Current drive for Loudspeakers

Jump to Latest
#201
The second unused voice coil represent its displacement (magnetic and coil has ideally constant displacement in coil voltage function, this principle also used in hard disk head control), and in loudspeaker it is normally distorted mostly in box resonance (acoustic) and free air loudspeaker resonance (mass+spring)

Moving masses has its own energy, also the spring, that is why the current is distorted like that.

Something difficult is that loudspeaker displacement distortion may different from acoustic distortion produced.
 
#202
ontoaba said:
The second unused voice coil represent its displacement (magnetic and coil has ideally constant displacement in coil voltage function, this principle also used in hard disk head control), and in loudspeaker it is normally distorted mostly in box resonance (acoustic) and free air loudspeaker resonance (mass+spring)

Moving masses has its own energy, also the spring, that is why the current is distorted like that.

Something difficult is that loudspeaker displacement distortion may different from acoustic distortion produced.
Click to expand...
The second voice coil voltage represents its velocity + some induced voltage introduced by the transformer effect with the first one.
 
#203
inventor said:
Many thanks for your rapid reply.



I agree with your comments emboldened above.

However, I disagree with your comment:



My understanding is that Vs (purple traces in the scope pictures) represents a combination of both motional back-EMF voltage and the voltage induced in the second (sense) coil by the current flowing in the first coil due to the transformer action (i.e. the magnetic coupling) between the two coils. In other words, the sensed voltage Vs is NOT equal to the motional back-EMF.

I am ignoring the transformer effect. It could be measured if you can immobilize the coil motion. The transformer effect is not interesting in and of itself.


If my understanding above is correct, your comment does not properly answer my question 1 (what is the difference, why is there a difference, between Vd and Vs?)

I believe my answer substantially answers your question.

As a consequence of this disagreement, I also disagree with your comment below as an answer to my question 2:

So be it. I am measuring stuff in order to be able to calculate the non-linear characteristics of BL, Cms, Le, etc.

(What you are stating above may be correct but, I think, is not the full or correct answer to my question 2.)

I understand that, in the voltage-driven case that we are discussing with an unloaded/undriven second sense-coil, the current in the first drive-coil does inevitably contain distortion (as evident in the scope shots). Likewise, the motional back-emf portion of the voltage induced in the second sense-coil also contains distortion (owing to nonlinear Blv). The combination of these two should result in a highly distorted waveform

There is an additional distortion. The voice coil inductance is changing with excursion. This may be evident before the BL and/or Cms gets too non-linear. I have measured the distortion and it is 2nd and 3rd mostly. At full power, H1, H2 and H3 are all pretty much equal in the current waveform.

for Vs, the voltage induced in the second sense-coil. However, such distortion is almost invisible in all the purple (Vs) traces in your scope shots (except for a tiny bit in the first scope shot). WHY?

Good question.

I agree with your last comment:



However, this still does not answer the question "WHY?" at the end of the paragraph above. If we look at your second scope shot (below):



the current in the drive-coil (blue trace) is still highly distorted (although not so much as in the maximum-excursion case). Nevertheless, the sensed voltage Vs (purple trace) has quite an undistorted waveform. WHY? Isn't there the transformer action, i.e. mutual coupling, between the two coils? How come does the highly-distorted current in the driven coil NOT affect the voltage induced in the second sense-coil?

Besides, the magnitude difference between the drive voltage Vd (yellow trace) and the sensed voltage Vs (purple trace) seen in the scope shots still remains unexplained.

When I have a perfect model from rest to Xmax and can measure all the non-linear parameters I will let you know.

My favorite quote is by Lord Kelvin:
"If you can measure something you can claim to know something about it....."

Thank you for your attention.
Click to expand...
 
#204
the transformer effect. It could be measured if you can immobilize the coil motion. The transformer effect is not interesting in and of itself.
Click to expand...
Dan Ferguson in his servo-sub project in AudioXpress circa 2003 studied this transformer effect by clamping two DVC drivers cone to cone à la "isobarik". The very small air gap between the cones allows to consider they have the same movement, at least in the frequency zone in interest in the current context. One of the coil of the first driver is driven by an amp, the signal of the other one is compared to the signal of one of the voice-coils of the second driver. This second driver is not driven at all and the signal delivered by its voice coil does not suffer from any transformer effect.
 
