Cart
Quantity :
Loading...
If there were ever a contest for the most influential multi-effects pedals of the recent past, Eventide's H9 Harmonizer® would be a top contender. Released in 2013, the pedal made decades of the high-end, sought-after DSP algorithms Eventide was known for accessible and portable for a wider range of musicians than ever. Guitarists, vocalists, bassists, synth players—all carved out a dedicated spot in their rigs for the H9. In depth, versatility, and quality of effects, very little came close. A lot has changed at Eventide since then: the H9's younger but bigger sibling, the H90, and the flagship H9000 are built around powerful ARM processors—and now the H9 itself joins them, with a much-desired overhaul.
To celebrate, we caught up with Eventide's long-standing DSP and sound design engineer, and the company's new President and CTO, Russell Wedelich. Starting with Russell's winding career path, we dug into his sound design wisdom, the new H9, and how an innovation-driven company like Eventide navigates the ever-changing landscape of music and sound technology.
Perfect Circuit: Hi Russell, thanks for taking the time. Before we get into the gear, let's start with you — you joined Eventide nearly two decades ago. Can you walk us through that journey, and how you ended up running the company?
Russell Wedelich: I graduated from Texas A&M University with an Electrical Engineering degree in 2003. I was obsessed with playing music (bass guitar) at the time and didn't want to work for a big defense contractor like many of my other EE friends, so I just kept taking gigs. I was able to squeak a living out as a musician for the next three years. However, part of my brain missed the challenge of deep problem solving, and being on the road takes its toll, so I decided to go back to graduate school in 2006. I graduated in 2008 with an M.S. in Music Tech from University of Miami.
I dove head first into signal processing in graduate school, devouring the math and programming for advanced digital filter theory and pattern classification (pre-AI Bayesian statistical stuff). I met Tony Agnello at the AES 2007 Conference (when the H910 was being inducted into the TEC Awards Hall of Fame!) at a career panel and I must have impressed him somehow because they eventually offered me a job. I hadn’t heard of Eventide before that (wild, I know), but the musical nature of the company was appealing to me. I wanted to stay close to musicians and the creative process for a career.
The first two weeks all they gave me was an empty desk outside of Tony Agnello’s office and a manual for the Motorola 56k series DSP! “Learn to write assembly for this chip kid, we got stuff we need you to do.” Ha! To this day, Eventide is an environment where a person hungry to solve big open ended problems can thrive. I didn’t really get to work on music products for the first couple of years, but I learned valuable technical skills that needed to be second nature by the time I did.
Dan Gillespie (now owner of Newfangled Audio, a partner company) was my direct boss at the time, and after a few years of working hard, I was finally rewarded with a big music project -> lead DSP and sound design for the Space reverb stompbox. We’ve always worked well together, and I can safely say this side of history, we knocked that one out of the park. I’ve loved reverb design ever since, it’s so artistically rewarding. We got a TEC award for Space and I believe that product put me on this path to running the company. From there I’ve just been living the “product lifestyle,” one after the other, trying my best to always be a student along the way. I’ve had the pleasure of working with so many amazing humans here.
×
PC: After many years of writing DSP and overseeing the creation of a number of Eventide products, what are the most exciting questions for you to work on today? Is there a class of problem or an effect category (pitch, time, space, dynamics) that you keep coming back to, or that you particularly like working with? Or perhaps it's other aspects of working on a product?
RW: I’ve always loved tearing apart signals, finding musically useful abstractions, and putting them back together in new ways. This is akin to source separation, but also includes filterbanks, source/filter separation and frequency domain transformations, etc. The tonal / transient split in SplitEQ fits the bill perfectly.
Recently Woody Herman, a senior DSP engineer, was able to use a technique in a class of processing called “modal reverb” to separate any reverb into early and late reflections in a plugin called Temperance. I loved hearing that a real plate reverb doesn’t have much early field at all (as theorized) or that most of the early reflections of a spring reverb are in the high frequency transversal mechanical vibrations (whereas the dominant “drip” late reflection sound is caused by torsional mechanical waves).
My favorite days are when Eventide finds a new class of sound I’ve never experienced. When we fully engage both our engineer and artist selves this almost always happens, even with effect categories where folks consider all new ground has been broken.
PC: The new H9 is the big news right now. Over its thirteen years on the market, the original earned a devoted following. What made this the right moment to transition to version 2?
RW: Easy answer. I thought the H90 would take up the H9 mantle, but I was wrong, and it ended up being its own thing alongside the H9 Gen1. The original H9 is our last product running on the 56k DSP and we pushed that platform as far as it could go. The H9 needed a rebirth, luckily its older siblings (H9000, H90) had a lot to draw from. We have this opportunity to have one cohesive hardware family that can share sounds at multiple price points, so I jumped on it.
PC: One of the first things you notice about Gen 2 is the more immediate, hands-on interface and the larger screen, which is essentially a direct response to perhaps the only consistent criticism of the original. What was your goal with the new design, and what were the main challenges of getting it right?
