All Podcast

Podcast

Practically Useful DSP w/ Sam Fischmann (Musik Hack) | Ep 34

Musik Hack’s Sam Fischmann on designing plugins by ear, choosing the right filters, when oversampling actually matters, and why no DSP idea is ever “done”.

Three Things You Can Take Away From This Episode

  • Your job as a DSP developer is reverse engineering vibes – the words people use for sound rarely map to maths, so listen to what they want, not how they describe it.
  • Start with a state variable filter and a Butterworth response, then move only when the sound tells you to.
  • Most of a great plugin’s development time goes into tuning the feel, not building the DSP.

Sam Fischmann came to plugin development from two directions at once. He studied music, spent years as a software developer, and then moonlighted in LA recording sessions for R&B and hip-hop. When COVID hit, he joined The Audio Programmer Discord, merged the two careers, and went on to co-found Musik Hack.

In this conversation, Sam and Josh dig into the practical side of digital signal processing: how Musik Hack turns complex mixing processes into simple, playable tools, which filters Sam reaches for and why, and the business realities of running a small plugin company.

If you’re learning DSP, or you can build the maths but struggle to make it feel good, this one is full of hard-won practical advice.

1. If a Control Doesn’t Make Sense Sonically, It Doesn’t Exist

Sam’s design philosophy started at a production camp. While everyone else was making music, he was building plugins with generic JUCE controls, and a producer who went by Sauce sat down and tried them. His feedback stuck: if he moves a control and can’t hear what it’s doing, he moves on. It doesn’t matter what it does technically.

That shaped everything at Musik Hack. A compressor user doesn’t want 3:1 with a tempo-synced release. They want to fit a sound into a mix louder without smashing the drum’s pop. Musik Hack’s tools aim straight at that musical goal, so people get the result they’re after faster, whether they’re first-time producers or working professionals.

2. You’re Reverse Engineering Vibes

People describing sound use words like punchy, warm, 3D, even colours. None of those have a mathematical meaning, and they don’t mean the same thing to two different people.

Sam’s advice is to gather sound examples, keep asking questions, and listen closely to what someone wants from the sound. He’s also clear that the person describing it may confidently explain the wrong technical cause. Great mixers know how to get a sound, but not necessarily what’s happening under the hood. Your job is to hear past the explanation to the goal.

3. From Idea to Parameters – Listen, Maths, Listen, Guess

Musik Hack’s process starts with a broad idea, usually shaped with co-founder Stan, who mixes and masters every day. Sweetie Q’s high control is a good example: people use all sorts of tools to add energy to the top end, and nearly all of them bring harshness with it.

Sam tries lots of approaches, reads white papers, runs listening tests and gradually narrows in on something that works across a wide range of material. Then comes the long part: tuning how multiple processes move together under one simple control so the sweet spot is as wide as possible. There’s no formula for that, which is why Musik Hack releases roughly one product a year.

4. Where AI Helps, and Where It Doesn’t

Everything at Musik Hack is written in C++, but Sam has recently started using Codex to try out ideas faster. It’s useful for translating MATLAB into C++, or explaining a paper eight different ways until something clicks.

What it can’t do is tell you what sounds good, or solve the kind of problem where filter overshoot is quietly making a limiter overreact. Sam’s point is that you still need a working understanding of DSP building blocks to know what to ask for, and to know when the answer is wrong.

5. A Practical Guide to Choosing Filters

Sam spends a good chunk of the episode on filters, which he sees as one of the most important things to understand early on. Filters change both amplitude and phase, and forgetting the phase side is what trips people up when they split a signal and recombine it.

His rough decision path: IIR filters for zero latency, minimum phase unless you really need phase coherence, biquads for fixed filters, and state variable filters for anything that moves. He points to Andy Simper’s free state variable filter designs from Cytomic as a great starting point. Then choose a response. Bessel is smooth with minimal phase distortion, Butterworth is flat up to the cutoff, Chebyshev trades ripple for steepness, and elliptic gives the steepest cutoff of all.

6. The Hidden Superpower of State Variable Filters

One reason Sam defaults to state variable filters is that they give you high-pass, low-pass and all-pass responses from one calculation. The all-pass output has the same phase response as the filtered one, without changing the frequencies.

That means you can process one path, take the all-pass version for the other, and mix them back together without phasing problems. You get phase compensation for free, without falling back on linear phase filters and the latency and pre-ringing they bring.

7. You Often Don’t Need Oversampling

Oversampling only matters when you do nonlinear processing, like distortion or saturation, that creates new frequencies. If those frequencies go above Nyquist, they fold back down as aliasing, the “guinea pig chirps” you hear when you sweep a sine wave through distortion.

Oversampling moves that ceiling higher. But on real-world material, aliasing is often too quiet to hear, and Sam wishes people wouldn’t switch oversampling on by default. If you do hear a problem, start with JUCE’s built-in oversampling and its minimum phase filter, and only go further if you have a specific reason to.

8. Nothing Has “Already Been Done”

Asked whether all the DSP has already been discovered, Sam disagrees completely. You don’t need new maths to make something new. How you combine building blocks and present them to people is what unlocks new ways of making music.

He compares it to pop music reusing the same four chords while pushing production forward. He also challenges the habit of treating analog gear as the benchmark for quality. What makes great analog gear great is its playability and generous sweet spots, and a digital tool can achieve the same thing through good UX.

9. Running a Small Plugin Company

On the business side, Sam is honest. Pricing is part feel and part comparison with what others charge, weighted by how often people will use the tool. Rent-to-own helps make tools accessible, and fighting piracy too hard is a distraction from making the product better.

For feedback, he relies on a small group of beta testers on Discord and gives extra weight to people making records every day. The biggest lesson was about help: a small company is better served by several reliable people handling support, socials and design than by one expensive specialist, because everything you can stop worrying about frees up headspace for the work only you can do.

Closing

What ties Sam’s approach together is a focus on the person using the tool. Filters, oversampling and saturation curves all matter, but only in service of something that sounds great and feels wonderful to play. Sam credits The Audio Programmer community with keeping him going in the early days, and his path from Discord newcomer to respected plugin maker is a good reminder that the building blocks are there for anyone willing to keep listening and iterating.

‍

‍

Audio Software Development
Digital Signal Processing
Audio Effect Design
Music Tech
Business
UI & UX
SHARE THIS

More Podcasts

View All
View All