Last edited:
#205
forr said:
The second voice coil voltage represents its velocity + some induced voltage introduced by the transformer effect with the first one.
Click to expand...

Yes that is, the difference is the transformer placed in high magnetic field, when you move the coil, it has voltage between coil wires, when coil wires are short circuited the coil won't move as easy as before.

Actually, loudspeaker distortion and inefficiencies are mostly coming from mechanical_vibration -> sound_waves conversion, only small in electric -> mechanic conversion except it is in resonance.
 
#206
forr said:
Dan Ferguson in his servo-sub project in AudioXpress circa 2003 studied this transformer effect by clamping two DVC drivers cone to cone à la "isobarik". The very small air gap between the cones allows to consider they have the same movement, at least in the frequency zone in interest in the current context. One of the coil of the first driver is driven by an amp, the signal of the other one is compared to the signal of one of the voice-coils of the second driver. This second driver is not driven at all and the signal delivered by its voice coil does not suffer from any transformer effect.
Click to expand...

This is a very clever trick and will work for small displacements. Several problems arise.

1. The magnetic field at the bottom of the pole piece (back plate) is smaller than the field at the top of the stroke. The cones being front-to-front means that the driving woofer magnetic field is strong when the receiving woofer magnetic field is weak.

2. The spiders of the two drivers are not identical. X amount of push by the driving woofer will not result in X amount of movement by the receiving woofer.

3. The magnets will not be identical.

A much better idea is to immobilize the cone and measure transformer effect directly.
 
#207
ontoaba said:
Yes that is, the difference is the transformer placed in high magnetic field, when you move the coil, it has voltage between coil wires, when coil wires are short circuited the coil won't move as easy as before.

Actually, loudspeaker distortion and inefficiencies are mostly coming from mechanical_vibration -> sound_waves conversion, only small in electric -> mechanic conversion except it is in resonance.
Click to expand...

Can you provide a citation for this assertion?
 
#208
I am afraid not, since there are only common test ever published, but if you try it your self, just catch some trees insect from the wood, they playing so loud hearable to more than 50meters in distance. Record it and reproduced sound with similar loudness needs 6watts in high efficient loudspeaker, I am sure this small insect doesn't have even 1watt power.
The sound reproduced also far different, the insect sounded much better and real.

Edit Some common test.
Just common test of loudspeaker comparison like this is available in records.
Speaker Efficiency: An Initial Review Article By Peter W. Mitchell
http://www.sengpielaudio.com/calculator-efficiency.htm
 
Last edited:
#209
ontoaba said:
I am afraid not, since there are only common test ever published, but if you try it your self, just catch some trees insect from the wood, they playing so loud hearable to more than 50meters in distance. Record it and reproduced sound with similar loudness needs 6watts in high efficient loudspeaker, I am sure this small insect doesn't have even 1watt power.
The sound reproduced also far different, the insect sounded much better and real.

Edit Some common test.
Just common test of loudspeaker comparison like this is available in records.
Speaker Efficiency: An Initial Review Article By Peter W. Mitchell
Efficiency and sensitivity conversion - loudspeaker percent and dB per watt and meter loudspeaker efficiency versus sensitivity vs speaker sensitivity 1 watt = 2,83 volt box - sengpielaudio Sengpiel Berlin
Click to expand...

High efficiency speaker are notoriously non-linear. They usually have very short voice coils and distort with any excursion. Horns are even worse. Frequency and phase response are not good at all. I have measured these kinds of speakers.

I do not think that your observations support your conclusion about acoustic cause of current distortion.

My measurements of current distortion are of a heavy cone woofer suspended in free air at resonance. I doubt that there is any appreciable acoustic coupling to create current distortion.
 
#210
woofertester said:
This is a very clever trick and will work for small displacements. Several problems arise.

1. The magnetic field at the bottom of the pole piece (back plate) is smaller than the field at the top of the stroke. The cones being front-to-front means that the driving woofer magnetic field is strong when the receiving woofer magnetic field is weak.

2. The spiders of the two drivers are not identical. X amount of push by the driving woofer will not result in X amount of movement by the receiving woofer.

