AudioXpress 11/03 Servo Sub article

[tiroth]

I'm curious about a few things so I thought I might start a discussion.

Any comments on this design? From the scope traces it looks like the reduction in distortion is an order of magnitude better than the Analog Devices accelerometer-based servo sub. I suspect AD's implementation suffered so badly from phase shift that most of its value is eliminated; I don't think a 10-20% reduction in THD is really that significant in a sub.

1. There are seperate adjustments for current sense gain and sense coil gain. The author states he adjusts these until he finds a combination where they are "similar"...similar to what? What is the proper adjustment procedure?

Since R17 is not affecting the feedback to signal ratio I would guess that that sense coil adjust determines the amount of feedback and the current sense adjust is used to null the error. But how do we determine when the error is lowest? The author doesnt seem to offer any hints.

I wonder if adding another adjustable buffer stage after the differentiator might not be a bad idea. That way you wouldn't have to null the error again if you change the feedback ratio.

2. This design appears to have been tested at low power, although that is not stated anywhere. When applied to 12"-15" 250-1000W design, will the higher currents cause the design assumptions to fail?

Any other comments? The magnitude of distortion reduction means I will definitely be trying this at some point, but I wonder if it will be as effective with my Shiva as with the 8" peerless.

[moamps]

Originally posted by tiroth
There are seperate adjustments for current sense gain and sense coil gain. The author states he adjusts these until he finds a combination where they are "similar"...similar to what? What is the proper adjustment procedure?]

Hi,

IMHO:
In the equation for sensor voice-coil voltage Vs(t)=c(dx/dt) + M(di1/dt), the first part may be made equal to zero if we block the coil from moving. The result is that the Vs(t) is equal to M(di1/dt). When we apply signal to the speaker (low level - 1W or so), we can set null at the output of the opamp D, with the relative changing of R1 and R9.

Since R17 is not affecting the feedback to signal ratio I would guess that that sense coil adjust determines the amount of feedback and the current sense adjust is used to null the error. But how do we determine when the error is lowest? The author doesnt seem to offer any hints.]

By measuring distortion with microphone and by setting the absolute levels of R1 and R9 to minimum distortion.

I wonder if adding another adjustable buffer stage after the differentiator might not be a bad idea. That way you wouldn't have to null the error again if you change the feedback ratio.]

Or placing a trimer resistor (10k) instead of R15 to do the job.

This design appears to have been tested at low power, although that is not stated anywhere. When applied to 12"-15" 250-1000W design, will the higher currents cause the design assumptions to fail?]

IMHO, applying servo makes sense only in low-power (HI-FI) designs.

Regards

[MarkMcK]

Although I work in a coherent transient world instead of a steady state world (sine waves) or chaotic world (noise) I am responding to the query for “any comments.”


For me it was a thought-provoking article. In that way the article had value even if I don’t build one of the circuits.

I like the idea of trying to cancel out voice coil cross talk. I am not certain the assumption of a relationship between current flow in the drive coil and cross talk to the sensing coil is valid. If it is, then at system resonance, where feedback might be most useful, almost no cancellation will occur. In a sealed system, the impedance goes high at system resonance, little current will flow and almost no difference voltage will appear across R7. If that is the case, considering the low pass filter knee point of about 100 Hz, then I am uncertain at what operating frequencies the cross talk cancellation circuit will be valuable. From my own experimentation, cross talk at system resonance is not a concern.

