Impedance and current-source amplification

[Svante]

About current control of a loudspeaker, I don't have the refernce handy but there was in JAES a paper indicating that current control of a loudspeaker can result in a dramatic decrease in distortion in the medium...

Yes, that is true. Any effects of the nonlinearity in the voice coil inductance will be eliminated. On the other hand, driving the driver with a negative output impedance of -Re will completely eliminate any nonlinearities of the cone suspension and box air cavity.

Both will also affect the frequency response a lot.

There are different sources of distortion, and they dominate in different frequency regions. They cannot all be simultaneously eliminated by changing the driving impedance, but they are all affected by doing so.

Therefore, the optimal driving impedance must be determined for each frequency and one must not forget the more important demands on the frequency response when doing so. This is very tricky, but for really carefullly designed speakers, one can reduce the distortion by applying the right driving impedance.

 

[kelticwizard]

I think this is a misconception of what is really happening versus mathematics.

Would you elaborate?

Are you saying that the source of the sound is not a quarter wavelength behind the voice coil, or my interpretation of what that means is not accurate?

 

[kelticwizard]

Therefore, the optimal driving impedance must be determined for each frequency and one must not forget the more important demands on the frequency response when doing so. This is very tricky, but for really carefullly designed speakers, one can reduce the distortion by applying the right driving impedance.

I was sort of thinking something along the same lines. In this era of inexpensive sub crossover/amps, the huge bump at resonance from a current driven amp need not be a deal-killer. It might be interesting for the bass region to be handled by a subwoofer driven by a voltage drive crossover/amp, then 150 Hz on up to be handled by a current driven amp, whether full range or crossed over to a tweeter.

That way the bump at resonance can be avoided entirely, if the midbass' Fc is low enough.

I suspect that a passive crossover to a tweeter, if one is used, might not resemble a passive crossover used with voltage drive.

The midbass on up is where most of the delicate detail is anyway, so the current driven setup would be used where it's advantages would be most apparent.

 

[Svante]

kelticwiz:

Yes, that would be the really advanced/elaborate way of doing it. On the other hand, there already is some room for experiments with distortion control in passive crossovers; consider the circuit below, it is a standard second order LP filter. The driving impedance that the tweeter sees has a peak at the resonance between L and C. This is a potential opportunity to reduce distortion al low frequencies (for the tweeter, ie 1-2kHz)´if the source of distortion is a non-linear voice coil inductance.

An externally hosted image should be here but it was not working when we last tested it.

 

[soongsc]

Would you elaborate?

Are you saying that the source of the sound is not a quarter wavelength behind the voice coil, or my interpretation of what that means is not accurate?

For example, we refer to the source of a signal to be the out of an amplifier. With speakers, the source is actually at the face where the diaphragm pushes the air. For purely analysis purposes, it is sometimes convenient assume an imaginary point source, which is good for deriving theories etc. In reality, due to diaphragm deformation, even for one given frequency, you could end up needing multiple sources to correctly model what is happening in reality. For design work, it is benificial to pick only one point as the signal source, which is usually determined by determining the time of flight when looking at a minimum phase plot. From test results, I have found that the plane where the VC cylinder mates with the diaphragm is a very good first guess reference.

 

[Jmmlc]

Hello soongsc,


In a first approach it is convenient to use the timeflight and this leads to take the origin of the sound waves from a loudspeaker around the voice coil. OK

It is now very easy using a microphone, a sound card and a personnal computer to time align a multiways system with pulses. Roughly this is done considering the timeflights of the different ways.

Timeflight is commonly measured by the arrival of a pulse,but rememeber that doing this we give more importance to the high frequency content as generally the high frequency arrive before the mow frequency of a same driver (due to crossover, acoustically reactive load...).

But what Drisko found is a bit different, its method based on phase figure at different frequency shows that a common point could eventually be precisely found for the origin of the different frequencies (remember he used an annular tweeter) but this common origin is nowhere near the coil!

What puzzled me is that Harvey replied that this is normal (but never considered) and explained that in the mass control region the waves are in quadrature with the signal at every frequency!

So the question is now: did Harvey is wrong or not?

Best regards from Paris,

Jean-Michel Le Cléac'h

 

[kelticwizard]

I have a question about building a practical current driven setup.

To briefly review a voltage driven setup, a 20 watt amplifier consists of + / - 10 volt rails with a 0 V ground. The current is 3 amps or so. The speaker is 8 ohms. This is just a rough setup.

In the discussion so far, current drive has been achieved by taking a 500 or 1,000 ohm reistor and placing it in series with the 8 ohm loudspeaker. I did a simulation on this using Subwoofer simulator, and see that this cuts sensitivity down to 65 dB @ 2.83 V from 89 dB.

