[Joules]

Hello Hugh and All - It's been a long time

Hugh I've been thinking about your concerns  about Back EMF from speakers causing havoc with the output stage of an amp and your efforts to combat this problem. (I find it all very fascinating)
So ... would it be possible to accomplish a poor mans reduction in Back EMF effects by using high efficiency speakers with very low Q and then add back some resistance in series with the speaker to bring the Q back up to reasonable levels, there by making the load look more resistive and less reactive. With a high 90's dB/w speaker it seams as tho one could reduce reactive effects by 3 - 6 or more dB and still have a reasonable efficiencies.

Thanks

[AKSA]

Hi David,

Yes, a long time, I trust you are well and your Lifeforces are batting on nicely!

This is a difficult topic, because it divides into many areas depending on whether you look at it from the speaker or the amp POV.

There are gaps in my knowledge, so I will confine myself to what I know, and try not to speculate.

Back EMF from a speaker comes from the inductance (a typical voice coil is around 2mH) and from the momentum of the cone after excitation producing an EMF across the voice coil which finds it way back to the amp.  The first is electrical, fast, and very small, but the second is mechanical, slow, and produces a voltage at the driver which does not reflect what is happening with the signal - that is, it's delayed by the mechanical system.  Both systems produce a small voltage, related to input signal but delayed, and this voltage is fed back to the feedback system of the amp (assuming its not a Zero FB amp) where it has the potential to produce problems, literally, since the fb loop is controlled by voltage and nothing more.

Both amps and speakers are designed for voltage feedback.  Many argue the benefits of current fb, but that's another story.  It's used in very fast opamps, but not generally in audio amplifiers.  But voltage is only voltage IF it's impressed across a resistance, or in this case, an impedance.  The driver has a source impedance, an ability to drive a load, and typically it's around 6 ohms.

The impedance seen by the speaker is the amplifier output stage.  It's extremely low due to feedback, and on the AKSA and Lifeforce it's around 20 milliohms.  So any voltage impressed on the amp by the drivers will be immediately shunted to ground, effectively, because this voltage has a source impedance of 6R and a load impedance, the amp, of only 0.02R. In fact, the voltage impressed across the amp will be 0.02/(6+0.02), which is around 0.33% of that originally produced, unloaded, at the driver.  Another way of considering this is that only around 1/300th of the back emf at the speaker voice coil finds its way into the output stage of the amp to corrupt the fb signal, and this we can live with.  Shorting rings within the magnetic structure of many drivers further reduce this back emf, and this creates a high Q for back emf, damping it out quickly, so it ain't all bad........

This is typically low enough to prevent back emf from interfering with real time signals coming into the speaker from the amp and it is probably the reason why driver manufacturers stick to the voltage feedback standard.  Most amps use vfb, damping factors are always high (that is the amp has an effective low impedance to a driver) and this is the time honoured convention.  No reason to change.

Enter the amp with high output impedance, like a single ended triode.  Zout is high, two or even four ohms, and so back emf might be a problem.

But in truth, it's not, because the single ended triode has no global feedback, so the amp input stages do not see what is happening at the speaker interface at all.  It's blind, so marches on regardless, attempting to pin the tail on the donkey.  OTOH, no feedback can distort the picture, so the amp just slavishly follows the input signal.  All is not lost.......

Now, consider your high efficiency low Q speaker.  It requires very little drive voltage - but this is merely a question of magnitude, so the broad phenomenon does not much change.  The low Q does mean it produces a lot of back emf, granted, and this would be a problem for a feedback amplifier, but if you place a resistor in series with the amp, it would merely chop the efficiency in half (for a 6R resistor!) massively compressing the signal, lowering output and negating the benefits of high efficiency drivers by 6dB.

