new topology for currrent drive

[capslock]

The merits and downsides of current driving loudspeaker drivers have been discussed before and should not be part of this thread (cf.
transconductance power amplifier for current-drive loudspeaker?
output stage with high impedance
Current Driven Loudspeakers and Tranconductance Amplifiers )

Just as a summary:
+ lower distortion due to elimination of Bxl term
+ better HF extension due to eliminaton of Le
+ lower IMD distortion due to elimination of Le and hence delta Le
- Qt = Qm, hence need to control behavior around resonance (use Linkwitz transform or add mechanical damping)
- Qt = Qm, hence slightly slower decay at f >> fs
- only makes sense for active concepts

What I want to discuss here is an amplifier topology that has inherently high output impedance.

Again a summary of what has been discussed (Hawksford wrong on amp topology for current driving loudspeakers? ):
Hawksford and Mills have proposed a complicated design with floating power supplies and no overall feedback that I do not find compelling in terms of distortion behavior.
The standard approach is to use a voltage feedback power amp in non-inverting mode and connect the loudspeaker between output and inverting input and have an additional sense resistor from there to ground. I have calculated that a humble LM3886 achieves sufficiently high output impedances, and it is clear that an amp with higher GBW will have even better performance.

However, it is not very elegant to rely on feedback to convert an inherently low output impedance (around 0.1 R) to something in the 10 kR range.

So, what are other possible approaches

1) Use a voltage controlled current source (VAS) or current mirror, with or without global feedback, and run it in class A against a constant current source. I bet Nelson will love this approach, but I do have a bad concience about wasting electricity.

2) Use the bidirectional class AB current mirror that is found in so many integrated CFB op amps. I have not thought enough about this to discuss it now.

3) Use a conventional voltage feedback design and remove the output follower. In other words, the VAS is the output. For this to work in class AB, it must be a complementary VAS so that both current source transistors can be controlled. This is achieved through a complementary input stage or through some folded cascode design. The VAS transistors would be cascoded, with a fast small signal or TO-126 transistor as the current source device and a power transistor as the cascode device. Overall feedback is achieved, much like in the conventional current drive approach, by connecting the loudspeaker between VAS out and inverting input and using a sense resistor between there and ground. As in the conventional approach, it may be necessary to bypass the speaker with a resistor and capacitor connected in series to keep feedback impedance and hence voltage gain low at high frequency.

OK, what do we gain from this?
+ Output impedance without feedback would be in the kR range for small currents (< 8 A for modern devices) and I suppose still in the 10-100R range close to rated current where beta nonlineary is rampant.
+ The follower stage is omitted which only slows down the amp.

Where are the pitfalls (or why haven't I seen this approch before)?
- compensation may work only for one specific driver (not a problem in an active configuration)
- load generated back EMF - how do the VAS and the feedback circuit cope?
- nonlinearity of cascode transistors cannot be compensated for and is probably greatest contributor to open loop output impedance variations
- VAS will enter class B, i.e. one side will be switched off temporarily. This should be dealt with in a non-switching design or at least with a few anti-saturation diodes. I have not yet thought about this enough.

What do you think??

[cunningham]

If driving a higher impeadence is your goal, then why not make a class A VAS biased with a current source that biases maybe 1mA or so, and then use a class AB current sourced biased EF stage as your output. If you are going to drive a high Z then just use some TO-92 style transistors for this stage. output maybe 10mA peak. They can be biased by 40-50uA, current source. This will give you an overall bias much less than using a higher biased class A VAS. More parts, but better results. If you want my opinion on a VAS transistor for this application, I would recommend an ECG26 type device. Vce 120V, Pd 200mW Hfe 300 Lo-noise audio amp. Great voltage amp, low current.


This type of circuit would be perfect to drive a current source biased darlington output stage.

[AKSA]

Eric,

I haven't read too much of this thread, but it is very interesting. Can I add my quick and dirty 2c here?

I agree pretty much with all your expected benefits. However, there are some mitigating factors which must be considered:

1. A high impedance source driving a speaker must rely on high mechanical damping since there is little electrical damping. This is, as we all know, a SET style of speaker, highly efficient, low excursion, limited spL, horn construction, low power, very peaky.
2. Most speaker drivers are designed for global, voltage NFB. Moving to current feedback will have its problems!
3. In all my experiments some years back with current source, SE circuits using CFB from a sense resistor in the speaker circuit, I noticed wonderful bass control but a severe congestion of the image. I'm not sure of this, but it's a problem.
4. If you go VAS to speaker direct, that's gotta be 1 big mutha of a VAS, and your notions of power efficiency are out the window!


Cheers,

Hugh

[Nelson Pass]

Your points are well taken Hugh, and in fact there are practically
no speaker drivers that fare best with a pure current source -
rather we find that under some conditions some speakers see
improvement with higher source impedances. It varies all over
the map, but haven't found any drivers that require source
impedances greater than about 50 ohms.

Obviously the vast majority of drivers want a low source
impedance, and only a small percentage want high source
impedances. These tend to be the high efficiency types.

