Audio DSP is at the heart of music software and digital hardware, yet it is often overlooked in various ways by product makers. Developing high-quality DSP algorithms is indeed error-prone, time-consuming, it requires deep expertise in computer programming and extensive knowledge in a large number of other subjects. This easily translates into substantial costs when hiring capable DSP engineers or somehow acquiring the required skills. Furthermore, it can be difficult to objectively communicate desired sound characteristics and quantify sound quality. No wonder digital products had a bad reputation for a long time.
While there are some peculiarities that make music DSP development unfortunately more complicated than other engineering tasks, we can still apply lessons learnt in other fields by tastefully adapting them to our case. I'll describe two potential complementary approaches: on one hand I'll draw a parallel between software libraries, object-oriented programming abstractions, and patching systems (such as Pure Data and Max/MSP); on the other, I'll discuss DSP programming languages in their compositional aspects. Code reuse, and thus modularization, will be the common theme, and it will be tackled also from a "cultural" point of view.
Hopefully this talk will help:
- startuppers to use their budgets more wisely, avoiding taking shortcuts that lead nowhere or spending too much on custom DSP development;
- companies with little internal DSP knowledge or few resources to catch sudden market occasions;
- established companies to rationalize effort and commit resources where it does really make a difference.
IF YOU ARE ATTENDING ONLINE, ALL TALK SESSIONS CAN BE ACCESSED FROM THE MAIN LOBBY:
https://conference.audio.dev