I assume the resistor method has been used to convert a voltage drive to current drive for illustration purposes.

I would hope there is a current drive setup with an efficiency near the voltage drive setup.

If so, can anyone give a quick outline of such a current drive setup, (amp and speaker), and what the voltage and current of the rails would be, and what the optimum ohmage of the speaker would be?

 

[soongsc]

Hello soongsc,


In a first approach it is convenient to use the timeflight and this leads to take the origin of the sound waves from a loudspeaker around the voice coil. OK

It is now very easy using a microphone, a sound card and a personnal computer to time align a multiways system with pulses. Roughly this is done considering the timeflights of the different ways.

Timeflight is commonly measured by the arrival of a pulse,but rememeber that doing this we give more importance to the high frequency content as generally the high frequency arrive before the mow frequency of a same driver (due to crossover, acoustically reactive load...).

But what Drisko found is a bit different, its method based on phase figure at different frequency shows that a common point could eventually be precisely found for the origin of the different frequencies (remember he used an annular tweeter) but this common origin is nowhere near the coil!

What puzzled me is that Harvey replied that this is normal (but never considered) and explained that in the mass control region the waves are in quadrature with the signal at every frequency!

So the question is now: did Harvey is wrong or not?

Best regards from Paris,

Jean-Michel Le Cléac'h

I think that if you look at the measured phase of any driver, you will find that below a certain frequency, there is actually a phase lead, which is pretty much consistent in linear high pass systems as we know it. If the same thing happens purley in an electrinc circuit, we would not say the source has changed with frequency. Without looking at specific measured data for a specific driver, so many things in drivers can influence the response such that cannot be covered in a general discussion. I could add some stuff on certain areas of the cone, and you can see the phase diagram swing a certain way in far field measurements with almost no change in the frequency, and we would call that a change in acoustic center?

I'm really don't get into the who's right or wrong issue, but just discuss what seems reasonable. From a comparitive point of view, if you assume a specific receiving point to have the same minimum phase value for all frequencies, then you will come to a conclusion that different frequencies will have different acoustic centers.

If we take other XOs and electronic circuits into the picture, it becomes more complicated and I would think this is not close to the subject of this thread. But I recall that Nelson Pass did a very good explanation and simulation of the effects of voltage source versus current source, I'm sure it can be found at his site.

 

[Svante]

I would hope there is a current drive setup with an efficiency near the voltage drive setup.

If so, can anyone give a quick outline of such a current drive setup, (amp and speaker), and what the voltage and current of the rails would be, and what the optimum ohmage of the speaker would be?

I posted such a circuit on the first page of this thread:

An externally hosted image should be here but it was not working when we last tested it.

Adding an amplification of the feedback signal would decrease the sensitivity of the circuit (it might be too high in this configuration).

 

[gpapag]

An easy experiment (Merits of Variable Output Impedance Amplifiers)

Hi all
I have asked myself many times why I enjoy so much listening to music through my 6L6 push-pull no global-feedback amplifier.
To be more specific, why is it so that I prefer listening to this amplifier and not so much any other transistor or MOSFET output or chip amplifier (all with global voltage feedback) that I have built?
Well, I could have built the same tube amplifier WITH global voltage feedback to compare, but I have not.
I understand that bringing such a subject in the forum in such a blatant way, is like asking for trouble.
There are myriad of variables involved and each one has to be identified and treated in isolation.
But my queered mind stuck in one of them: Amplifier’s Output Impedance.
Why?
There is a gross difference in this specification between this tube amp (Rout~13 Ohm)and the rest (Rout~0.01 Ohm).
Resulting Damping Factor with 6 Ohm load is ~0.46 for the tube and ~600 for the rest.

I had to build two amplifiers which should differ only in their output impedance to find out.
Rod Elliot’s Project 56 (http://sound.westhost.com/project56.htm )provides a method to vary the output impedance of an amplifier through a combination of voltage and current feedback.
I set-up two bare basic LM1875 amplifiers as shown in attachment.

 

[gpapag]

They both have the same gain.
Quick and dirty implementation seen in attachment.
I listened to them in mono. Sound-wise they possess the difference I perceive when comparing the sound of my tube amp and the rest of my amps.

 

[gpapag]

Then I modified a stereo LM1875 amplifier (schematic attached)
Measurements correlate very well to the simulated data (very basic simulation with an old Electronic WorkBench).
Increasing the input signal, measured Rout decreases.
I am listening to this modified amplifier for four days now through my Jordan JX92S (in mass loaded quarter wave enclosure). Very pleasing sound.
Unmodified (voltage feedback only), this amplifier was producing a very anemic sound with these speakers.
I am in the process of modifying a more powerfull amplifier (my GB150D) accordingly.

