3 and Figure 4 are tests accomplished with these limits. From these new curves you can clearly see that there is an asymmetry that needs to be evaluated. This can be done using the two new Bl and Kms symmetry range curves shown in Figure 5 and Figure 6.
Figure 3: The 6.25” test driver with Redrock Acoustics’ Klippel Bl(x) graph.
Figure 4: The 6.25” test driver with Redrock Acoustics’ Klippel Kms(x) graph.
Figure 5: The 6.25” test driver with Redrock Acoustics’ Klippel Bl(x) Symmetry Range graph.
Figure 6: The 6.25” test driver with Redrock Acoustics’ Klippel Kms(x) Symmetry Range graph.
The Bl(x) symmetry curve shows some offset as the excursion passed 8mm. The range below this is essentially a window of error caused by the highly linear Bl(x) (nice motor design!). The lower Kms(x) symmetry range curve shows a clear bias in the down direction. Both curves show about 1.4mm of bias. Because both curves show this bias, experience suggest that there is a mechanical offset.
Physical examination of the moving parts show a dished-out spider (the lead out tinsel wire was stitched onto the spider). This can be caused by many factors, but typically it is the result of a mechanical stack-up error. In this case, cone assembly parts do not allow a passive ± = 0 rest position. My conclusion is that the cone neck is too long, the spider neck up design wasn’t calculated correctly, and so forth.
If the Bl(x) symmetry curve looked correct, it would have also been possible that the stitched lead wire was causing asymmetrical limits in the spider motion. This is always a concern when using this type of spider. This can also mean a mechanical weakness after long periods of high excursion. The lead wires tighten more in one direction than the other and the joint between coil and the lead wire can fracture and cause and open short (Photo 1).
Photo 1: Close-up image of the broken tinsel lead joint.
As part of the analysis of this driver, I performed a simple stress test exercising the driver at Fo and 10mm of excursion. After about 5 minutes the driver failed and an examination of the connection point at the neck showed the expected fracture. There are several ways to fix this problem, but these are discussions to be had with the manufacturer. My recommendations are to use a neck down spider design with lead wires stitched to the top of the spider rolls. This should allow a cone assembly stack-up without bias and a far stronger lead wire connection at high excursions.