3. The magnets will not be identical.

A much better idea is to immobilize the cone and measure transformer effect directly.
Click to expand...

I made my own tests with two DVC drivers, but did not want sacrifice one of them by gluing the voice coil in the gap.

I also got some data with a naked dual voice coilbut it would have been really informative only if I could have compared them with a dual voice coil immobilized in the magnetic gap.

Using a common DVC in a velocity servo experimental project is quite interesting for didactic purposes. The great drawback is the great loss of efficiency. There was such a project by Russel Breden which has been published in Electronics World : jeff macaulay's audio amp designs

I am pretty sure that the DVC driver used in the famous Mills and Hawksford's study is not a common DVC driver but a specially made one where the velocity sensing voice coil was a very small additional voice-coil just as it is in the Ryhtmik drivers for subs. It is probably the easiest and best way to build servo-ed loudspeakers.

In his AudioXpress servo project, Dan Ferguson seems to have finally thought that the transformer effect is not worth to be compensated for if the driver range is limited to the bass frequencies.
 
#211
Every time I look at this forum, I learn something new. Thanks to all the people who post messages here, especially those who provide information and have spent certain effort towards the content of their messages.

In this respect, thanks to "woofertester" for posting the oscilloscope screen shots that show the waveforms of 1- the voltage and current of the driven voice-coil (Vd), 2- the voltage sensed from the undriven voice-coil (Vs) of a Dual-Voice-Coil (DVC) subwoofer .

The previous message from "forr" (message #210) indicates to me the relevance of a sense coil (either a specially-made sense coil or using one of the two coils of a DVC driver as a sense coil) to the present topic. It is because of this relevance that I am interested in a correct interpretation of the Vs waveforms posted by woofertester.

I like the quote from Lord Kelvin:
"If you can measure something you can claim to know something about it....."
Click to expand...
Nevertheless, I think that we must firstly know what it is that we are measuring. Otherwise, what we are measuring may not actually be what we think we are measuring.

My understanding is that the voltage across the sense coil is a combination of 1- motional back emf and 2- the voltage induced by the driven coil through transformer effect. Both of these voltage components are likely to be in the same order of magnitude.

Hence, why ignore the transformer-induced voltage component:

woofertester said:
I am ignoring the transformer effect.... The transformer effect is not interesting in and of itself.
Click to expand...

or why disregard it in any compensation scheme:

forr said:
Dan Ferguson ... thought that the transformer effect is not worth to be compensated for if the driver range is limited to the bass frequencies.
Click to expand...

I think that the problem is in separating these two voltage components from within the single voltage measured across the sense coil.

I have looked at the Russel Breden's paper:
forr said:
which has been published in Electronics World : jeff macaulay's audio amp designs
Click to expand...

where the sense-coil voltage is fed back after being processed by R22, R24 and C19 (in Figure 1 of the paper). The paper states that the manipulation of the sense-coil voltage in this way produces motional feedback.

If I understood it correctly, the paper claims that the filter formed of R22, R24 and C19 separates the 'motional back emf' voltage component from the transformer-induced voltage component, both of which exist simultaneously in the sense-coil voltage. I fail to understand how this is achieved.

I have also looked at the "Hawskford & Mills" paper (Distortion Reduction in Moving-Coil Loudspeaker Systems Using Current-Drive Technology), which also uses "motional back emf" feedback with a "current-drive" amplifier (see Section 3.2 of the paper).

In this paper, there seems to be a better attempt to extract the "motional back emf" component from the sense-coil voltage. However, the paper does not make it crystal clear how this is achieved, e.g. the contents of the "coupling error compensator" block in Fig. 18 of the paper is not disclosed. Furthermore, the analysis of the system is based on a simplified model (in Fig.19 of the paper), for which the authors state: "No transformer coupling effects were included in the model.". This approach does not make sense to me.

In my next message, I will try to explain the discrepancy that exists between the sense-coil voltage (Vs) waveforms posted by 'woofertester' and my mathematical understanding of a DVC driver. Surely, I must be wrong somewhere in my understanding. If somebody points out where I am going wrong, this may help all the interested people to achieve a better understanding of "motional feedback" with "current-drive" amplifiers.