RW: Listening to our customers' feedback mostly sums up our main goal. We wanted a more immediate and accessible Eventide experience. You can now drive the pedal easily from the main UI, but of course you can still go deep and tweak there too. All that and having the family of FX units I mentioned above means this new paradigm has to be product agnostic at the Algorithm and Preset level.
The main challenge of getting the UI right was keeping the small form factor of the original H9 while holding balance with the new H90 hierarchical mode paradigm, i.e. Play Modes: Select / Perform and Edit Modes: Presets / Parameters. We also wanted more user configurable options: assignable meta controllers like Hotknobs, more Hotswitches, and more Play Mode Quick Knobs, as well as completely configurable footswitch options in Perform Mode.
×
PC: The new H9 runs on the same ARM platform as the H90, which is a big leap from the original's hardware. Besides the shared algorithm library across the two units, what does this jump open up for you as developers?
RW: That’s the main big advantage. A unified algorithm and sound design workflow across all our products keeps us in the “flow state” of creativity and technical innovation. This becomes a huge customer advantage, single alg H9, dual alg H90, and 16 alg H9000 all have access to similar Eventide innovations.
PC: With Gen 2 you've dropped the multi-tier model, upgrading every H9 owner to H9 Max for free on the way out. Why was it important to you to do it that way?
RW: Seemed like the right thing to do. Obviously we’re hoping those customers stick with us for the long haul by buying into the new platform one day when they are ready. But the original H9s are out there and still helping folks make great art. We’ll do our best to support them for as long as is feasible. If customers didn’t already have all 52 algorithms on their H9, we figure they were just missing out on some fun inspiration. The next 22 we’ve done for the new platform continue the inspiration!
PC: What are your personal favorite updates in Gen 2?
RW: The ones we’re planning on releasing in the future! Seriously though, as a DSP engineer I’m a fan of all the new Algorithms first and foremost. So much new sound exploration in that small package. As a personal example, not that folks are asking for it, but I can take my simplest four string bass and an H9 Harmonizer® Gen2 running Cosmic Web and put on a whole solo ambient bass performance (again no one wants this, but it’s a fun rainy day activity).
My other personal favorite update is Spillover on all Preset changes. It feels so smooth and professional when you play live.
×
PC: When an algorithm moves across hardware generations, what are the challenges of keeping its sonic character consistent? Do you ever find yourself engineering artifacts and glitches back in, once new hardware no longer produces them naturally?
RW: We obviously didn’t want to change the sound of Algorithms like Blackhole. This particular move was less about designing stuff back in (recent work on modelling H3000 converters was more that, but that’s a different story) and more about porting code in a manner cognizant of the underlying platform processor differences.
Our legacy Motorola 56k DSP code was written in the specific assembly language for that line of DSPs. Luckily, the lion’s share of that porting work was completed when we made the jump from H8000 to H9000 and native plugins. So for H9 Harmonizer® Gen2, it’s essentially a done deal.
Going deeper: The 56k is a fixed point processor (ask your friendly neighborhood CS / EE) and we moved to floating point ARM processors using C++. Therein lies the greatest challenge. These are two different ways processors represent and do math. I know it’s hard to believe, but math can have different answers for the same operation if the number representations are different! Pure mathematics assumes infinite precision and that’s just not possible. For example, have you ever heard the phrase “There’s no such thing as a perfect circle?” I bet you folks at Perfect Circuit have!
Many of our Algorithms are packed full of both nonlinear (saturation and distortion) and time-varying (modulation) goodies which present a further challenge. A linear time invariant system (a basic EQ) is easy to measure differences on, just subtract two system output signals, if the answer is very close to zero, they likely sound the same. For nonlinear and time varying systems we actually developed a perceptual machine learning algorithm to help point us in the right direction for the thousands of algorithms and dsp modules we needed to port. From there it’s about picking the right underlying DSP structures and process so that sounds are perceptually as similar as possible. Most times though, it would just catch porting bugs.
PC: Let's talk about your Spectral Instantaneous Frequency Tracking (SIFT) technology for polyphonic pitch shifting. One of the hardest problems in pitch tracking is responding accurately to very different sound sources—H9 and H90 players run everything from guitars and voices to synths and amplified household objects into these boxes. How do you approach that problem, knowing the range, and what still feels out of reach?
RW: I can’t say too much about this in any detail. I’ll say this though: engineering is all about acceptable tradeoffs. One of our guidelines for Eventide Audio signal processing is, if your algorithm must fail, make sure it fails musically. In playing music, most folks don’t notice the bad note in your solo if you don’t make a face.
In time-frequency processing, there are universe level rules, you can’t have good resolution in both time and frequency. Good frequency resolution will cost you time (latency), good time resolution will cost you some ability to resolve individual frequencies (notes or harmonics really in pitch tracking). But…. that doesn’t stop us from trying to bend the rules towards an artistic or musical end. Our greatest challenge of course is achieving something close to real time performance in a stompbox. Does it work perfectly? Well nothing does, but we musically mitigate the known failures (algorithm guitar solo face!).
