Servo sub basics

[Amphiprion]

Hey guys,

I'm looking into designing a servo-based subwoofer. I know there are multiple ways of going about this (accelerometer, DVC, SVC with sensing coil, and more). What I am looking for are circuit schematics (say for use with the often used ACH-01) and comments on practical considerations with the various implementation methods. All I've been able to find in my early google searches is really basic block diagrams and basic theoretical descriptions.

I've got a degree in electrical engineering, so feel free to refer me to technical references (AES journals etc), but most of what I've done with my degree has been DSP related and screwing around with matlab (there's almost no focus on analog electronics in modern EE curiculums and I'm a pretty recent grad).

Thanks for any help!

[Ludvig]

Setting up a servo circuitry is rather straight forward, and can be derived from various block diagrams. The design challange is to optimize the feedback loop. I recommend that you first spend some time learning and understanding control and regulation systems before you get into hardware. That's what a servo system is all about. To design a feedback loop that effectively linearizes the cone movement without getting into instability problems regardless what the operational circumstances may be.

[HMenke]

Ludvig, in servo subwoofer applications, are PID (proportional, integral, differential) circuits typically employed? Or only PD? If I remember control theory, "I" is used to correct small offset errors and "D" is used to enhance transient response. In a dynamic application such as a subwoofer, "I" perhaps is not required? I also remember that "D" has to be set carefully to avoid instalbility under step-input conditions.

Henry

[Ludvig]

That's correct. Integral control is unnecessary (I hate to spell that word, it never gets right) since the application target is to correct AC errors, not DC errors of the woofer cone position. I don't think that the need for derivation control is that big in low frequency systems since the result of speeding up the control scheme only increases the bandwidth of the loop. If the bandwidth is large enough with propotional control only, it will be much easier to optimize.

However, increasing bandwidth also means that you increase your stability margin so you could use more feedback gain which will increase the efficiency of the correction. Tuned right, a PD control yields better correction, but if it's not optimized correctly a simpler P control will give better results.

If one decide to use derivation control, one can always limit the slewrate of the derivation part of the circuit to prevent the derivation part to bottom out in case of input transients.

[capslock]

Location: Northern KY near Cincinnati

Why do I have the feeling you don't want to be thought of as a Southerner? :P


On a more serious note: there is a whole bunch of useful patents and papers by Backes & Müller who pioneered Servo speakers (including Servo midrange and tweeter!) in the 70s. Unfortunately, most are in German

If you do your control loop without any integral part, chances are that you end up with a low frequency oscillation of the rest position. This means
a) you have less usable excursion
b) the differential linearity becomes worse, hence the loop has to work harder get get the output signal right

[dyazdani]

Why do I have the feeling you don't want to be thought of as a Southerner? :P

Hey, what's wrong with the South?

[ThomasW]

Hey, what's wrong with the South?

It's to hot and humid, but the food is great! ...:wink:

There was an article years ago in Speaker Builder (now AudioXpress) about servo subs. I don't keep issues that old around.

There's probably tons of stuff about servo design available from the AES.

[HMenke]

Why do I have the feeling you don't want to be thought of as a Southerner? :P

8O You must be a damn Yankee! :P I love horses, hills, and bluegrass! I just want to locate myself nationally, but I definitely don't want to leave the impression that I live in Cinci-nasty. And as a native Virginian I shout, Sic semper tyrannis! WARNING: Don't get me started on the War of Northern Aggression...

Now that we have the introductions out of the way...

When I was studying control system theory back in '91, I was able to do some very cool modeling in MathCAD. Once you have the system equations, it's a piece of cake. I remember I modeled a DC motor-generator set to simulate an old-style elevator control system. The trick was to make the elevator come to a stop smoothly, but as fast as possible. If you dialed in too much "D", you got overshoot and ringing. Not enough "I", and it took forever for the elevator to stop.

You're recommending using "I" and the reasons sound convincing. What would be, in practice, the source of low frequency oscillation that would need to be tamed by some "I"?

Henry

[dyazdani]

When I was studying control system theory back in '91, I was able to do some very cool modeling in MathCAD. Once you have the system equations, it's a piece of cake. I remember I modeled a DC motor-generator set to simulate an old-style elevator control system. The trick was to make the elevator come to a stop smoothly, but as fast as possible. If you dialed in too much "D", you got overshoot and ringing. Not enough "I", and it took forever for the elevator to stop.

I did the same, but in MatLab... pretty neat, what you can do with it.