As a consequence you can easily create the high source
impedance you might want by placing a power resistor in series
with the output of a voltage amp. The high efficiency of the
driver will usually let you get away with this with a conventional
100 watt amp, and since the amp is seing high load impedance,
it usually will operate at quite low distortion. This is not
necessarily what I recommend, but it has the virtue that you
can try it without building a special amplifier.

I have not been thrilled by the performance of a current-feedback
amplifier either (and of course in this case I mean feedback used
to create a high output impedance - not the other kind).

Many of the full-range high-efficiency drivers are peaky, but not
all, and in many cases there are easy techniques to passively
control this.

[phase_accurate]

Maybe a mixed solution between current- and voltage drive should be considered: Voltage drive from DC up to just over the resonance area and current-drive further above.

Regards

Charles

[capslock]

Hi folks,

please let's keep this thread to amp topology. The pros and cons of current driving loudspeakers have been discussed exhaustively in the linked threads. To me, it is clear that current drive only makes sense in a fully active setup.

Cunningham: what is a current source biased follower? I can imagine a few common topologies, but all have pretty low output impedance.

Regards,

Eric

[Konrad]

Originally posted by capslock
Hi folks,

please let's keep this thread to amp topology. The pros and cons of current driving loudspeakers have been discussed exhaustively in the linked threads. To me, it is clear that current drive only makes sense in a fully active setup.

Cunningham: what is a current source biased follower? I can imagine a few common topologies, but all have pretty low output impedance.

Regards,

Eric

agree !

It just came to me that using a normal amp, and just modify the feedback so that we add current feedback in a passband so that the voltage gain is halved within a passband. Thus i belive in using simple things i belive in using (Rvc / Rsence) filtergain equal to the voltage gain.

After testing and measuring (!) the next step wold bee to pass out the voltage feedback in the 'most used passband' ie the normalised.
Then hopfully giving the possibility to ekstend the elements freq response in both ends, but limiting it to voltage gain outside a passband.

What is left is to find the passband for current feedback and (maybee not equal to voltage bandstop filtering settings but i think wil bee normalised to as different designa have their requierements).

2 times voltage gain outside the passband remember the SPL limits. I dont think it's advisable to exstend it further (under most sircumstances) without looking at filtering etc UNLESS/Or Yo'r using a lot of elements. But again then fft response analyse and designing from a different aspect again gives other popps & corns
Remember at 20 Hz you would need a bucket or to of air to flow to give consert levels at home ( unless you live in a carlike appartment )

[Konrad]

The sugested value for the sense resistor is already given in som of the above post's . After all the power losses are of consern, but calculation of system's new Q at the low end will still be 'easy'.

If we compare what i have found and seen so far in MFB treads ( servo subs accelerometers and DVC elements ) One of the ways is to pass out the feedback 1.order hi and low from a refrense freq.
Also i think current drive at low freq is equal voltage drive as they become more and more in phase as freq is lowered.

For some of us this stops wit a litle teory, for other just before measuring and for most of us by bouth.

In the bok of Martin Collums he shows sevelal advantages as redused distortion etc with current drive compared to voltage drive.

Getting rid of what i belive is that motional induced voltage is correct only 1/2 of the time it may bee god reasons for getting rid of the phenomena, and aply the current ( the correct amount of current, not the distorted by the movement of the element )
I belive in putting current to the element regardless of movement than rather than putting it the wrong way just because the element happens to move the vrong way caused by somthings happened in the past. whitch i belive is tim
and please comments are apreasiated.

[pmkap]

In an amazing display of non self promotion and propriety, Nelson failed to make reference to his new whitepaper that is quite relevant. As this thread is specific to topologies, the paper is quite on point, as it refers to the use of an rlc network in parallel to an amplifier's output to get you there....
http://www.6moons.com/audioreviews/f...firstwatt.html
FWIW

[cunningham]

Originally posted by capslock

Cunningham: what is a current source biased follower? I can imagine a few common topologies, but all have pretty low output impedance.

Regards,

Eric

Here is a circuit topology for an emitter follower circuit with current mirror bias. Sorry this is for a lower output impeadence, but it can be adapted for a higer impeadence by adjusting the resistor values for less current, Vcc, and transistors used, but will work the same way.

mount the servo to the output's heatsink to achieve thermal stability. I included the equation for I-bias on drawing. The input impeadence is determained by how much current the driver transistors need to operate, not I-bias. The amount of current the drivers will need will depend on the output current and gain...Hfe.

BTW...This is not a true darlington circuit, nor is any darlington transistors used here. The 1A diodes have a larger forward bias voltage than what takes to bias on the output devices so bias is controled by pot on emitter/base of servo transistor.

As for the current source\current mirror circuit, Q1 is saturated because collector is connected to base. Current through this saturated transisor is set by (Vcc-0.6)\Rx. Q1 will not allow any more current than this to flow through Q2 because this would turn Q2 off due to internal resistance between the base and emitter. Therefore current through Q2 is equal to current through Q1 and is mirrored through the load of Q2(I-bias).

This is used often to bias a differential amp circuit.

Hope this helps to explain...

Chris