output stage with high impedance

Why is not more publications and devises with this solution?

Click to expand...

Are you asking why more is not discussed about current-driven loudspeakers? The accepted interface standard between amplifiers and speakers is voltage. I imagine this is because it was historically easier to convert a voltage into a cone displacement than a current, and still is for magnetic, voice-coil designs.

What is your interest in a current source amplifier?

I imagine this is because it was historically easier to convert a voltage into a cone displacement than a current, and still is for magnetic, voice-coil designs.

Click to expand...

This is definitely NOT true: The force generated by the voice-coil is proportional to the current and not the voltage.

The (only) adavantage of voltage-drive is that taming of the driver's fundamental resonance is achieved more easily (i.e. cheaply).

Regards

Charles

This is definitely NOT true:

Click to expand...

I never said it was. You misquoted me.

sorry

Charles

if you are concerned about sending a current to a speaker, there is an easier way of doing that. Put a sampling resistor between the speaker the and ground. and move the feedback pickup line from top of the speaker to top of the resistor (or between the speaker and the resistor). This way, the feedback will be proportional to the voltage drop on the sampling resistor which in turn is proportional to the current doing through the resistor (and the speaker).

This has been done for many years, going back to the vaccum amplifier days. It is said to improve the damping factor and control of the coil. My personal experience with this topology is that it does clearify bass a little but otherwise has no audioable difference.

It depends heavily on the polarity of the feedback you take from the current sensing resistor. With positive feedback you make your amp's output impedance negative, increasing damping (i.e. decreasing Qes).
With negative feedback you will increase the output impedance of your amp, lowering damping (i.e. increasing Qes).

With frequency-dependant feedback you will get a frequency dependant output impedance, allowing more parameters to be altered artificially than just Qes.

But the current-sensing resistor is definitely the most common approach to make an amp behave as a current source.



Regards

Charles

I read the Mills and Hawksford paper. By using current drive, they indeed obtained reduced distortion and a reduction of the compression due to voice coil heating and temperature coefficient. However, they had to apply motional feedback with a special second voice coil to get a reasonable damping of the loudspeaker resonance. Perhaps you can apply this to loudspeakers with double voice coils intended for subwoofers, using one coil to drive the loudspeaker and the other for motional feedback.

It is very easy to make an amp with current output (nearly infinite output impedance) instead of voltage output - just find a principle of V/I converter with opamp. But does it make any sense? Every standard electrodynamic speaker must be supplied from a voltage output (zero output impedance) to achieve the best transfer and transient characteristics. Electrodynamic speaker is not designed to be driven from current output.

If you take an ordinary amplifier and take your feedback from a current sensing resistor (connected between "cold" wire of load and ground) instead of the amp's output - then you'll end up with a voltage controlled current source.

There are only three things to take care of:

-Le of your driver might affect stability.
-If no load is connected, the output voltage is perfectly undefined.
-You loose control of the driver's fundamental resonance.


Regards

Charles

alex0 said:

Mills P.G.L. , Hawksford M.O. Transconductance Power Amplifier System for Current-Driven Loudspeakers. JAES,
Vol. 37, 1989, 110 p. 809-822.

Please, send me copy of this article. I can’t get it in Moscow
in central technics libraries.

Click to expand...

It's rather cheap (for a swede) and very easy to buy articles directly from AES. 4 dollars for members and 5 dollars for non-members. Fast service and decent scanned pdf files (of old articles).
http://www.aes.org/journal/search.html

phase_accurate said:

-If no load is connected, the output voltage is perfectly undefined.

Click to expand...

for such a design, people usually put a small resistor in parallel with the load.

traderbam said:

I imagine this is because it was historically easier to convert a voltage into a cone displacement than a current, and still is for magnetic, voice-coil designs.

Click to expand...

hmmm... ...a voltage to current conversion is done by the impedance of the speaker. ie, I = V/R (or Icomplex = V/Z).

