Home Tweakers' Asylum

Tweaks for systems, rooms and Do It Yourself (DIY) help. FAQ.

[ Follow Ups ] Thread: [ Display  All  Email ] [ Tweakers' Asylum ]
[ Alert Moderator ]

Re: MOSFETs work fine too...

Likewise, if the cone is moving in the opposite direction (to a transient thump, say a drum sound) because of a lower frequency tone and assuming it has some momentum, then the load on the amplifier is greater as the impedance will be lower at that point.

You seem to be assuming here that the loudspeaker isn't responding to the transient. If it's not responding to the transient, then it's not reproducing it. And if it's not reproducing it, we can't hear it. So what's the point?

Now if you assume a loudspeaker that is responding to the transient, then the impedance will be whatever the loudspeaker's impedance is at the frequency/frequencies involved with the transient.

Of course the speaker responds, but not in the same way that it would if it weren’t moving in the opposite direction to the transient voltage in the first place. Before it can move in the opposite direction it must first stop.

Loudspeakers with instant response have not been invented yet. If the speaker is not moving in the first place then it does not first have to stop. The delay in changing directions is exactly what I am talking about. In that instant, if there were a resistor in series with the speaker the voltage drop would be greater for the cone moving in the opposite direction to the driving voltage than a cone at rest when that driving voltage was applied. That translates to a lower than usual impedance which a MOSFET amplifier with feedback responds to (see demonstration below).

I don't know what circles you move in, but where I come from, back-EMF is rather routinely described in relation to loudspeakers. In fact, you can't discuss a loudspeaker's electrical impedance without implicitly discussing back-EMF. It's precisely back-EMF that makes your loudspeaker a reactive load as opposed to a nice static, purely resistive load.

Yes, but you are including the voltage produced by an overshooting speaker as back EMF. It isn’t. Charge a coil and then take away the charging voltage then a voltage will be produced by the coil as the field collapses (lookup Lenz’s Law). If the applied voltage is AC then we can see this on a trace as a phase change. Obviously speaker coils are inductors as well. But the overshoot voltage is not back EMF (not caused by the collapsing of a field) unless the term is used more loosely than I had realised.

Square wave and triangular waves are not sine waves.

No, they're not. I don't recall saying or implying that they were. Though I might mention that square waves and triangular waves can be described and produced as sums of multiple sine waves.

Your obviously from the pre-digital era. Fourier transform reduces
to sine waves, Wigner transform reduces to square waves. Look it up.

But the driving voltage will meet a voltage that has the opposite polarity.

Not if the cone (or more specifically the voice coil) moves in the opposite direction as you state above. If the voice coil moves in the opposite direction due to the driving voltage being reversed, then it will see an opposing voltage. I.e. back-EMF.

Before it can move in the opposite direction it must stop. Before it can stop its momentum must be overcome. This momentum does not have to be overcome if the speaker is not moving in the first place. Before the cone stops moving, it will produce a voltage that you call Back EMF that is the opposite in polarity to the voltage that is trying to send the cone off in the opposite direction. This is the same as a transient low impedance that a MOSFET amp with feedback can correct for (though not perfectly).

What the amplifier is trying to do, whether it uses feedback or not, is to maintain a voltage across its output identical to that across its input except higher in amplitude (assuming the amplifier has voltage gain greater than 1). Period. In order for it to "push harder" it would have to increase the voltage across its output to greater than it would be otherwise (i.e. input signal times voltage gain). And if it's increasing the voltage across its output beyond this point, i.e. "pushing harder" then it's doing something it shouldn't be doing.

Not so. For a MOSFET amp (of the Hitachi MOSFET variety), the output changes in response to the load as I’ve maintained all along. Feedback is required to maintain a constant voltage so when the load increases, the output pushes harder, that is, the current rises. If the speaker produces a voltage, either from the back EMF caused by a collapsing field or voltage produced by the overshooting speaker, the amplifier corrects for those as well.

When the impedance of a load suddenly changes, as clearly outlined above, the output voltage changes and so, via the feedback circuit, the gate voltage is raised - the amplifier pushes harder.

In conclusion, if the driver was reproducing a low frequency sine wave at the same time a short high amplitude signal representing the leading edge of a drum sound or other percussive sound, then the amplifier without feedback will fail to control the driver properly - the change in impedance will be ignored. The amplifier with feedback will correct for some of the speaker overshoot (no amplifier/speaker is perfect) and produce a more faithful sound (in this case, a drum sound).

The amplifier controls the driver ultimately by virtue of its low output impedance. While feedback can help reduce output impedance, you can still achieve low output impedances without it. And even relatively high output impedances don't have a huge effect on controlling the cone's motion. That's largely taken care of by the voice coil's DC resistance.

An amplifier with an output impedance of 8 ohms (a damping factor of 1 into a nominal 8 ohm speaker) only has marginally less control over cone motion than an amplifier with an output impedance thousands of times lower.

Maybe for a dozy old low frequency sine wave. For transients, maybe its down to listening. I’m not convinced. If I repair an amplifier that has no feedback and it ends up being a bit dull on the transients then it is doing what it supposed to do. A transistor amplifier restored to its gritty worst - the customer will be happy. Properly designed MOSFET - now that’s much better.

BTW for a driver that is producing a single sine wave, and assuming the driver is liable to overshoot, then as the voltage delivered by the amplifier changes (toward the maximum positive end of the cycle, for instance), then the drivers impedance drops (it produces a voltage of its own).