Regards
George

 

[gpapag]

There are some very interesting articles in “Audio Anthology” (Audio Amateur Press)related to the synergetic effects of amplifier Output Impedance and Speaker Load.
In Audio Anthology Volume Two:
“How Far Can I Mismatch?” by Saul J. White.
“A New Approach to Loudspeaker Damping” by Warner Clements
“It’s Positive Feedback” by Warner Clements
In Audio Anthology Volume Three:
“Zero Impedance Output Stage” by Raymond G. Anthes
“Feedback and Loudspeaker Damping” by John A. Mulvey
In Audio Anthology Volume Four:
”What’s all this About Damping?” by N.H. Crowhurst

Regards
George

 

[Svante]

Re: An easy experiment (Merits of Variable Output Impedance Amplifiers)

Hi all
I have asked myself many times why I enjoy so much listening to music through my 6L6 push-pull no global-feedback amplifier.
To be more specific, why is it so that I prefer listening to this amplifier and not so much any other transistor or MOSFET output or chip amplifier (all with global voltage feedback) that I have built?
Well, I could have built the same tube amplifier WITH global voltage feedback to compare, but I have not.
I understand that bringing such a subject in the forum in such a blatant way, is like asking for trouble.
There are myriad of variables involved and each one has to be identified and treated in isolation.
But my queered mind stuck in one of them: Amplifier’s Output Impedance.
Why?
There is a gross difference in this specification between this tube amp (Rout~13 Ohm)and the rest (Rout~0.01 Ohm).
Resulting Damping Factor with 6 Ohm load is ~0.46 for the tube and ~600 for the rest.

I had to build two amplifiers which should differ only in their output impedance to find out.
Rod Elliot’s Project 56 (http://sound.westhost.com/project56.htm )provides a method to vary the output impedance of an amplifier through a combination of voltage and current feedback.
I set-up two bare basic LM1875 amplifiers as shown in attachment.

The reason you hear a difference between these two amplifiers is the output impedance, combined with the speaker's varying impedance. The combination will affect the frequency response, and that is the most audible effect.

I am not surprised that you perceive the sound more like the tube amplifier, but I am surprised that you prefer the high output impedance. This is because most speakers are designed for a 0 ohm driving impedance.

 

[gpapag]

Hi Svante

..but I am surprised that you prefer the high output impedance. This is because most speakers are designed for a 0 ohm driving impedance.

From my limited experience:
Multispeaker units with complex cross-over circuits do match well with nearly 0 Ohm driving impedance, but again not always.
Single unit, no cross-over speakers are breathing better with a positive driving impedance.
With open baffle and rear horn loading it is more evident.
Least evident is with small volume sealed and small volume bass reflex.
Thus, how much high output impedance gives best results, depends on speaker unit and enclosure type.
Also on kind of music to be reproduced.
Rock, techno, rap are best with very controlled mid bass and good attack (low driving impedance).
Symphonic, opera, need extended freq. response, balance, space. For this, higher driving impedance is beneficial.

(I have settled upon my (sad) conclusion that I have yet to hear a system that performs very well with any kind of music).

With the help of Rod Elliot's article, it is feasible to test how much driving impedance is good for everyone's case.

Regards
George

 

[soongsc]

In various books, when people talked about impedance matching, it is always illustrated with a series resistor with load of the same value. So I tend to think that the transconductance nature of the source and the nature of the load all need to be matched in a certain way.

To make things more complicated, just recently playing around with different transformers in a class D amplifier made significant differences, aspecially in the low frequencies. So rather than trying to just let the ear decide, there can be a technical explanation in different stages of the audio chain that effect what we hear. What we really can try to do is sort each stage out and technically improve them one at a time an listen to the improvements.

 

[Svante]

In various books, when people talked about impedance matching, it is always illustrated with a series resistor with load of the same value. So I tend to think that the transconductance nature of the source and the nature of the load all need to be matched in a certain way.

To make things more complicated, just recently playing around with different transformers in a class D amplifier made significant differences, aspecially in the low frequencies. So rather than trying to just let the ear decide, there can be a technical explanation in different stages of the audio chain that effect what we hear. What we really can try to do is sort each stage out and technically improve them one at a time an listen to the improvements.

Well, those books on impedance matching probably (hopefully) don't deal with the amplifier-loudspeaker interface, do they?