But it would lower the back emf reaching the speaker, and on a global nfb amplifier, this could be an advantage if damping factor is not high.  Conclusion:  At a loss of efficiency around 6dB, a 6R series resistor would indeed reduce back emf reaching the amp, by around half, and thus might make it good IF the amp had poor damping factor AND global negative feedback.   A 50W amp can swing around 60Vpp, so if half this swing were lost across the resistor and half across the load, if the driver were efficient it might sound rather nice.  Slam and impact would be compromised, but it would sound rather twee - a bit like a SET!

Hope this helps, difficult topic without resorting to math, which I'd rather not do!

Cheers,

Hugh

But it would

[Daryl]

Hi guy's,

Back EMF is a concept you learn of in the biginning of electronics study to help explain the behaviour of inductors and motors.

It only applies to a single instant in time and so if you try to describe what takes place when a signal is applied which is constantly changing you end up chasing your tail because nothing will hold still so you can dwell on what's happening and quantize it.

You would only think of back EMF while first learning of the concept of inductors and motors, beyond that there is no use for the term back EMF.

Once you have learned these concepts you will use an impedance vs. frequency curve in place of back EMF from then on.

Impedance vs. frequency is a frequency domain curve rather than a time domain curve.

Where before I mentioned 'chasing your tail' trying to grasp what's going on because nothing will hold still, that disappears in the frequency domain because time does not exist in the frequency domain and each point on the curve is for all time and now every thing holds still for you to quantize it.

The most important thing is that the time domain and frequency domain are exactly the same information.

So frequency response tells you impulse response and vise/versa.

Same goes for back EMF and impedance, the fourrier transform of back EMF is impedance and vise/versa.

Impedance describes back EMF completely and no consideration need be given to any type of delay or what the speaker cone is doing and so on.

What this leaves you with is simply your speakers impedance and how it affects your amplifiers feedback loop.

Adding resistance in series with a speaker is going to effectively increase your amplifiers output impedance and allow your speakers impedance curve to be superimposed upon it's frequency response as AKSA already said.

You could measure the impedance vs. frequency of your speakers and build a conjugate network that you connect in parallel with your speakers that will make their load resistive across the entire spectrum.

This requires equipment and electronics knowledge.

I you just wanted to apply a brute force approach you could connect some resistance in parallel.

8 ohm in parallel with an 8 ohm speaker would level it's impedance considerably but would also increase the load to your amplifier.

[AKSA]

Thanks Daryl,

And this introduces the next topic......  impedance of speakers, and how it varies with frequency.

This is REALLY important, and when you look into it, you find many speaker manufactures who compromise the impedance curve so severely that the load becomes very difficult to drive with conventional amplifiers.

This is a hobby horse of mine;  the speaker with an impedance peak of 45R and a dip to 1.5R.  Don't laugh. These things are common, and they play merry hell with amps.

The other problematic speaker is the electrostatic.  These are almost completely capacitive - very difficult to drive - and a capacitive load has a huge phase shift, up to 90 degrees, between applied voltage and measured current.  Think on this a while;  a simple 6R resistor will pass 2A when 12V is applied across it.  The current will always keep step with the applied voltage, and amps love this!  Problems start when the applied voltage and measured current get out of step, and this happens with speakers.....

I raise this issue because I have in fact found a way to completely immunize an amp from its speaker - the reactance has NO effect at all.  I use this technology in my Soraya CB105 amp, and the improvement in sonics and musicality is extraordinary.

Cheers,

Hugh

[Steve Eddy]

Daryl brings up some good points.

Instead of thinking in such anachronistic terms as "back EMF," it's better to look at the speaker as an RLC resonant circuit as that's how it fundamentally behaves electrically.

Considering a single driver speaker, "back EMF" is greatest at resonance, which is illustrated by the large peak in the impedance plot of such a speaker. But also at this point, the speaker appears purely resistive and voltage and current are in phase.

Below resonance it appears more inductive and above resonance it appears more capacitive until you reach the voicecoil's DCR and then it starts appearing more inductive as the voicecoil's inductance starts taking over.

