Theoretical Design - Current Feed Active Dipole Speaker

[ThorstenL]

Konnichiwa,

Elsewhere in the most abstract manner we discuss a rant about the drawbacks and benefits of current drive. What I'd like to do here is to present an easy and quick "testbed" for current driven Speakers.

For starters, if we use a "gainclone" amplifier converted to current operation we will have a pretty decent Amplifier per driver without breaking the bank. The circuit to that is here:

http://sound.westhost.com/project56.htm



The righthand side circuit is the current amplifier.

Now as discussed elsewhere (Current Driven Loudspeakers and Tranconductance Amplifiers ), feeding a driver using a "current source" amplifier eliminates all electrical damping and as shown by Hawkesford and Mills can lower distortion in cone and dome drivers of conventional construction by quite substantial amounts, plus it cancels out any dynamic compression.

With all electrical damping gone the Qe of our drivers becomes close to infinity and thus for a good approximation we hold that the drivers Qt = Qm.

I had noticed previously that Seas makes many drivers with a fairly low Qm. Looking around and thinking of a simple design that would have some general appeal I decided to come up with a fairly sensible, floorstanding dipole design (I like dipoles) that combines the various features, namely Dipole operation, Low Qm Drivers and Current Drive operation to good use. A tall column appears a format universally liked and I like Coaxial Drivers.

So usong a 6.5" Seas Coax and 4 pcs 6.5" Seas Woofers came to mind as "natural" solution. The drivers chosen are the T18RE/XTVFC as "fullrange coaxial" and the CA18RLY as 4pcs Woofer.

The woofers have a Fs of 40Hz and a Qe of around 2. In other words, fed with "constant current" they will exhibit a 4.5db lift at 45Hz. This lift can be used to compensate part (or all) of the dipole rolloff, below 40Hz however the SPL will drop like a stone at a rate faster than 18db/8ve.

Using 4 Drivers may give enough displacement to use a Linkwitz transform circuit to shift the LF cutoff down, however I feel I would suggest as preferable a sealed box subwoofer to be cut in, probably with a crossover point (-6db) around 30...35Hz and a 3rd order lowpass, which would integrate pretty well in most domestic rooms.

The main Coax will show a more modest LF lift of 1db @ 55Hz. We would need to EQ a lot to keep this driver operating to low frequencies in a small dipole.

In order to keep things sensible in size we will use a simple plan baffle, 25cm wide and 1m high. This is a rather small dipole, but with 4pcs of woofers and the floor image we get a good deal of available SPL.

We can use "The Edge" (how about stuffed crust?) to get a general idea what the SPL distribution will be like, the recent version can handle 4 Woofers on a narrow baffle, but still does not account for the floor image.

"The Edge" is here:

http://www.tolvan.com/edge/

For our baffle

To estimate the effect of the floor Image a piece of Software from Roy Allison and RA Labs called bestplace. You can find one download here:

http://www.arsenal.net/speakers/allison/allison.htm

Now, with our woofer column's acoustic center around 40cm (16") from the floor bestplace suggests we gain a 3db flat boost in SPL below around 50Hz which in turn slowly reduces to 0db @ 200Hz.

So, our drivers together with the floor image will give around 7.5db boost at 45Hz and return to a flat response at around 200Hz. Now "The Edge" suggests a total SPL loss compared to the Midband (@500Hz) of around 12db @ 50Hz. After accounting for around 7db boost from the Drivers high Q and the floor image we get around -5db @ 50Hz.

So, boosting 40Hz by around 6db on top of the existing system should result in a flat response. This kind of boost can be easily engineered into the feedback loop of the amplifiers for each individual driver, using a LM3886 or LM3875 should allow set of woofers to generate > 97db/1m @ 50Hz, despite being on a small dipole, at 100Hz and above the maximum SPL should exceed 103db/1m well matching the Mid/Hi section as we shall see later.

Now, where do we need to cross over the Coaxial driver?

Using "The Edge" suggests that our driver will be down 3db compared to midband output at around 300Hz, falling off with a first order function down over at least two octaves. If we add a simple 1st order 300Hz highpass (to limit excursion) to this we will require a 2nd order crossover (lowpass for the woofers) to match this slope, the crossover being down 6db around 300Hz.

The coaxial Driver has a pretty flat response up to around 4KHz followed by a fairly steep rolloff. This means that the rising impednce from the inductance is compensated mechanicall, so under current drive the Coax Cone driver will show a rise of likely around 6db towards 4KHz, before rolling off, with the response again flat around 300Hz. This suggest again a simple compensation via a -6db shelf @ 4KHz in the Woofer Amplifiers feedback loop.