The problem of polyphonic pitch FX is easier for offline processing, where you have all the latency you need, like in a DAW. That’s a better place to try to pitch shift dense bandwidth complete song material. I’d say that type of signal is slightly out of reach for a real time live performance. But then again, feel free to try it, you might discover a sound we never anticipated.
In summary, try not to let your algorithms flinch when they fail, or on the other hand, as Miles Davis put so well: "Once is a mistake. Twice is an idea. Three times is style."
×
PC: You've mentioned in other interviews that the interactive aspect matters a lot to you—what makes turning a knob feel playable. With instruments that's a natural question, but with effects it's not an obvious expectation. How do you think about playability in effect design, and does the answer change depending on the type of effect, or is it a general set of principles that runs through everything?
RW: This is connected to the “fail musically” diatribe above. Parameters should be smooth to change and produce interesting results throughout the entire range with usable minimum and maximum values. I believe there's a psychological feedback loop here. A musician learns to trust their instrument. FX should behave like they belong in that loop.
From there it really does depend on the type of effect or even specific parameter. We use our musical sensibilities to make judgement calls. Every engineer on the team is an outstanding musician. Playing through our FX is very much part of our development process at all steps along the way. The meetings where we make these judgement calls are some of the most fun, most challenging, most productive meetings.
Re: smooth to change: Our main general principle is NEVER have parameters click or pop or crackle when you manipulate them. They should be as smooth and automatable as possible. If I’m a stickler about anything, it’s this. And it’s because Dan was stickler with me on it while working on Space.
There’s a standard set of tools we can draw from to smooth out changes in systems that aren’t inclined to be time varying (and that’s most digital signal processing by the way). For example you can slew rate limit, filter, or cross fade parameter changes. Each of those has various advantages, disadvantages and their own sets of tunable parameters that must be carefully chosen. If those don’t work, we get creative, and it’s usually just hard trial and error work.
Re: the entire range: Not only do start and stop points matter, but tapering in between them is something to pay careful attention to. You don’t want any part of the range to be boring or feel like it’s not accomplishing anything. It should be fun to explore!
PC: Eventide's name was built on sound processing, and the Harmonizer® series above all. But recent work like Misha, the Newfangled Audio excursions into chaotic synthesis, and the Laurie Spiegel Music Mouse revival pulls toward instruments and performance interfaces. Is that a deliberate direction, and one you plan to explore further?
RW: Deliberate is a strong word here! I don’t think I could plan these things if I tried. Tony and Richard have a lifetime of relationships with some good people who also happen to be great artists. We are all very grateful. We make connections and products along the way.
We try to stay open in this creative journey, and we have so many creative friends like Dan Gillespie, owner of Newfangled, Leon Gruenbaum, inventor of Samchillian (which Misha uses), and Laurie Spiegel (no intro needed!) that have walked along with us in the journey at various times. I find the partnerships are a great way to stretch out of your comfort zone as a company. We are always open to exploration. And, afterall, Richard was the chairman of the SETI League!
(BTW If we end up talking to aliens, the first thing I’d ask them to do is play some of their music. Well first I’d make sure they’re from an atmosphere that supports sound vibration as we know it, even if not, we’re engineers, we can modulate any wave based artform to human being audio rate!).
PC: The landscape of music technology is changing rapidly, with AI looming over much of it. How do you experience this moment as a company, and how do you experience it personally, as a creative engineer?
RW: My stock answer to the AI question is “Dammit Jim, I’m an engineer not a prophet!” Ha! But thankfully you didn’t ask me where I think this is all going.
As an engineering forward company, we’re adjusting. That’s what we’ve always done. We iterate and adapt. As long as I’ve been here and before “AI” was even a common phrase we’ve been using machine learning and will continue to do so in ways that match our values (creativity, integrity, and trying to make a small, hopefully positive difference). I mentioned one example above, our perceptual listening algorithm for porting over legacy DSP code. For Temperance we also developed a machine learning algorithm to convert room impulse responses into modal responses.
Personally, as a creative engineer, the processing power available these days and the audio signal processing applications for training neural nets (e.g. “AI”) is wide open, so in that way, it’s fun! More discovery, new ideas, yay! Smaller networks (not everything is a chat bot) can solve meaningfully interesting problems and create new sounds.
I’ve had some surprisingly interesting DSP “conversations” with some LLMs. They aren’t slackers that’s for sure, they know their stuff. I do have ethical issues with the data sourcing of the big LLM and Generative AI companies. I’m no lawyer, but they likely broke some laws. It won’t matter, they are likely too big for the slap on the wrist they’ll get to even sting. Hopefully we all benefit in the long run.
No AI was used in these answers by the way! I enjoy my own thinking process too much at this point to farm it out or augment it.
PC: Before we wrap up, what's next for Eventide?
RW: I won’t say, but there’s always something! That’s why we keep doing it, so we can do the next thing.
The information below is required for social login
Sign In
Create New Account