KEF's Breakthrough VECO Speaker Technology to Lower Distortion–Explained by Its Inventors

[Cbdb2]

Amplifier instability with arbitrary loads? The driver, amp, and velocity feedback loop are a tightly integrated system and aren't intended to be mucked with by the end user.

The load is constantly changing as the voice coil temp. changes, Especially at high power.

 

[Cbdb2]

Power compression as a means of protecting the driver? For a system like this, much better results can surely be had by incorporating even a basic thermal model and limiting mechanism.

So you replace the built in power compression protection with a electronic compressor that requires temperature feedback from the coil? Dosnt sound basic to me.

 

[bmc0]

The load is constantly changing as the voice coil temp. changes, Especially at high power.

Not nearly enough to cause stability problems.

So you replace the built in power compression protection with a electronic compressor that requires temperature feedback from the coil? Dosnt sound basic to me.

No real temperature measurement is required, only a fairly simple thermal model which can be trivially implemented in the digital domain. The voice coil temperature can be calculated approximately from the known input current, the real part of the impedance (which is somewhat dependent on temperature, of course), two or three time constants based on known thermal masses and conductivity, and possibly the coil velocity/excursion (for motors which rely on forced air cooling from diaphragm motion).

If necessary, it's possible to calculate the voice coil temperature directly from its DC resistance, which can (with some tricks) be measured in real time during music playback. No need for an additional sensor.

 

[Cbdb2]

Not nearly enough to cause stability problems.

You an expert on current drive feedback?

No real temperature measurement is required, only a fairly simple thermal model which can be trivially implemented in the digital domain. The voice coil temperature can be calculated approximately from the known input current, the real part of the impedance (which is somewhat dependent on temperature, of course), two or three time constants based on known thermal masses and conductivity, and possibly the coil velocity/excursion (for motors which rely on forced air cooling from diaphragm motion).

If necessary, it's possible to calculate the voice coil temperature directly from its DC resistance, which can (with some tricks) be measured in real time during music playback. No need for an additional sensor.

You sure make it sound easy. Have you ever designed a compressor/limiter? You have no idea what devils are in the details. Its still a compressor/limiter and will sound like one, why do you think it will sound better?

 

[bmc0]

You an expert on current drive feedback?

No, but I've toyed with it some and I know a bit about audio amplifier design. Stability considerations are outlined in the article you linked. If you read it, you should realize that it's not a problem for a soundbar or other powered speaker.

Have you ever designed a compressor/limiter?

Technically yes, although it was an exceedingly simple one a long time ago. Have you?

why do you think it will sound better?

My claim was that it would be easy to do a better job at protecting the driver than relying on power compression.

 

[Audionaut]

I think it's debatable whether the advantages of servo controlled drivers are "immense".

If the loudspeaker chassis is otherwise well designed, you have an additional control level with which you can modulate the signal as desired in real time.

At low frequencies, I think the advantages are immense, yes.

 

[Mnyb]

Could modelling help ? In my line of work with VSD drives , modern ones of any brand identify the electrical parameters by measuring them and construct a model based on that and the label data , a sort of equivalent circuit. Then during running you need only to measure two currents and three voltages and churn trough the mathematical model and get pretty darn close to the motors actuall physical status

 

[Timcognito]

Could modelling help ? In my line of work with VSD drives , modern ones of any brand identify the electrical parameters by measuring them and construct a model based on that and the label data , a sort of equivalent circuit. Then during running you need only to measure two currents and three voltages and churn trough the mathematical model and get pretty darn close to the motors actuall physical status

This kind of sounds like what was said in the video?

 

[Mnyb]

Watching again I also realise that’s there similarities to a speaker driver and transformers and motors and generators.

For example there commutation coils in old DC motors , there the field in a motor twist a bit with large currents and you get sparks in the brushes but it’s compensated by a an coil in series with the current but the field of it is positioned so that it undoes the effect . DC motors also have compound windings that compensate for the loss of field due to current in the rotor itself counteracting the field coils.

There where field coil drivers in the past . But wonder if any speaker driver ever had voice coil windings not on the moving part ( to counter modulate the field with the signal ) ?
Or if tricks with shorting rings and very well saturated fields makes such things unecessary.

On topic wonder if a future sucessor to the LS60 ( which I own, great speakers ) will have this technology going forward

 

[jackocleebrown]

This kind of sounds like what was said in the video?

Yes, that's right. We, and many others in the industry, have products that have a DSP model of the driver+cabinet and then calculate feed-forward correction to reduce the distortion in the radiated sound. The approach works quite well but there are some aspects of the driver behaviour that are difficult to predict. Some such systems are fixed at production and others are adaptive and monitor a state variable, such as voice coil current, and minimise the error between the internal driver model and the real behaviour. One of the motivations for the sensor was to superseded this approach, although the sensor can also be used to augment this approach by providing another state variable to the model.