General comments:
First, the technique will work with other drivers just as well as it works with the Peerless driver.
Two, if you go higher in power watch for clipping or overloading of the op amps. Voltage off the sensing voice coil will come close to the voltage applied to the drive coil. I also am uncertain how much headroom is provided with the so-called differentiator circuit. These are not fatal flaws, just things to watch.
Three, this is a complex circuit with lots of op amps. Watch for oscillations in any of these op amps. The breadboard photograph shows LF347N IC s with no high frequency power supply by-pass caps on the board. This is a circuit for the person who loves to “tweak,” has access to a high bandwidth oscilloscope, and some way to protect both amp and driver during initial set up.
Four, I wish the author had better instrumentation available to test his design. Single frequency sine wave testing can be misleading. A matter of author’s choice was involved in choosing photographed sine wave frequencies for the article. Even the student paper on AD’s Web site showed frequency dependent differences in distortion reduction.
Five, I wish we were told how much feedback was actually achieved using the design. Looking at the RTA graphs, you cannot tell. If closed loop feedback is high, then the author has adjusted for the decreased sensitivity.
Six, I wanted to see a “phase” comparison between the drive sine wave and the differentiator output in addition to the comparison between the sensing coil output and the differentiator output. How close is the processed sensing signal to temporally matching the drive voltage and how much mismatch is there between the drive signal and the sensing coil signal? With a complex, broadband signal, a mismatch can produce ripple in the acoustic output that 1/3 octave analysis is unlikely to detect.

In sum, the circuit will work to provide MFB for a woofer, but while the cross talk canceling circuit is a nice idea it may not provide any real benefits. I am also suspect of how the combination differentiator/integrator circuit will work in the transient world. I can’t promise I will ever build and test this circuit, but if I do I will post results to this thread.

Mark

[zelter]

Hi tiroth,

is it possible to find that article somewhere in the internet?
It's quite hard to get a copy of audioXpress here in Germany.

Zelter

[tiroth]

Not to my knowledge. I'm afraid that I do not have a scanner or I would be happy to send it to you. Perhaps someone else can oblige?

[JoeBob]

I recieved a copy, from Rarkov (excuse my spelling, it's from memeory) I do believe, it was mentioned in another thread. I had many doubts about the differentiator circuit, and would like a better way of doing it, maybe an OPA541 could play a role here (or simmilar OPA54X chips, not sure which is suited for this) for +/- 15V is not enough for my use, a normal opamp will clip in this situation especially if you're eq'ing the bottom end.

The crosstalk did seem to be maybe wirthwhile if it indeed does have a large effect, but I wished the author had taken measurements with both parts of the circuit appart (crosstalk elimination and differentiation).

I'd like to try the differentiation, for with normally a 90 degree phase difference between the input (into the first voice coil) and output (out of the second voice coil)) signals is rather large and even if not perfect, I think that differentiation would be able to improve on that. I think I'll go the OPA54X route and see what I can achieve, unless anyone has an idea for a better or easier way (not enough time for all these projects)...

[zelter]

Hi everibody,

could someone post a copy of the article or send it to me private?
I'm just working with a dual voice coil driver to create a regulated subwoofer.

Zelter

[herm]

FYI;

The same driver that Dan Ferguson used in the article
(Peerless 831858) is on the Madisound sale page for
27.50 USD

Hey moamps,

How would you block the driver to stop the coil from moving?
I'd like to give this a try.

herm

[moamps]

Originally posted by herm
How would you block the driver to stop the coil from moving?

Hi,

Practicaly speaking, this isn't easy task.
I was trying this with one cheap Monacor TC speaker by pushing cone maximum inside the speaker, and then applying small power (half watt) on it. Generated voltage (unloaded) in receiver coil was ca 10 times smaller then on generator coil in low frequency range. This isn't the best method especially if coil isn't long type.(M is changing with coil position..... anyway this is "part of linearity problem").

Regards

[MarkMcK]

MOAMPS wrote:

I was trying this with one cheap Monacor TC speaker by pushing cone maximum inside the speaker, and then applying small power (half watt) on it. Generated voltage (unloaded) in receiver coil was ca 10 times smaller then on generator coil in low frequency range.

Risking oversimplification, isn't this an approximation of crosstalk?

If so, then we can expect crosstalk to be at least 20 db down (10 to 1 voltage ratio) at low frequencies.

How much MFB do we need? If it is less than or equal to 20 db, then perhaps we could do away with the crosstalk circuits.

Just wondering,

Mark