The cone section should provide a maximum SPL of > 105db/1m within it's operating range driven by a LM3886 or LM3875.

The tweeter should cross over at around 2KHz, my choice would be to wrap the tweeter filter directly around the Amplifier, as 3rd order lowpass and to derive the woofers lowpass filter from the tweeter as subtractive X-Over. And add a modest amount of EQ for the tweeters rolloff above 10KHz, again a shelf filter with 6db Boost 15KHz and 3db point at 10KHz would seem to suit.

As discussed, a further 1st order 300Hz High pass is needed for the Coax Section, this would be easily implemented as series capacitor ahead of the dual Chip Amp section which would cover crossover, current drive and basic driver EQ.

Then a simple 2nd order lowpass for the woofer section with individual current drive amplifiers per driver incorporating the 6db 1st order lift @ 40Hz in the feedback loop and you have a fairly simple to design and build speaker.

The resulting speaker should in a number of areas substantially outperform conventional speakers, be it in the area of compression and distortion at higher levels or in terms of room interactions in the modal and above range. I hope this little thought excercise will be of use to some....

Sayonara

[Elias]

Se on moro!

According to my understanding main distortion contributors in bass and low frequencies in a dynamic driver are the nonlinearities of the spider, cone surround and non-homogeneous magnetic field. Current drive can not improve any of these nonlinearities.

Where the current drive is (might be) useful is at higher frequencies by correcting the nonlinearities caused by impedance modulation of the voice coil inductance.

Usually, if one can use such a word, voice coil inductance of the bass driver starts to dominate the impedance around >1kHz.

To get linear bass you need linear drivers, or motional/acoustic feedback.

- Elias

[ThorstenL]

Konnichiwa,

Originally posted by Elias
According to my understanding main distortion contributors in bass and low frequencies in a dynamic driver are the nonlinearities of the spider, cone surround and non-homogeneous magnetic field. Current drive can not improve any of these nonlinearities.

These are SOME of the main contributors, yes. However, as Hawkesford & Mills show, at 100Hz (the Kick Drum is tuned around there, as are tympani) and +9dbw going from voltage to current drive on a fairly generic 6.5" Woofer improved 2nd harmonics almost 10db and higher harmonics by around 3 - 4db. So even at failry low frequencies do we attain benefits from current drive.

As the ear becomes progressively less sensitive to distortion with falling frequencies I would not worry excessively over very low distortion bass, but the reduced compression from current drive should be very much more notable.

Originally posted by Elias
Where the current drive is (might be) useful is at higher frequencies

As commented earlier, benefits are realised even at fairly low frequencies, but yes, with the example driver used by Hawkesford and Mills the improvement at 3KHz (where the ear is much more sensitive to distortion) is around 26db, but even a 10db reduction in distortion at LF for in essence FREE in an active speaker is not a bad thing.

Originally posted by Elias
To get linear bass you need linear drivers,

I am unaware of any LINEAR driver, only drivers with varying degrees of non-linearity. I find the best solution to retaining linearity is o make sure the driver cone moves as little as possible. I tried to go a part of the way by using 4 6.5" woofers instead of one, which should for a given SPL reduce the excursion to 1/4. Given that with acoustical music the maximum SPL falls notably below 100Hz compared to the Midband this should be sufficient in many situations. But I agree, scaling the LF system up to say a pair of 10" or 12" Woofers would be preferred, but would lead to problems with the "slim" siluette of the design.

Originally posted by Elias
or motional/acoustic feedback.

Motional feedback is a good idea and would be the next step, but it is comparably harder to implement than current feed and as Hawkesford & Mills show, the improvement with current feed & MFB over current feed alone at least at 100Hz/+9dbw is modest.

At 2nd harmonics their experiements showed no improvement, higher harmonics where attenuated a further around 10db over current feed, in my view a much smaller return on investment, compared to the current drive case which requires minimal investment to apply.

Sayonara

[phase_accurate]

I haven't tried current feed with one of the favoured IC amps. But one has to watch out that stability is maintained at higher frequencies, since the series RL circuit of the driver will introduce a first order pole in the feedback part of the NFB loop.
This is first desirable from the response-compensation point-of-view. But it is also detrimental to amp stability.

So one has to be careful when trying it out.

If it IS unstable I have to proposals:

Either use a voltage feedback resistor as well, setting a quite high gain. You won't have a pure current source anymore but a voltage source with a high output impedance.
The other solution is using a series RC (I do deliberately NOT call it a Zobel) accross the driver. Depending on dimensioning you will loose some of the Lvc-response-compensation at the top end.

