Dirac Research has been heavily involved in the automotive industry since the early days. As a research and software company, our mission is to deliver outstanding tuning algorithms and tools that are second to none. In order to live up to this goal, it’s of utter importance that we reflect on and understand the role software plays in the tuning process: What are the expectations for a tuning tool? And who will use it?
Approaches to tuning
Over the years, many different opinions and ideas regarding the requirements and goals of software-based automotive sound tuning have come to my attention.
On one end of the tuning scale, we have the manual tuning approach. ‘Manual’ here refers to the fact that the software being used is dumb— it doesn’t suggest any parameters or it won’t use any advanced algorithms. Instead, the software solely provides certain information and functionality to the tuning engineer. The tuning engineer will interpret the given information, then manually set required filters, gains, delays, etc. The manual tuning approach is highly dependent on skilled and experienced tuning engineers, whereas the software only needs to provide rather basic functionality.
On the other end of the automotive tuning scale we have the automatic tuning approach. Automatic here means that the software autonomously performs the entire tuning. The automatic tuning approach does not require an educated tuning engineer but is highly dependent on the presence of a robust software design.
Between the two ends of the scale we have space for sheer endless combinations of the two approaches. These combinations are best described as assisted tuning approaches, where the level and type of software assistance available to the tuning engineer can vary significantly. Further, I would like to mention the emerging virtual tuning approaches, which are not directly covered here, but can also be viewed as a type of assisted tuning.
Influencing factors: Sound quality, time/cost, and consistency
Now let’s set the stage for a deeper discussion by taking a look at some examples of the factors and contexts that influence the tuning approaches we take.
Through talks with different stakeholders in the context of presenting our products to customers, I’ve been confronted with contrasting arguments for each of the tuning approaches. For the sake of the argument, let me introduce two extreme positions. Managers tend to argue for automated tunings, expecting consistent and high quality sound to be delivered in almost no time. While engineers tend to plead for manual tunings, fearing their skills and craftsmanship will become an unnecessary product once the software has done the job for them.
Without a doubt, a skilled tuning engineer can meet and even beat sound quality expectations and perform a perfectly executed tuning. Manual tuning approaches are often an important part of a brand identity, which can be highly appreciated or demanded by an OEM (Original Equipment Manufacturer). However, manual tunings are by definition very dependent on the engineer, and quite time-consuming. Not all of the required tuning work is fun and demanding for the engineer, and much time is spent on rather boring and less challenging parts. The manager might be unhappy because the engineers become locked into a project for a long time and the costs for manual tunings are high. Securing a car in the audio lab for a longer period of time can also be challenging. Furthermore, the outcome is dependent on the tuning engineer themselves, so brand consistency can become an issue. For entry-level systems the reachable sound quality is limited by sound system design and components, and a skilled tuning engineer reaches a ‘good-enough’ result, fast. Nonetheless, tuning entry-level systems is more annoying than fun. The manager might also be reluctant to assign his engineers to such projects, preferring to use their expertise for high-level systems.
This is where a completely automated tuning approach would be very beneficial. The basic idea is to ask any person on hand to follow an easy manual: Put microphones in the car, press a button, and then, after some calculations, the tuning parameters are written into the hardware. The idea is very tantalizing to management because it’s fast, cheap, and doesn’t require trained personnel. To some extent, this approach is surely feasible. Constraints on the automated tuning approach are robustness, handling of versatile system layouts, and algorithm limitations due to limited DSP capacities. In order to ensure that an automated process is robust for all (or more realistically let’s say most) cases, the tuning must be performed very carefully and the full potential of each individual system will not be reached. Default settings that are too aggressive may work in the first or second test, yet they will most likely fail in the third or fourth and for sure in a later one. Furthermore, the lower the number of tuning parameters, the more difficult it is for an algorithm to determine meaningful settings.
For example, let’s assume that we have four biquad filters for each channel of a 4-channel sound system. A tuning engineer will easily use these to fix the worst flaws in the spectrum by inspection of measurements. For an algorithm, this is a very challenging task. The more parameters that are provided, the easier it gets for the algorithm. For entry-level systems, it’s likely acceptable not to reach the full potential and an automated tuning can be expected to be better than no tuning at all. Yet, the limitations of an automated tuning should be understood before choosing this method. Inspection and adjustments by a professional engineer are likely a good idea in order to secure a satisfactory tuning result, which renders the automated approach to be less automated.
For a high-level system automated tunings are, in theory, a possibility too. The algorithm has many parameters to work with and thus, could create great filters. Yet again, automated tunings will not reach the full potential of the sound system, which is a big deal for high-level systems, which easily set the customer back by 4 digits.
There is however a third position, which myself and many of my peers strongly promote: The assisted tuning approach. The software solves the parts that are either too boring or time-consuming to perform manually. The Engineer solves the fun parts which fully require their expertise and craftsmanship.
Given a good algorithm, the tuning can be automated to the extent that 80-90% of the performance is reached (numbers sourced from professional Dirac Unison user feedback). Engineers can then focus on the last 10-20%, contributing their tuning expertise where it matters most and keeping the spirit of manual tunings alive. The total tuning time is reduced significantly, and the consistency between tunings is raised. Certainly, different ratios between the automated and manual part are required for entry or high-level systems.
Nevertheless, I argue that the assisted tuning approach is generally a sound approach. It reduces the total time and costs for tuning, provides value for the OEM and marketing, strengthens brand identity through higher consistency across tunings, guarantees full potential of the sound quality, and lets the Engineers work with what they’re best at— tuning car sound systems with the assistance of smart software solutions.
– Adrian Bahne, Product Manager Automotive at Dirac Research