To The Builders/Manufacturers of the Forum

[Impious 1,476]

Hoping someone like Scott @ Fi or some of the other manufacturers/designers on this forum will chime in here.

When designing a driver how are small signal parameters and large signal peformance considered or correlated throughout the design and testing phase of the product?

For example; Are the target (small signal) T/S parameters and design analyized via computer program and then large signal performance tested simply by "playing" with and testing the physical driver? Or are the large and small signal performance predicted via computer and then the physical driver tested? Do you have any sort of target large parameter shift when designing a driver compared to the small signal T/S parameters?

Basically, what is the actual process in designing a driver and how do the differences between small and large signal analysis/performance fall into that process?

[Tirefryr 1,742]

Dude! That's a freaking awesome question. Let's see how correct VD is.

[DevilDriver 28]

Here is some general info I have gathered in conversation with some speaker engineers.

First step is to decide a target. What are you making and how much is it going to cost? What is the margin of profitability?

As an engineer, there is an inherent knowledge of working parts. FEA is often used at this point to simulate the end product, as virtually everything can be modelled pretty accurately in FEA if you are experienced with it. At this point, you can achieve simulated small signal thiele/small parameters as well as the BL, Cms, and Le curves of the driver and get a general idea of any thermal power compression that you will experience. This will give you a great idea as to the large signal measurements.

Some engineers I have spoken with are extremely concerned with the comparative values of the small and large signal parameters, and as such, all forms of power compression are big concerns. Others are willing to grant that power compression is a tough constant to avoid and will merely accept a certain tolerance.

Once the simulated designs are achieved, you can move on to prototyping. From here, there are a number of tests that can be performed, depending on how serious you are: THD and IMD measurements (although use of the GedLee metric becomes more supported all the time), subjective listening tests, and if you've got the wallet, you either send your driver off to Klippel, or you test it on the Klippel machine that you have in-house. This is great for providing a real-world expectation of the level of power compression you can expect.

If this meets your requirements, it's on to production. A few production units will be tested as well beyond the normal quality control program.

The level at which the power compression is analyzed will vary dependent on the R&D costs you're willing to put in. If your intention is to make a W7, then compression is both simulated and measured (JL is fortunate to be able to afford their own Klippel machine). If you're intending to make the 1000th TC2 variant, you probably won't be too concerned about it based on previous experiences and the relatively low-level concerns that a low-budget driver presents.

Of course, I have never personally designed a driver and have only played with FEA minimally as a 1st year engineering student, but this is what I have learned from a select few in the industry. I seem to recall Scott specifically mentioning that he utilizes FEA (and in fact, you will see many speakers marketed as "FEA optimized"), but take my opinion for what it's worth.

[Impious 1,476]

Very nice response Neil. From your description, it sounds like they use FEA aswell as their previously learned knowledge to come up with a pretty accurate idea of both small and large signal performance before any physical parts start going together.

Would still like to get some others playing in the sandbox though Since, as you mentioned, different engineers take slightly different approaches.

[///M5 2,833]

One other step beyond FEA and that is verification. Here is a snapshot from the B&W measurements of their cone showing no breakup. This is usually done after prototyping to verify that the modelling was done correctly or to adjust the model. They measure almost every part of their driver to optimize it; however, the downside is that a system to do this costs an order of magnitude+ beyond a Klippel. If you get their 800 brochure and DVD there is a really neat blurb on how they use the technology.

Of course, this is what I do for a living. Verify FEA models or build a measured model of a part instead.

[Aaron Clinton 2,372]

Great topic!