DSP = Dutch Super Power. With Grimm's track record of recording-studio master clocks, active monitors and celebrated MU1 server, that's a statement of fact, not some dark Grimmoire fiction. To explain their DMF aka digital motion feedback, designer Rob Munning Schmidt published this PDF. Rob worked at Philips R&D for 20 years, spent 10 equivalent years at ASML then taught mechatronics at the Delft tech university. Today he runs RMS Acoustics. In this most current white paper written especially for lay people, Rob explains what can be measured which then determines what can be corrected. He distinguishes between positional, velocity and acceleration feedback. He explains that the first kind increases stiffness, upshifts its driver's resonant frequency so reduces bass reach. For a sub that's counterproductive. The second feedback kind increases damping to minimize the resonance peak but offers no further advantages if the driver is well damped already.
His choice is acceleration feedback. It "decreases the resonance frequency with an effective increase of virtual mass. As a result, the flat response above the resonance frequency extends to lower frequencies as the preferred behaviour. Positional feedback is useless for loudspeakers, velocity feedback useful only around the resonance frequency. Acceleration feedback is useful over a wider frequency band so also reduces harmonic distortion at higher frequencies.
"Acceleration sensors are best to correct errors in speaker drivers by means of active feedback control. Unfortunately it's also why motional feedback in loudspeakers has never gone mainstream. It requires the skillful integration of an acceleration sensor with vulnerable wiring in the driver to become more expensive and prone to damage. As cheap alternative it is possible to derive a velocity signal from the voltage/current through the motor as is sometimes done in velocity feedback. In theory this derives the acceleration from a velocity signal by means of electronic differentiation. Unfortunately it increases noise at higher frequencies which will transfer to the driver.
"For those who think that they've invented the perfect positional sensor, for instance one based on a reflected laser beam or bending element… even if such sensors were perfect for positional sensing, an acceleration signal requires twice the differentiation (!) which increases the noise even more. Unfortunately cheap modern MEMS accelerometers with integrated A/D conversion as used in smart phones are not suitable yet because of their limited dynamic range, noise and bandwidth. We need at least 16-bit/96dB S/NR performance with a bandwidth of over 3kHz. So analogue piezo-electric sensors remain best and are what we use in our DMF subwoofers."
We now appreciate that the SB1 isn't another stock woofer in a cubed box. Instead it thinks real hard on why, if particularly amplifier circuits use feedback to minimize distortion, speakers whose distortion magnitude is far higher do not. To implement speaker feedback obviously relies on first going active. When most distortion occurs in the bass, applying corrective feedback to active subwoofers is the most obvious strategy. With the SB1, there's no port or passive radiator. With the LS1's sub a known quantity since 2017, the SB1 also is a next-gen effort with new now black woofer, new sensor, new sensor mounting, new DMF electronics, power supply and higher digital IQ for even lower distortion. What it's not is a hulking side table or extreme contraption à la Magico, Wilson or YG Acoustics. It's compact and with its grill perfectly salonfähig. Its closest cosmetic competitor is probably the also upfiring but bigger Wilson Benesch Torus MkII. Whilst considering a simple business end of XLR mono input, IEC mains and mains thru-put, we wonder. How does one set the SB1's volume relative to the mains? Others could ask whether one may turn the SB1 onto its side to back up against a wall; or parallel to the main speakers like an externalized sidefiring woofer; or turned to fire at the listener? Or would its asymmetrical forces go walkabout like the original Sunfire subs?
"At frequencies below 70Hz there's no acoustic difference in the woofer orientation since wavelengths are much longer than the cabinet size. The waves just bend around the box. With the SB1 sitting on the floor, its virtual acoustic center for frequencies below 70Hz rises to ~1m in height. This is independent of its orientation. The main impact of the orientation is moving mass. You'd rather not see it produce resonances into the floor or wall or main speaker system. We designed the SB1 to fire upward because that's how it transfers the minimum of energy into the legs of our main system below. It simply needs a sturdy floor. If you have a wooden floor or downstairs neighbors, you may consider turning the SB1 onto its side to place against a wall. It's a pretty heavy box so I don't expect it to move even freestanding. But it could help to place it on a thin rubber mat which would also protect its surface. Balanced-force woofers neatly solve the moving mass problem. We tested such a design but aesthetically it was too big to fit nicely between our LS1 legs. If after your review we see market interest for our sub in 3rd party systems, we might reconsider building it."
Eelco Grimm added that "our 15kg 358 x 340 x 213mm enclosure is MDF with very thick walls. We use a muli-layered bonded sandwich of CNC'd MDF. That's probably as expensive to make as a solid-metal enclosure but far sturdier and more inert. Do note that some manufacturers use feed-forward error correction or DSP pre-correction instead of feedback. Apparently KEF do this, too. To suppress the same amount of distortion is difficult with feed-forward because driver parameters can vary with temperature and age. It's an interesting technique but has even more pitfalls than acceleration sensor feedback. The requirement for a near-perfect model of the real world is only the start. Relative to your question on latency, feedback systems cannot have serious latency because latency determines the highest frequency that feedback can be applied without stability issues. A natural amplitude roll-off always comes with a phase shift. If the phase shift goes through 180° and you have negative feedback with gain, it flips into positive feedback to oscillate. The SB1 has an effective range up to 250Hz. However, the system should remain stable to much higher frequencies. The DSP we use has a total latency—incl. A/D and D/A—of just 38μs so is really negligible. Feed-forward systems often show far higher latency because the heavy calculations of the pre-correction rely on FFT filters with much longer processing delays." The Norwegian Arendal subs have DSP latency of 8ms which equates to a 2.7m path-length difference when set up equidistant with the main speakers. To correct for time alignment, their subs would have to sit 2.7 meters closer to our seat than the mains.
In anticipatory overview, the SB1 seems to be a very high-IQ proposition. But what else to expect from Eindhoven? Some call it Holland's Silicon Valley because it's a densely packed hub of hi-tech think tanks and manufacturing. With their sub, Grimm Audio continue where Phillips left off in the 1980 with motional feedback on active speakers. For 2021, the new SB1 promises bass of exceptionally low distortion and superior control from a compact enclosure. Rather than predictive DSP based on modeling a golden sample, its sensor-based feedback measures the actual driver's behavior against the very real and present signal. Still, "the SB1 is not ideal to use as a generic sub like a REL or such. There's no crossover adjustment, no volume control. Systems with an active variable-gain crossover or AV receiver cover that. It's then that the SB1 becomes an interesting option. And yes, you may run an RCA cable through a plug adapter if a true XLR cable connection isn't available." My next lesson in how to sub was afoot. Just then Doug Schneider had published his 2.2 system feature to prompt readers to try the sat/sub route for themselves. Perfect timing.
Would you like to download the current image?
Would you like to download the current image?