Of course things get a bit more complicated when you start adding more drivers and crossover networks but the point is the same.

se

[andyr]

OK youse guys (Hugh & Steve),

Can you answer a Q I have. 

"Are impedance-correction circuits (there is a name for them but I fergeddit) only relevant to drivers which are driven by passive XOs?  Or do they apply equally in active setups - where the amp is directly connected to the driver?"

It seems to me (as an amateur) that, given that an impedance-correction circuit flattens the otherwise-rising driver impedance (which increases with frequency), it is absolutely necessary for a passive XO ... as otherwise the nominal X ohms DCR which sets the XO component values according to the standard Butterworth etc. filter formulae, will vary with frequency (so the theorectical filter slope will not be achieved in real life).

However, it seems to me that an impedance-correction circuit is just as necessary for an active setup.     (Hence that adjustment, Hugh, which you made for marcus's Orion mid-range driver?)  Otherwise the amount of power which the amp is able to put out will vary widely with frequency, if the impedance ranges widely (like from 1.5R to 45R)?  Which won't sound any good at all!   

Regards,

Andy

[Steve Eddy]

It seems to me (as an amateur) that, given that an impedance-correction circuit flattens the otherwise-rising driver impedance (which increases with frequency), it is absolutely necessary for a passive XO ... as otherwise the nominal X ohms DCR which sets the XO component values according to the standard Butterworth etc. filter formulae, will vary with frequency (so the theorectical filter slope will not be achieved in real life).

Yes. That's why real crossover design isn't as simple as one might think.

However, it seems to me that an impedance-correction circuit is just as necessary for an active setup.     (Hence that adjustment, Hugh, which you made for marcus's Orion mid-range driver?)  Otherwise the amount of power which the amp is able to put out will vary widely with frequency, if the impedance ranges widely (like from 1.5R to 45R)?  Which won't sound any good at all!   

Don't see that that's particularly an issue. I mean, the power the amp puts out will vary widely with frequency just due to the music signal itself.

The variation in impedance is more an issue where amplifier output impedance is concerned where if it's high enough you can get significant variations in frequency response due to good ol' Ohm's Law.

Check out some of the Stereophile reviews of tube amps. John tests amplifier frequency response both into a straight 8 ohm resistive load as well as a "simulated" loudspeaker load (2uF in parallel with 8 ohms). You'll see the variations in frequency response due to the variations in the load impedance with frequency.

se

[andyr]

I mean, the power the amp puts out will vary widely with frequency just due to the music signal itself.

se

Thanks, Steve ... but I'm afraid I don't quite understand what you're trying to say, here. 

OK, my logic goes as follows:
1. If an amp can put out 100w rms into 8 ohms then it can produce a max of +/- 'X' volts rms at the output terminals.  (Sorry, I forget the maths.  )
2. If it was powering a driver whose impedance was a constant 8 ohms then the rated input voltage would deliver the same 'X' output voltage at all frequencies.
3. However, if the driver's impedance has risen to 16 ohms at 10Kz then the amp is not capable of outputting 'X' volts at 10Khz ... for the same input signal level, the output voltage will be considerable less (half?  quarter?).
4. Thus the driver is less loud at 10Khz than it is at 1Khz ... which is not very good! 
5. Hence, an impedance-correction circuit is required to smooth the impedance of the driver, so that it registers a constant impedance at all frequencies ... so that the amplifier produces a constand 'X' volts at all frequencies.

How is the above logic flawed?

Regards,

Andy

[Steve Eddy]

Thanks, Steve ... but I'm afraid I don't quite understand what you're trying to say, here. 