It is true that the output power from a source is maximized if the driving and load impedances are equal. In the case of sound, however there are several things that makes that rule hopelessly useless. For example:

- Almost all loudspeakers are designed for 0 ohms of driving impedance. Using another driving impedance would result in frequency response issues that should be bad if the loudspeaker designer has done his job properly.

- Most amplifiers have an inherent output impedance that is much higher than the effective output impedance with feedback applied. So if the amplifier has a "damping factor" of 100, the effective output impedance is 8/100=0.08 ohms. In practice, the amplifier without feedback would have an output impedance of maybe half a ohm or so, but feedback lowers the apparent impedance. However, feedback cannot eliminate the fact that the voltage source will have to overcome the voltage drop over the real output impedance. This complicates the maximizing of output power, it cannot be determined by the closed loop output impedance only.

- Most amplifiers have a maximum output voltage and a maximum output current. These have nothing to do with the output impedance, but one of them will limit the maximum output power. Actually, maximizing output power is a question of knowing these limits and U/I equals the optimum load impedance, if max power is all that counts.

Finally I agree that each step in the chain should be optimized, and usually frequency response is the number one factor to take care of. Listening tests can be useful, but it must be kept in mind that FR issues will effectively mask other differences in most cases if they exist. So, if one setup is preferred over the other, first look at the FR changes and see if they can explain the audible differences. Actually, differences as small as tenths of a dB can explain much.

And output impedance of amplifiers is a typical example of where the FR is affected and where these FR changes explain most of the audible differences.

There is nothing magic about it, and an eq could do the same thing.

 

[soongsc]

Then we would have more problems with speaker cables than most people think, because speaker cable characteristic impedance is much higher than most speaker drivers. Additionally, when an XO is added, the additional impedance is much more than the amplifier output impedance, which normally isn't even published for the whole audio spectrum.

Of the speaker cable specs that I have seen, only some Alpha Core cables come close to the low impedance of most speakers have.

So maybe current source output to some degree does have some advantages? I was hoping to find some kind of resolution for better phase response in wide range drivers above 10KHz.

 

[Svante]

Then we would have more problems with speaker cables than most people think, because speaker cable characteristic impedance is much higher than most speaker drivers. Additionally, when an XO is added, the additional impedance is much more than the amplifier output impedance, which normally isn't even published for the whole audio spectrum.

Of the speaker cable specs that I have seen, only some Alpha Core cables come close to the low impedance of most speakers have.

So maybe current source output to some degree does have some advantages? I was hoping to find some kind of resolution for better phase response in wide range drivers above 10KHz.

The characteristic impedance of a cable is the resistance one sees into an infinitely long cable. It is important to know this impedance if the cable is significantly longer than the (electromagnetic) wavelength in the cable, which in case of a 20 kHz audio signal is roughly 200 000 000 / 20 000 = 10 000 metres . Obviously the characteristic impedance is only important* when frequencies very much higher than those of audio are involved, such as in data communication. For shorter lengths, what matters is parallel capacitance, series inductance and series resistance. The series resistance, which is something other than characteristic impedance, can be measured with a multimeter; to do this, short the remote end and measure between the wires of the near end and you will typically find a resistance in the 0,1-0,5 ohm range or less.

If it is a copper wire you can calculate it as Rs=2*L*0,017/S, where L is the length in metres and S is the area in mm². You can also approximate how big effect this series resistance has on the response in decibels by L=20*log(Zmin/(Zmin+Rs)) where Rs is the series resistance and Zmin is the minimum impedance of the speaker.

So, for 5 metres of 0.75 mm² wire, the series resistance would be 0.23 ohms, and if the speaker has a minimum impedance of 3 ohms, the frequency response would be altered by 0.6 dB.

In my opinion, phase response at higher frequencies is completely unimportant except near the crossover frequency where more than one driver contributes to the sound. If you hear differences between systems with current vs voltage drive, you will find the explanation in the frequency response, or possibly but far less likely in the distortion.

*If the amplifier has a tendency to oscillate im the MHz region, cable impedance, or rather termination can be important. Well designed amplifiers do not have this tendency, however, but this explains why there may be audible differences between cables, if poor amplifiers are used.

 

[soongsc]

Svante, I'm there is another thread that talks about cable differences. I do think each different cable may have differences. For whatever technical reason, we probably should discuss it in that thread. But I think the impedance effects does influence the result even though it may seem very small. It's only a matter of how small until the effects are not audible.

Phase response as well as frequency response are equally important. We have been able to identify many good qualities associated with both.

I am also quite interested in how driver back EMF effects current source versus voltage source amplifiers.