The two feedback systems are not as different as claimed. They both involve generating a signal based on driver movement which is fed back to the amp to control the woofer excursions. The current accelerometers manufactured (as used in the current Genesis lineup) should be as quick to respond to the signal as the separate coil approach for direct servo. The implication that the direct servo has no feedback is a bit misleading, as the signal from the coil is a form of feedback (which controls the woofer). In other words, both approaches should be more than quick enough to do the job correctly, with no measured gain in actualperformancee from one approach to the other.
We never claim that we have no feedback. We have feedback, servo feedback.
Until you actually implement a servo system, you wouldn't realize how true the saying "the devil is in the detail" is. I designed my first accelerometer-based servo system when I was in graduate school. Before that I have built at least more than 5 different types of amps, including one with variable bias power amp that reduces crossover distortion. At that time Nelson Pass was already well known for his own variable bias circuit and I thought I can built one without reading this patent. My design was indeed different from Nelson's. The reason why I didn't commercialize it? I gave it to my brother and 2 times in 3 years, that amp fried the output stage transistors. Any common sense will tell you that is the design you should stay away unless you want to make yourself broke. So lesson one, robustness of a design is paramount.
What shocked me when I first brought up my first accelerometer-based servo sub was it took more than 5 seconds for the system to settle after I powered up my lab grade power supply. That thing went through a low frequency oscillation (I mean the woofer cone just go in and out at 1/2hz) for full 4 seconds. I repeated several times, on and off the power supply and it was the same thing. In contrast, in all of my early power amp building experience, the only oscillation that I needed to watch out was high frequency one (anyone who build their own amp knows what I mean) and now I have to adjust my mind to adapt phase margin analysis at high frequency to low frequency. Then I began to dig into the spec of the accelerometers. There is a spec called low end frequency roll-off point. Then I found that there are accelerometers of 1-axis, 2-axis, and 3-axis. Then I found the accelerometers are built with a mass glued to piezo film and the piezo film is aresponsiveeponsive to pressure.
Each of these features brings in a challenge to to a stable and robust design. First this low end frequency roll-off point is a result of this mass-spring struture. It is a 2nd order high pass filter. So at DC, the phase shift is 180 degrees. What is the phase margin? zero. That is not the worst yet. The signal from piezo film is so weak, it needs amplification, an active device with DC offset. You cannot feed that directly into amp without a decoupling cap. That set back the phase margin by another 90 degrees. So at DC, the phase shift is 270 degrees. Try make a system like that be stable.
Second is number of axis. Most people think the voice coil makes piston like ideal stroke. But in reality, there can be lateral movement in particular at high excursion. If the accelerometer picks up lateral movemnt and feed it back to the system, would it cancel the lateral movement? No. It introduces unwanted signal and makes wrong adjustment. The accelerometer should ignore lateral cone movement which it cannot correct. What if the cone is in break-up mod?
Third, acceleromter does not respond to acceleration directly, instead it is through a spring-mass system with piezo film attached with a loading mass. One immediately notices our driver is also a spring-mass system. We are trying to address a problem created by it and in order solve it, we created another spring mass? But a more immediate problem is if the housing can withstand the externale from extenal. If you google, there are tons of solution proposed for how to make the housing immune to external pressure. But solutions conflict with the requirement that the accelerometer needs to be light weight to be suitble for speakers.
Lastly, there are some comment that servo feedback signal has a lag. I always defend this for our servo system by saying that our sensing coil is at exactly the same location where the electromagnetic force takes place (that is, the driving coil) and there is no physical distance between the sensor and driver coil where the action takes place. And after the sensing coil picks up the voice coil movement (it is not even cone), the signal travel at light speed and race back to the amplifier. However, I cannot say that for accemerometer-based systems. Accelerometers cannot be placed with the voice coil and it has to go the the cone under the dust cover. It takes time for the wave to travel through the voice coil former to the neck of cone and from cone into the location of accelerometer. That may be a distance of 1" or even 2". Wave proprogation speed in former and cone? Maybe 3000ft/sec. It will still take 30 micro second to response. That is definitely not negligible.
Now each of these problems mentioned above is where our system excels. The phase shift of our system at DC is 90 degrees. There is no way our system will become unstable. Our sensing coil is a natural 1-axis sensor. It will ignore all lateral movement and ignore cone breakup. It will remain stable even if cone is detached from voice coil. lastly the sensor is directly responsive to velocity, not pressure.
Again, my strong suspicion is that the two subs would sound more alike than different, and measurements would point to both designs having very low distortion.
Well, our system has more microscopic dynamics. Our design has not changed from day 1 of our business. On the other hand, Velodyne has gradually reduced the feedback amount of accelerometer feedback signal, in part to cope with the problems that I mentioned above. Their current design is already very different from their early day ULD-15 or HGS design.