Since current is the driving force generating the magnetic field, it would stand to reason an improvement could be had if you removed the affect of Zspeaker from the output. Therefore you could have speaker drive current proportional to input voltage. By turning the amp into a current source, you move the V/Z to I conversion to a different Z, (From the speaker Z to a Z in the amp, which would hopefully be more linear.)

However, my guess is since you are radiating mechanical power from the driver, and that the power input to the speaker will always remain P = I * V, complex impedance of the speaker would still affect the signal (because V in turn would be generated by V = I*Z.... One way or another the inductance and capacitance of the driver would seem to affect the signal.

MarcelvdG said:

However, they had to apply motional feedback with a special second voice coil to get a reasonable damping of the loudspeaker resonance. Perhaps you can apply this to loudspeakers with double voice coils intended for subwoofers, using one coil to drive the loudspeaker and the other for motional feedback.

Click to expand...

I wonder how much cross talk you would get between the two coils in any of these modern woofers?

-Dan

I've done some experiments with a double voice coil subwoofer, current drive and motional feedback a couple of years ago. The crosstalk was pretty bad at high frequencies, but you only need the motional feedback near the resonance of the loudspeaker. By putting the right filters in the motional feedback loop, I could make it work.

Good note. Amplifiers are to be tested (for stability and distortion) to dummy loudspeaker load or better to real load (X-over etc.). Small example - Wilson Audio Maxx. This excellent speaker declared as 8 Ohms falls to some 2.2 Ohms somewhere near to 250 Hz. Just imagine the current that is to be delivered in this region. You are perfectly right, measurement into resistive load is only first attempt and approach. But every good amp designer must be aware of this fact .

Pavel

phase_accurate said:

If you take an ordinary amplifier and take your feedback from a current sensing resistor (connected between "cold" wire of load and ground) instead of the amp's output - then you'll end up with a voltage controlled current source.

There are only three things to take care of:

-Le of your driver might affect stability.
-If no load is connected, the output voltage is perfectly undefined.
-You loose control of the driver's fundamental resonance.


Regards

Charles

Click to expand...

- A Zobel (R-C-combination) in parallel to the load is definitely needed to keep the HF impedance down.
- A parellel resistor will do the job. A low value will be highly inefficient in terms of power consumption, while a high value might mean a large DC offset without a load.
- If you're using the driver well above resonance (i.e. midbass used with 200 Hz highpass or tweeter with 600 Hz resonance chamber used above 1.5 kHz, this doesn't hurt). If resonance is inside the passband, one could try to linearize the impedance with a paralell RLC combination. It might be possible to achieve resonably linear frequency response this way, but the driver sees essentially voltage drive then (the excess current goes into the external resonant circuit) which kind of defeats the purpose of lowering distortions.


Any idea if a Manger would have lower distortion with current drive? I do suspect the distortion peaks are due to overtones in the membrane rather than problems of the motor, so it probably would not change things...

Regards,

Eric

A Zobel (R-C-combination) in parallel to the load is definitely needed to keep the HF impedance down.

Click to expand...

This depends on what you intend to do. If you want to compensate for the effect of Lvc then you should definitely not use a Zobel network.

If resonance is inside the passband, one could try to linearize the impedance with a paralell RLC combination.

Click to expand...

Another possibility would be the use of frequency dependant feedback.

Any idea if a Manger would have lower distortion with current drive?

Click to expand...

I was already wondering also. I assume it might help to make it a tick faster due to the elimination of the effect of Lvc. The downside would be a higher risk for driving the amp into clipping.

Regards

Charles

if you base the feedback on the current going through the load, output power will be proportional to the load. That is less power output on a smaller load.

That is just the opposite of a traditional voltage feedback amp where a smaller load gets higher output.

so it is possible to combine the two and use both current and voltage feedback to reduce or eliminate the fluctuation in power output caused by variations in the load. That is, such an amp will output the (roughly?) the same amount of power to different kinds of load.

Wouldn't that be desirable?