The Eventide H9000 Network Effects Platform is our new flagship multi-effects processor—our fifth flagship processor in 50 years. The H9000 project has been, and continues to be, by far the longest and most costly in Eventide history, but it’s time and effort well spent.
In 1982, we introduced the SP2016. There was nothing like it at the time. There were dedicated delays, pitch changers and reverbs on the market but the SP2016 could be programmed to be anything—it was the world’s first general purpose digital audio box. It was a reverb, a multitap delay, a vocoder, it ‘did’ granular, it was a looper, and anyone could write effects by using Eventide’s primitive ‘development system’ called SPUD (Signal Processing User Development). It accepted hardware ‘plug-ins’ (ROMs) from Eventide and others.
The SP2016 was designed before DSP chips existed. In 1980, CPUs were way too slow for real-time audio processing. In order to do anything even vaguely useful digitally, you had to roll your own custom DSP-style computing system in a modular way—by designing and connecting primitive add/subtract chips, multiply chips and a few memory banks. This type of distributed processing system was called an “array processor.”
By the late 1980s, TI’s 16-bit TMS320 became available, paving the way for our first DSP chip-based flagship, the landmark H3000. Our third-generation flagship, the H4000, was more powerful than H3000. The fourth-generation H8000, now ~20 years old, was more powerful still and also used dedicated DSP chips.
All of our effects are built using modules—building blocks that we research and develop and optimize for the particular architecture of processing in use. We use our library of modules and our proprietary application, VSig, to build algorithms/effects by interconnecting the building blocks. The downside of using DSP chips is that the development tools are less advanced/feature-rich than are the tools for general purpose CPUs like ARMs and Intel processors. It’s a rather big downside and we’ve longed for the day when general purpose CPUs would be fast enough to do the job.
By the late aughts, it became clear that ARM chips had become so powerful that they could do the job and, finally, DSP chips and their tools were no longer necessary. Near 10 years after we made the decision to migrate to floating point ARM processors, in 2018, we shipped our first ARM-based flagship, the H9000. Why did it take so long?
One of the main challenges was migrating all of our fixed point modules to floating point. This was a nontrivial task because the arithmetic matters—it can change the sound. The migration took several years and involved developing an automated ‘listener’ application which we presented at DAFx a few years ago (home-brewed AI!). It would be nearly impossible for a human (or even a big bunch of humans) to listen to each of our 1,600+ algorithms, changing every combination of parameter values while comparing the sound with every turn of every knob. We are confident that the algorithms in the H9000, running on ARM, sound the same as they do on our older integer arithmetic, DSP-based hardware. We have also rewritten VSig for the H9000, so now ARM-based, developing new algorithms should be less painful.
Each generation of our rackmount flagships has enjoyed a long life and we expect the H9000 to be the platform for new algorithm development for at least the next decade. New, faster ARM processors are introduced at a much more rapid pace than dedicated DSP chips—a key advantage. For ‘future proofing’ and expandability, we decided to mount the ARM DSP engines on four plug-in modules, each holding a quad core ARM processor. Each core serves as a DSP engine, giving the H9000 16 DSP engines. Owners of the H9000 will be able to update to faster processors over time.
Another challenge was designing a UI and remote application. With 16 DSP engines, 96 channels of audio, a host of formats, sample rates and whatnot, we struggled with devising an interface that was flexible, expandable and intuitive. The original UI spec was discarded after a round of user trials and we were driven back to the drawing board. The new UI introduced the concept of FX Chains to assist users in creating processing blocks that combine algorithms, handle multiple audio channels and support various control and automation capabilities.
Our new flagship leapfrogs its predecessors. Its eight channels of analog I/O has better specs than any of our previous products. With its 16 DSP engines, the H9000 can do more stuff, handle more channels of audio and unlike the H8000, it is network aware. It can support 96 channels of audio I/O over USB, ADAT, Dante, MADI and so on. It can be controlled from its front panel or controlled by our remote app/plug-in, Emote, by connecting over LAN or WIFI. While its raison d’etre is running effects, the H9000 can be used as a world-class audio I/O box—and digitally stream 8 channels of pristine analog audio.
The H9000 has three expansion slots designed to accommodate customer-specific I/O requirements. Today, we’re offering Dante and MADI expansion cards, but the slots are designed to accept additional audio channels, mic/instrument pres, other audio formats and more, which we can develop based on customer demand and market opportunity.
For the foreseeable future, the H9000 will remain a work in progress, with priorities influenced by ongoing input from leading audio pros from all sectors of pro audio and now our growing H9000 customer base as well. Completing the platform’s feature set is our highest priority. Currently, it is not possible to route between FX Chains and that’s an important feature. Also, we can’t easily distribute processing across modules—we have some applications in mind that would require all 16 cores running in concert processing a single algorithm (a ‘proof of concept’ algorithm that we created for one customer runs a 22.2 limiter!). We have also started working on surround sound Dolby Atmos applications. We continue to work on improving Emote with the goal of tailoring it for various ‘use cases.’ OK, now that I’m tossing around marketing jargon, I’m gonna wrap up.
The H9000 project is at the heart of everything that we do. It’s the platform that will spawn our future hardware and software projects. Getting off DSP chips and on to ARM was a huge step for us; now the road ahead is wide open and we’re eager to take the next steps forward.
Tony Agnello is managing director and co-owner of Eventide.
Eventide • www.eventideaudio.com