In the meantime, can anybody point me to a patent (or patents) that Rhytmik Audio Inc claim their patented "servo" technology is based on? (Without knowing the title or the inventor's name, a search on patents granted or assigned to Rhytmik Audio yields no result.)

Finally, I have just noticed the website Servobass, which claims to combine current-drive amplifier and motional feedback (using an accelerometer device rather than sense-coil) via high-speed (ADI Sharc) DSP technology.

Thanks for your attention.
 
#212
Here's the transformer effect I measured on an out-of-gap dual voice-coil :
Uin = voltage across the driven voice coil.
Uout = voltage across the sensing voice coil.

Code: 
Frequency   Uin       Uout      Uout/Uin in dB

20          4,000     0,012     -50,5
25          4,000     0,015     -48,5
32          4,000     0,020     -46,0
40          4,000     0,025     -44,1
50          4,000     0,033     -41,7
64          4,000     0,040     -40,0
80          4,000     0,050     -38,1
100         4,000     0,066     -35,7
125         4,000     0,080     -34,0
160         4,000     0,100     -32,0
200         4,000     0,130     -29,8
250         4,000     0,163     -27,8
320         4,000     0,210     -25,6
400         4,000     0,260     -23,7
500         4,000     0,330     -21,7
640         4,000     0,420     -19,6
800         4,000     0,520     -17,7
1000        4,000     0,650     -15,8
1250        4,000     0,760     -14,4
1600        4,000     0,940     -12,6
2000        4,000     1,200     -10,5
2500        4,000     1,480      -8,6
3200        4,000     1,800      -6,9
4000        4,000     2,160      -5,4
5000        4,000     2,560      -3,9
6400        4,000     2,800      -3,1
8000        4,000     3,200      -1,9
10000       4,000     3,500      -1,2
12500       4,000     3,600      -0,9
16000       4,000     3,810      -0,4
20000       4,000     3,940      -0,1
 
#214
Hi, here are some measurements I have conducted using current drive and voltage drive. Current drive is realized with 1:4 step-up transformer and ~66 ohms series resistance on the secondary side of the trafo in series with the driver. The amp is a regular high-biased, balanced voltage source amplifier.

Measurements are done with Scarlett 2i2 and EMC8000, at 1 meter away from the driver unless otherwise stated. The box is 160 litres, tuned to 20Hz.

Click on the links to open the plot and see the higher order harmonic reduction.

Current drive sounds so good and liquid compared to voltage drive, the difference is drastic!


Eminence Beta 12LTA:

Frequency Hz
Level dB: Voltage Drive THD % / Current Drive THD %


50Hz
90dB: Voltage Drive 12,4% / Current Drive 8,73%


100hz
90dB: Voltage Drive 1,81 % / Current Drive 0,978 %
95dB: Voltage Drive 3,65 % / Current Drive 2,01 %

200Hz
90dB: Voltage Drive 0,375 % / Current Drive 0,252 %
95dB: Voltage Drive 0,674 % / Current Drive 0,487 %

400Hz
90dB: Voltage Drive 0,181 % / Current Drive 0,0644 %
100dB: Voltage Drive 0,360 % / Current Drive 0,111 %

800Hz
90dB: Voltage Drive 0,202 % / Current Drive 0,0541 %
95dB: Voltage Drive 0,296 % / Current Drive 0,0665 %
100dB: Voltage Drive 0,572 % / Current Drive 0,116 %

1600Hz
90dB: Voltage Drive 0,485 % / Current Drive 0,197 %
95dB: Voltage Drive 0,879 % / Current Drive 0,256 %
100dB: Voltage Drive 1,78 % / Current Drive 0,414 %

3200Hz
90dB: Voltage Drive 0,401 % / Current Drive 0,201 %
95dB: Voltage Drive 0,758 % / Current Drive 0,387 %

6400Hz
90dB: Voltage Drive 0,092 % / Current Drive 0,0968 %
95dB: Voltage Drive 0,175 % / Current Drive 0,184 %
100dB: Voltage Drive 0,357 % / Current Drive 0,314


Celestion K12H-200TC

Measured at 50cm:

100Hz@95dB:
Voltage Drive 0,655% / Current Drive 0,608%

150Hz@100dB:
Voltage Drive 0,237% / Current Drive 0,176%


Measured at 100cm:

200Hz@100dB:
Voltage Drive 0,335% / Current Drive 0,124%

400Hz@100dB:
Voltage Drive 0,190% / Current Drive 0,0683%

800Hz@100dB:
Voltage Drive 1% / Current Drive 0,163%

1600Hz@100dB:
Voltage Drive 0,578% / Current Drive 0,220%

3200Hz@100dB:
Voltage Drive 0,374% / Current Drive 0,086%

6400Hz@100dB:
Voltage Drive 0,477% / Current Drive 0,436%
 
Last edited:
  • Like
Reactions: Freedom666
#216
inventor said:
Every time I look at this forum, I learn something new. Thanks to all the people who post messages here, especially those who provide information and have spent certain effort towards the content of their messages.

In this respect, thanks to "woofertester" for posting the oscilloscope screen shots that show the waveforms of 1- the voltage and current of the driven voice-coil (Vd), 2- the voltage sensed from the undriven voice-coil (Vs) of a Dual-Voice-Coil (DVC) subwoofer .

The previous message from "forr" (message #210) indicates to me the relevance of a sense coil (either a specially-made sense coil or using one of the two coils of a DVC driver as a sense coil) to the present topic. It is because of this relevance that I am interested in a correct interpretation of the Vs waveforms posted by woofertester.

I like the quote from Lord Kelvin: Nevertheless, I think that we must firstly know what it is that we are measuring. Otherwise, what we are measuring may not actually be what we think we are measuring.

My understanding is that the voltage across the sense coil is a combination of 1- motional back emf and 2- the voltage induced by the driven coil through transformer effect. Both of these voltage components are likely to be in the same order of magnitude.

Hence, why ignore the transformer-induced voltage component:

This is thread about current drive of loudspeakers. If you want to start a thread about transformer effect of DVC woofers, please do so and copy the responses above. I will also add to the thread when I can freeze my monster woofer and measure it. I can freeze it, but I have to fabricate a device and do not have a machine shop.

or why disregard it in any compensation scheme:

Because I am researching current drive of loudspeakers, not DVC transformer effects. Most woofers have one vice coil and I want to develop a model for behavior at large excursions.

I think that the problem is in separating these two voltage components from within the single voltage measured across the sense coil.

Yes, it is called "do the math".

I have looked at the Russel Breden's paper:


where the sense-coil voltage is fed back after being processed by R22, R24 and C19 (in Figure 1 of the paper). The paper states that the manipulation of the sense-coil voltage in this way produces motional feedback.

These systems that use the second coil are notoriously unstable and self-destruct. Infinity made such a system in the late 70s. I doubt that you can find one alive today.

If I understood it correctly, the paper claims that the filter formed of R22, R24 and C19 separates the 'motional back emf' voltage component from the transformer-induced voltage component, both of which exist simultaneously in the sense-coil voltage. I fail to understand how this is achieved.

I have also looked at the "Hawskford & Mills" paper (Distortion Reduction in Moving-Coil Loudspeaker Systems Using Current-Drive Technology), which also uses "motional back emf" feedback with a "current-drive" amplifier (see Section 3.2 of the paper).

The amp is complicated. Not interesting.

In this paper, there seems to be a better attempt to extract the "motional back emf" component from the sense-coil voltage. However, the paper does not make it crystal clear how this is achieved, e.g. the contents of the "coupling error compensator" block in Fig. 18 of the paper is not disclosed. Furthermore, the analysis of the system is based on a simplified model (in Fig.19 of the paper), for which the authors state: "No transformer coupling effects were included in the model.". This approach does not make sense to me.

I doubt that the system is great at extremely high SPL.

In my next message, I will try to explain the discrepancy that exists between the sense-coil voltage (Vs) waveforms posted by 'woofertester' and my mathematical understanding of a DVC driver. Surely, I must be wrong somewhere in my understanding. If somebody points out where I am going wrong, this may help all the interested people to achieve a better understanding of "motional feedback" with "current-drive" amplifiers.