Sometimes you'll have to make compromises.

Regards

Charles

[ThorstenL]

Konnichiwa,

Originally posted by phase_accurate
But one has to watch out that stability is maintained at higher frequencies, since the series RL circuit of the driver will introduce a first order pole in the feedback part of the NFB loop.

Yup. But if you look closely you will see that the Amplifier Chip will remain stable as long as the current sense resistor is Re/10 under all conditions. The increase of driver impedance with frequency will increase the systems noise gain to higher frequencies which is just what we want....

Originally posted by phase_accurate
This is first desirable from the response-compensation point-of-view.

Yup.

Originally posted by phase_accurate
But it is also detrimental to amp stability.

Nope, it is actually BENEFICIAL to Amp stability.

Sayonara

[JensRasmussen]

What happens if you play the amp with no load?

\Jens

[cocolino]

BTW: what happens when the speaker is not at all connected and
the feedback loop isn`t closed?
I believe in such a current drive amp there should be precautions to avoid this condition (or at least people should be aware that with current drive plugging in/out the speakers as they`re used to do with voltage-drive could result in damage to their speakers) .


Ooohps - haven`t seen that the question has already been asked by Jens!

[phase_accurate]

Nope, it is actually BENEFICIAL to Amp stability.

No it definitely isn't since it reduces phase margin.

Edit: I don't want to claim that it definitely has to be unstable but that there are some pitfalls one has to be aware of.
Another one is the fact that the (basically positive) feature of compensating the top end rollof (caused by Lvc) might increase the possibility of TIM. So be aware that you have some margin regarding output voltage swing.

Regards

Charles

[ThorstenL]

Konnichiwa,

Originally posted by JensRasmussen
What happens if you play the amp with no load?

You get no sound.

Originally posted by cocolino
BTW: what happens when the speaker is not at all connected and the feedback loop isn`t closed?

It will alternate between the rails according to the signal.

Originally posted by phase_accurate
No it definitely isn't since it reduces phase margin.

I don't think it reduces the phasemargin at high frequencies, capacitive elements do, inductive shouldn't as the phase will be shifted the opposite way.

Originally posted by phase_accurate
Another one is the fact that the (basically positive) feature of compensating the top end rollof (caused by Lvc) might increase the possibility of TIM. So be aware that you have some margin regarding output voltage swing.

Well, as long as you remain in the linear operating range of the Amplifier and the usual precautions against TIM are taken you will not have any increased risk of TIM over what is already there.

I notice that everyone tries to find the "problem" without ever stopping to think and analyse and to provide something constructive. Guy's, I used current drive with chip Amp's back in the 80's of the last millenium! Never yet blew out drivers or chips.

Maybe everyone feels that "If current drive was so good everyone would use it!" and hence seeks what's wrong? Have you not noticed commercial reality?

If cold, highly efficient, energy saving and extremely long life lighting was a good thing, surely everybody would be using it?

If the "Wankel Motor" (rotary engine) where a superior design, offering less fuel consumption with more power from a much smaller and cheaper to make engine, with much less operational noise and vibration surely everybody would be using it?

I can go on. Current commercial reality means that significantly "different" approaches, even if obviously superior and without drawbacks are being ignored in favour of what was made yesterday....

Sayonara

"they saw the founding of a great republic and proclamations hailing new gods named Due Process and Equal Rights for All. And they saw many in high places in the republic form a separate cult and worship Mammon and Power.

And the Republic became an Empire, and soon Due Process and Equal Rights for All were not worshipped, and even Mammon and Power were given only lip-service, for the true god of all was now the impotent 'What Can I Do' and his dull brother 'What We Did Yesterday' and his ugly and vicious sister 'Get Them Before They Get Us'."

From Robert Anthony Wilson "The Golden Apple"

http://www.rawilson.com/goldenapple.html

[JensRasmussen]

No sound without the speaker, well you got that right. Let me ask another way: what happens when you play the amp without load, and while playing connects the load?

My point is that because of no load there will be no feedback, and the amp's output will be rail to rail. When you connect the speaker I fear that there will be sound in a big way.

Regarding phase, I like the concept of true current feedback, because the speaker movement will be in phase with the input voltage. I used this in a project to get rid of the delay in the current vs. voltage through a coil.

I have not seen it implemented in an audio amplifier yet – don’t know why, maybe because of the no load scenario.

\Jens