OK, my logic goes as follows:
1. If an amp can put out 100w rms into 8 ohms then it can produce a max of +/- 'X' volts rms at the output terminals.  (Sorry, I forget the maths.  )
2. If it was powering a driver whose impedance was a constant 8 ohms then the rated input voltage would deliver the same 'X' output voltage at all frequencies.
3. However, if the driver's impedance has risen to 16 ohms at 10Kz then the amp is not capable of outputting 'X' volts at 10Khz ... for the same input signal level, the output voltage will be considerable less (half?  quarter?).
4. Thus the driver is less loud at 10Khz than it is at 1Khz ... which is not very good! 
5. Hence, an impedance-correction circuit is required to smooth the impedance of the driver, so that it registers a constant impedance at all frequencies ... so that the amplifier produces a constand 'X' volts at all frequencies.

How is the above logic flawed?

It's flawed in that it assumes that the amplifier's output in volts changes depending on the impedance of the speaker. But unless the amplifier's output impedance is quite high, then it's output voltage won't change depending on the impedance of the speaker. Instead, what changes is the current through the speaker.

se

[AKSA]

Laurie Menogue and I have just finished a two way speaker called the Aspen VSonics which have a minimum impedance of 5.28R and a maximum of 7.8R - across the entire range from 30Hz to 20KHz.

Together with the tuning of the speaker and the proper phase and amplitude matching of the drivers, the crossover was by far the most difficult design task and took around 8 months.

Thanks Steve, you explained it well.  And your work is artisan quality, if I didn't already remark on it!

Cheers,

Hugh

[Joules]

Thanks Hugh and others
It has always been a point of debate as to weather to apply a Zobel to a driver that has been connected directly to an amplifier. The overall opinions is inconclusive (as with eveything else in audio) when considering theory and subjectivity. The thoughts and opinions expressed here just give me something to think about when I have nothing else to do.
Disregarding the Back EMF- Inductance semantics thing (reactance happens) the concept was to explore ways to make a speaker appear more resistive and less reactive. I'll collect some ideas sit and think about them and through out the bad ones. Al tho given the right drivers the simplicity of just adding resisters (perhaps 32ga speaker wire) has some KISS appeal.

thanks again

[kyrill]

hi Joules
"I'll collect some ideas sit and think about them and through out the bad ones. Al tho given the right drivers the simplicity of just adding resisters (perhaps 32ga speaker wire) has some KISS appeal."

Can you let us know your findings at that time?

[Steve Eddy]

Thanks Steve, you explained it well.  And your work is artisan quality, if I didn't already remark on it!

Thanks, Hugh. And yes, you did. I believe it was your post I replied to where I brought up the carbon sequestration program.

se

[Steve Eddy]

Disregarding the Back EMF- Inductance semantics thing (reactance happens) the concept was to explore ways to make a speaker appear more resistive and less reactive. I'll collect some ideas sit and think about them and through out the bad ones. Al tho given the right drivers the simplicity of just adding resisters (perhaps 32ga speaker wire) has some KISS appeal.

Yeah, but simply adding resistors won't really do much toward making the speaker appear more resistive unless your resistances are much greater than the reactances. And given that the impedance at resonance can be upwards of 100 ohms for some speakers, you're going to need so much resistance not only will you end up with a horribly inefficient speaker, but you're also going to end up with a huge peak in bass response due to the speaker's mechanical Q (Q_ms).

You've heard the phrase "fight fire with fire"? Well it's rather the same here. Fight reactance with reactance. This come about from the fact that capacitance and inductance are ultimately opposites of each other. So if you want to null capacitive reactance, you use inductive reactance and vice versa.

se

[Joules]

Disregarding the Back EMF- Inductance semantics thing (reactance happens) the concept was to explore ways to make a speaker appear more resistive and less reactive. I'll collect some ideas sit and think about them and through out the bad ones. Al tho given the right drivers the simplicity of just adding resisters (perhaps 32ga speaker wire) has some KISS appeal.

Yeah, but simply adding resistors won't really do much toward making the speaker appear more resistive unless your resistances are much greater than the reactances. And given that the impedance at resonance can be upwards of 100 ohms for some speakers, you're going to need so much resistance not only will you end up with a horribly inefficient speaker, but you're also going to end up with a huge peak in bass response due to the speaker's mechanical Q (Q_ms).