This is incorrect. The point at which the driver has the MOST overshoot is at its fundamental resonance. And this is a point of MAXIMUM impedance. That "voltage of its own" you're talking about here is back-EMF. Which because it's in phase with the driving voltage manifests as a HIGHER impedance.

The fundamental resonance of a typical dynamic loudspeaker models as a parallel RLC resonant circuit across the amplifier's output. Do some modeling with that and perhaps you'll gain a bit better insight as to what's going on.

No, your overplaying the back EMF card. The coil will have a back EMF as any coil will, I never said it wouldn’t. But if you push on a speaker cone, it also produces a voltage. If the cone has some momentum then it will overshoot at the end of its cycle (moving in or out too far) and produces a voltage with the same polarity as you mention. This voltage makes the impedance appear to rise (effectively it does rise).

We agree on the fundamentals of these points. Consider a resistor in series with the driver. We see that the voltage drop over the resistor is small when the speaker overshoots because driving voltage (from the amplifier) and the voltage produced by the speaker have the same polarity.

Now what happens if, as the driver moves inward, the amplifier sends a signal with the opposite polarity? Lets call it a transient - the leading edge of a drum sound. The voltage drop over the resistor is going to be higher than if the driver were stationary when the thump signal came through. This only occurs for a fraction of a second after which the speaker is moving in the direction the amplifier is pushing it.

But that fraction of a second makes the difference between a dull thud and a crisp faithful reproduction. The momentary lower impedance causes MOSFET output to drop which is immediately corrected for by the feedback circuit.

I will perform a simple test right now. I will hook up a woofer to one of my amps and push down on the cone. This simulates a cone that momentarily does not move where it is supposed to, say, for whatever reason, it was moving in the opposite direction to the driving voltage. That motion must be overcome before it moves as it should (this only takes a fraction of a second).

(10 minutes later) I ran the test. I connected the scopemeter to the gate of the MOSFETs of the power amplifier and a small woofer to the output. I shorted the input of the amplifier. Now I pressed down on the cone. 100mV spike shows up in the gate (using Min/Max/average trace on the scopemeter - so that any spikes remain on the screen).

Now I connect the scopemeter only to the speaker. Pushing the cone in the same way I get up to 400mV.

Now I connect the speaker to the amplifier and check the voltage in the same way. When I push down on the speaker as hard as I can, no voltage shows up at all - the amplifier has corrected for the speaker’s voltage so that none shows up on the scopemeter. This would be typical of an op amp circuit with feedback - very hard to measure the correcting voltage as it all happens so fast - but a voltage does show up in the amplifier’s feedback circuit.

Just thought of a final test - with the amplifier off, do I still get no voltage at the speaker (when the speaker is connected)?

2 Minutes later. The Voltage is about half what it is for an open driver. Switched the amp back on and the driver voltage does not show no matter how hard I push on it.

This is proof positive - a very simple experiment to perform (anyone with a MOSFET amplifier and scope and a couple of minutes can perform it) and it shows that the amplifier is correcting for the voltages produced by the speaker.

So if it overshoots when reproducing a simple sine wave, say at its resonant frequency, the amplifiers driving voltage drops so that the voltage that speaker produces itself and the voltage of the output devices does not exceed the correct voltage (input multiplied by gain). Without feedback the voltage would rise.

The voltage does rise (for MOSFET amps of the Hitachi variety)- I measured it and reported the rise in my last post.

PS I just ran a quick test on a MOSFET amp I just finished (sounds sweet).
With 100mV input, I tested the gate voltage for various loads. If there were no feedback, the output would vary rather than the gate voltage. My dummy loads are all 10R 200W, so when wired in parallel I get 10, 5, and 3.3 ohms.
100mV in, 4.38V out (for all loads)
4.40 (gate voltage, no load)
4.99 10R load
5.42 5R load
5.80 3.3R load (2.4dB rise over no load voltage)

Yes. Again this illustrates that what your feedback network is doing is trying to compensate for the inherent limitations in the amplifier, not correcting problems with the speaker.

A ported enclosure with an 8 ohm speaker typically has an impedance ranging from 8 to 16 ohms, a sealed enclosure may have an even higher impedance peak. As illustrated above, without feedback the amplifiers output would change with frequency as the impedance of the speaker varies with frequency.

Electrostatic loudspeakers typically have an impedance dip at high frequency, typically at 10kHz, and at very low impedance, typically of 2 ohms or less. An amplifier without feedback would have an output varying by over 3dB.

Note that Hitachi MOSFETs (a typical MOSFET) varies in output according to load. The change in load is caused by the changing impedance of the driver. If the amplifier changes its output voltage in response to a change in the speakers impedance, then it is sensing the speaker and correcting for changes in the speakers impedance.

A low amplifier output impedance would achieve much the same thing, as you mention, but I’m not convinced that a MOSFET amp will perform properly without feedback and I’m also not convinced that corrections performed by the feedback circuit of a MOSFET amp does not control the loudspeaker to a degree (My systems are all active only, no passive X components).

But perhaps any finer points must be listened to (in a listening test, I mean). Or performed in controlled conditions (Lab conditions) which I’m not prepared to do even though I have the equipment to do a good approximation.

Kind Regards,
Robert Karl Stonjek.


This post is made possible by the generous support of people like you and our sponsors:
  Atma-Sphere Music Systems, Inc.  

Follow Ups Full Thread
Follow Ups
You can not post to an archived thread.

[ Top Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2024, Audio Asylum, all rights reserved ]
[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]