Please do this in a separate thread. We are discussing current drive here. I plan to post some measurements when I have a proper amp. I also have the facility to attach an accelerometer and plan to show excursion vs current for current drive and voltage drive. This is the holy grail (to me) that will allow direct measurement of BL vs excursion.

In the meantime, can anybody point me to a patent (or patents) that Rhytmik Audio Inc claim their patented "servo" technology is based on? (Without knowing the title or the inventor's name, a search on patents granted or assigned to Rhytmik Audio yields no result.)

Start a separate thread. Try the subwoofer forum.

Finally, I have just noticed the website Servobass, which claims to combine current-drive amplifier and motional feedback (using an accelerometer device rather than sense-coil) via high-speed (ADI Sharc) DSP technology.

Its a very good idea. I plan to have all of that working soon except for the SHARC. The ACH-01 is $33 from Digikey, My woofer is built so that I can attach an accelerometer. That is how I can also freeze the voice coil.

Thanks for your attention.
Click to expand...
 
#217
Legis said:
Hi, here are some measurements I have conducted using current drive and voltage drive.

Click on the links to open the plot and see the higher order harmonic reduction.

Current drive sounds so good and liquid compared to voltage drive, the difference is drastic!
Click to expand...

Hi,

What is the speaker parameter or reason one speaker distortion drops by 1/2 or so and the other driver by very little change in distortion??

When I applied my idea of series R thru fb path of amp ... the drop (at low freqs) was also about 1/2 or more.
And, I agree, the sound was a lot better.


THx-RNMarsh
 
Last edited:
#218
RNMarsh said:
Hi,

What is the speaker parameter or reason one speaker distortion drops by 1/2 or so and the other driver by very little change in distortion??

When I applied my idea of series R thru fb path of amp ... the drop (at low freqs) was also about 1/2 or more.
And, I agree, the sound was a lot better.


THx-RNMarsh
Click to expand...

The back emf is out of phase with the driving voltage. The low amp output impedance shorts this voltage into significant current which drives the voice coil out of phase compared the the input signal into the amp.

The back emf is worse than out of phase. It is created by non-linear BL and non-linear Cms.

Increasing the amp output impedance causes the back emf current to be reduced thereby directly reducing distortion in the driving current in the voice coil. If you make the amp output impedance infinite, you would reduce the back emf current distortion to zero. You would be left with the distortion caused by the non-linear BL and the non-linear Cms and all the other non-linearities, Re, Le etc.

It is ironic that amp designers strive for ppm distortion when the speaker distortions are mega-percent at appreciable excursion. The exception is, of course, a good push-pull electrostat. I lived with a pair of Martin Logan CLS for many years. They sounded like excellent headphones.

The most awesome damping factor amp causes the most back-emf distortion. (intended to elicit discussion)
 
Last edited:
#219
RNMarsh said:
Hi,

What is the speaker parameter or reason one speaker distortion drops by 1/2 or so and the other driver by very little change in distortion??

When I applied my idea of series R thru fb path of amp ... the drop (at low freqs) was also about 1/2 or more.
And, I agree, the sound was a lot better.


THx-RNMarsh
Click to expand...

Hi, I think depends on what's the cause of given distortion. There are so many mechanisms, from which current drive can only reduce some (quite many) of the causes of distortion. Other causes, for ex. suspension non-linearity and other mechanical stuff, it cannot reduce.

Also the impedance curve affects the result. The amount the current drive can reduce harmonics, is based on the magnitude how "strong" current drive is used. Rout + Rload / Rload, for example 66ohm + 8ohm / 8 ohm = 9,25:1 voltage ratio, meaning that the current drive can reduce harmonics approx. 19,3dB at the most. If/when the impedance curve of the driver is rising towards the HF, said voltage ratio decreases and the current drive cannot reduce harmonics as much. This might also explain some differences between drivers.

Also the material, for example, of the voice coil former, plays a role. The measurements I posted were made with drivers that both had non-conductive (Kapton) VC formers. Conductive former might affect how the driver responds to current drive.

There might be other factors as well. You should check out Esa Meriläinen's work: Current-Drive - The Natural Way of Loudspeaker Operation
 
Last edited:
You must log in or register to reply here.
Top Bottom