You've heard the phrase "fight fire with fire"? Well it's rather the same here. Fight reactance with reactance. This come about from the fact that capacitance and inductance are ultimately opposites of each other. So if you want to null capacitive reactance, you use inductive reactance and vice versa.

se

Hmmm Good point - however the application I have in mind would be a mid, high passed an octave and a half above Fo and low passed just below breakup, the impedance curve should be at is lowest and flattest.
but you are right I need to plot an impedance curve to really analyze this. or just give up and Zobel.

[AKSA]

Joules  (David),

You made this very sharp comment:

It has always been a point of debate as to weather to apply a Zobel to a driver that has been connected directly to an amplifier. The overall opinions is inconclusive (as with eveything else in audio) when considering theory and subjectivity.

In my opinion, it is ALWAYS worth adding a zobel.  Why?  Because, consistent with Steve's line about 'fighting fire with fire', the Zobel effectively reverses the phase shift introduced by the voice coil inductance, and in so doing, makes the driver look like a pure resistance electrically.

This means the amp sees very little phase shift across the driver.

In turn, this means voltage and current are in step, and this is GOOD for the feedback loop, enhancing stability and making the amp smile......

So, always use a Zobel.  Oh, and when you listen to an active system with and without Zobels on the drivers you hear quite marked differences in the imaging area.  This is because driver phase shifts have strong influence around crossover on gnfb amps, where all the spatial information is found, that vital first watt.

Cheers,

Hugh

[andyr]

In my opinion, it is ALWAYS worth adding a zobel.  Why?  Because, consistent with Steve's line about 'fighting fire with fire', the Zobel effectively reverses the phase shift introduced by the voice coil inductance, and in so doing, makes the driver look like a pure resistance electrically.

This means the amp sees very little phase shift across the driver.

In turn, this means voltage and current are in step, and this is GOOD for the feedback loop, enhancing stability and making the amp smile......

So, always use a Zobel.  Oh, and when you listen to an active system with and without Zobels on the drivers you hear quite marked differences in the imaging area.  This is because driver phase shifts have strong influence around crossover on gnfb amps, where all the spatial information is found, that vital first watt.

Cheers,

Hugh

Thanks for sorting out my question, Hugh.

Regards,

Andy

[Joules]

Joules  (David),

You made this very sharp comment:

It has always been a point of debate as to weather to apply a Zobel to a driver that has been connected directly to an amplifier. The overall opinions is inconclusive (as with eveything else in audio) when considering theory and subjectivity.

In my opinion, it is ALWAYS worth adding a zobel.  Why?  Because, consistent with Steve's line about 'fighting fire with fire', the Zobel effectively reverses the phase shift introduced by the voice coil inductance, and in so doing, makes the driver look like a pure resistance electrically.

This means the amp sees very little phase shift across the driver.

In turn, this means voltage and current are in step, and this is GOOD for the feedback loop, enhancing stability and making the amp smile......

So, always use a Zobel.  Oh, and when you listen to an active system with and without Zobels on the drivers you hear quite marked differences in the imaging area.  This is because driver phase shifts have strong influence around crossover on gnfb amps, where all the spatial information is found, that vital first watt.

Cheer

Hugh

Thank you Hugh - Coming from you this means everything.
I will experiment

[gerado]

Hugh does the Zobel argument apply for active crossovers that correct for phase ( eg DEQX)

[AKSA]

Hi Theo,

Absolutely, yes.  The Zobel at the speaker corrects for phase shift at the speaker, which has benefits for an amp with gnfb.  The DEQX corrects for phase shift at the line level around crossover, and this is at a totally different point in the circuit.

One must aim for both......

Answer your email later tonight, putting kits together!

Cheers,

Hugh