 

[jackocleebrown]

For a system like this, much better results can surely be had by incorporating even a basic thermal model and limiting mechanism.

Yes, that's correct. As mentioned power compression is eliminated so cannot be used to provide any kind of natural thermal limiting. However, we do this in the DSP. A simple thermal model of the driver is enough. There are some nice benefits compared to relying on power compression. First that in the DSP we can apply no attenuation until actually required to protect the drivers. Second, we can apply attenuation that's flat over the spectrum of the driver, whereas thermal attenuation from Re increase isn't flat. Third, we can attenuate the whole system by the same amount so that you don't have a situation where the woofer is attenuated and the tweeter is not (or vica versa).

 

[Mnyb]

Yes, that's right. We, and many others in the industry, have products that have a DSP model of the driver+cabinet and then calculate feed-forward correction to reduce the distortion in the radiated sound. The approach works quite well but there are some aspects of the driver behaviour that are difficult to predict. Some such systems are fixed at production and others are adaptive and monitor a state variable, such as voice coil current, and minimise the error between the internal driver model and the real behaviour. One of the motivations for the sensor was to superseded this approach, although the sensor can also be used to augment this approach by providing another state variable to the model.

Very cool ,similarly when our VSD ” motor modell ” does not cut it for the application we attach pulse encoders to the drive shafts and get better precision.
We historically had opposite approach model based systems where designed to get rid of extra speed sensors if possible

 

[miha]

If I am not mistaken, the founder of B&M has still an operating company called Silbersand employing the very sensor technology they invented in the 1970s. B&M still uses a similar principle as well at least for subwoofers and bass drivers, and so do other manufacturers.

Interestingly Müller´s solution of an electrodynamic velocity sensor seemingly circumvents most of the problems KEF has claimed to solve in a pretty complicated manner. His sensor coil is not following the shape of the voicecoil former (which would pose induction problems), but uses two separated windings on a little plate, placed in the middle of the voicecoil kind of crossing this area:

View attachment 462817

Sensor control service



Yes, I would not overestimate the advantages such a sensor and negative feedback loop can bring, but I would say they can be significant in a limited frequency band under certain conditions.

It might be interesting how that would affect the concept of active speakers as the general opinion expressed by developers using sensor technology already is that it does not make sense with bass reflex concepts, drivers with limited excursion capabilities and DSP crossovers with significant latency. Sounds to me as if a radically different active concept has to be designed around this little sensor.



Which is not really a problem to implement in a negative feedback loop if you know how these parameters are depending on each other.

CODA AUDIO also uses an electrodynamic sensor for some of their PA speakers

https://codaaudio.com/sensor-control/

 

[bmc0]

But wonder if any speaker driver ever had voice coil windings not on the moving part ( to counter modulate the field with the signal ) ?

18 Sound makes a few drivers like this. Here's a paper about it.

 

[Buckchester]

If the loudspeaker chassis is otherwise well designed, you have an additional control level with which you can modulate the signal as desired in real time.

At low frequencies, I think the advantages are immense, yes.

If that were the case, then wouldn’t more companies do it? Tom V at PSA has said the same things can be achieved with more modern DSP.

 

[gamerpaddy]

Backes & Müller even made a servo tweeter in some of their BM series long time ago. (including the BM3 in profile pic)
the tweeter was a metal dome in a cage for safety with around 300 volts applied to it.

they sent the signal trough HV modulator and onto the dome, the electrode that picked it up went trough some circuitry into the amplifier drving the voice coil.

nice gimmick but didnt do much i guess

 

[Arindal]

If that were the case, then wouldn’t more companies do it?

It is an enormous investment in terms of maturing the sensor technology, developing negative feedback routines which are stable under all conditions and mastering the implementation in the power amplifier. None of these steps really works with off-the-shelf components.

Products on the market so far had to rely on closed-box designs in what is de facto a subwoofer, which is narrowing down the potential candidates for this technology. I am very curious to see how KEF will achieve clear advantages of this technology when using vented designs and 2-way coaxials, as they have hinted. That would really mean a big step forward.

Backes & Müller even made a servo tweeter in some of their BM series long time ago. (including the BM3 in profile pic)
the tweeter was a metal dome in a cage for safety with around 300 volts applied to it.

If I recall it correctly, this was a capacitive sensor, and prone to instability. So not really a surprise this did not become mainstream tech.

 

[Ssss]

When can we expect to see a KEF speaker with VECO? (not the soundbar).

 

[TheBatsEar]

Looking at my crystal ball - Blades with this new tech, brand new BMW M3 or put your kid through college?

KEF LS/50 Meta VECO when?

 

[Vacceo]

KEF LS/50 Meta VECO when?

Probably the 60´s can benefit more as they can dig deeper in frequency than the 50´s or LSX.