Cuircuit for simplistic current amplifier

[svenr]

Hi folks,

I'm looking for a current amplifier circuit or topology which is simple to integrate. That's the background of the story:

- 2 x 4 channels of current amplifiers for an actively controlled monitor system
- 2 x 3 channels need to provide approx. 25W RMS to a minimum impedance of 5 ohms under steady state conditions
- 2 x 1 channels require approx. 50W RMS assuming the same load
- the 25W channels should if possible operate in Class A to avoid the hassle of compensating for crossover distortions and the need for elaborate phase inverters
- ideally, readily available ICs should be used for the sake of mechanical and thermal simplicity

I know this is not the kind of topic commonly discussed here, but I'm shure some of you will have some very useful suggestions. As to why use current amplifiers? Simply because they effectively circumvent the dominant source of non-linear distortion in direct radiator applications for monitoring systems - voice coil resistance changes. Given strict adherence to minimum-phase behaviour of the driver, the old objection of insufficient electrical damping is irrelevant as equalisation of the occuring transfer function peak near the resoance frequency solves the problem.

Cheers: Sven

 

[gyraf]

LM3886?

 

[abbey road d enfer]

As to why use current amplifiers? Simply because they effectively circumvent the dominant source of non-linear distortion in direct radiator applications for monitoring systems - voice coil resistance changes.

Voice coil resistance in itself is the cause for insufficient electrical damping of the driver, so it is somewhat a cause of non-linear distortion. Changes of VC res is the cause of thermal compression (decrease of efficiency) and alignment drift. The speaker/enclosure alignment changes when the Qts and Vas of the speaker changes. Changes of Qts are mainly due to change of VC res (change Qes), changes in Vas due to temperature elevation (change compliance). This alignment drift is not necessarily accompanied by an increase in THD. The only certain consequence is a change in the LF response. Careful design allows the consequences of this alignment drift to be minimised.

Given strict adherence to minimum-phase behaviour of the driver,

This is to be taken with a pinch of salt. The behaviour of a speaker can be assimilated to minimum-phase only in a limited frequency range (approximately the range where the wavelength is significantly smaller than the depth of the cone). I think it's ok to make this assumption since you seem to be concerned with the LF response mainly.

the old objection of insufficient electrical damping is irrelevant as equalisation of the occuring transfer function peak near the resonance frequency solves the problem.

Taking into account your requested spec of "minimum impedance of 5 ohms under steady state conditions" means that the resonance peaks would be damped by the amplifier, but not the changes in VC res. The most extreme VC res variations are 1:2 factor (from 0°C to 300°C), i.e. from 6 ohms to 12 ohms for an 8 ohms speaker. If you want to minimise this, you need at least 60 ohms source impedance (10 times the variation).

Loudspeaker damping is a complex subject. If a speaker was heavily electrically damped (very low VC res), the electromotive force would be equal to the drive voltage, so it would be a velocity transducer, a 6dB/octave differentiator. (dx/dt = sin wt => d²x/dt² = w cos wt)
Modern loudspeakers are built with an overhung coil, i.e. with a large part of the coil producing no force, which increases the DC resistance; in fact the speaker becomes a BLi driven system, where the force is the image of the drive voltage. In the mass-driven operation range, force equals sound pressure, which yields a somewhat flat LF response, at the cost of enormous power loss (most LF speakers are less than 5% efficient).

 

[tv]

You are probably looking for something like this:
http://sound.westhost.com/project83.htm

However, imho, you will find more practical advice on diyaudio.com boards - simply because this is where the class-A heads are hanging out and have actually built stuff like this for their own use....

 

[JohnRoberts]

That sounds like a current output amplifier. Would still be voltage input.

I suspect if you rearrange the feedback pickup, from a typical voltage output amp to instead put out current(look at a small resistor in series with output differentially, so you are effectively controlling the voltage drop across the resistor in series). 

You may need more output voltage swing than a typical voltage amp, since loudspeaker impedance peaks will require several times the voltage to make a nominal current.

I recall messing with a variant of this on a small 4" speaker back in the '70s and it was interesting, but this may be more of a science fair project than Hifi. I recall a tiny speaker in a tiny box making a bunch of bass, before it released it's smoke.

JR

 

[gyraf]

yep - the LM3886 can be used to run current-mode.

Jakob E.

 

[tv]

Ah, you were looking after this
http://sound.westhost.com/z-effects.htm
http://sound.westhost.com/project56.htm

 

[abbey road d enfer]

This circuit is better because the speaker can be grounded and can have gain (by increasing R2 & 4). Any power IC with an opamp structure will do 3886, 1875, 7294...even the STK hybrids.

 

[svenr]

@ at all
thanks for the suggestions so far, the LM3886 seems to fit very well.

As I understand the cited project 83, this is a voltage amplifier, the output impedance of this circuit will be very low.

@ abbey road d enfer
My use of the term 'non-linear distortion' may have been confusing. I understand that you refer to the enclosure compliance Cmb (or Vab) which changes with enclosure temperature. The alignment drift becomes increasingly less manageable the lower the VC's thermal capacity before failure. One solution is to steer well clear of the system's resonance frequency when designing the operation bandwidth of a driver. However, bearing the directivity in mind, this may result in an unacceptable number of 'ways' required.

The minimum-phase assumption certainly only applies to the effect of the current amplifier's large output impedance on the alignment. Resulting underdamped CB alignments can thus be remediated using conventional equalization. The minimum 5 ohms refer to the minimum load resistance, not the output impedance of the amplifier; this should of course be as high as possible.

 

[abbey road d enfer]

@ abbey road d enfer
My use of the term 'non-linear distortion' may have been confusing. I understand that you refer to the enclosure compliance Cmb (or Vab) which changes with enclosure temperature.

No, this is linear distortion (frequency response error). The non linear distortion I'm talking about is the accumulation of non-linearities in the BL product and in the suspension compliance, which are the main cause of THD in loudspeakers. A small Qes allows better control of the movement of the voice coil. the main parameter in Qes is the voice coil DCR. If an additional resistance is used, the resultant Qes becomes higher. A speaker system with an effective Qts > 0.7 will: A) ring badly, B) distort. A loudspeaker with a Qts of 0.3-0.4 will not ring too much. Putting an 8 ohms res in series with it will push Qts in resonant territory (and distort a tad more). Even more so with current drive.

 

[JohnRoberts]

Without passing judgement on the merits of this experiment,,,  the use of this feedback topology will bootstrap the output impedance much higher than even the apparent 1 ohm in the schematic cited.  BUT... how much higher will depend on the precision of all 4 of the resistors in the feedback network, and the loop gain margin of the amplifier.

Using 1% resistors (for the 10ks ) should push the 1 ohm up effectively 100x or to 100+ ohms. Since 1% resistors from the same batch should track better than 1% expect higher than 100 ohms.

The transfer function of 1V input= 1A output may be pretty close to what you want, since it looks equivalent to a voltage gain of 8x driving an 8 ohm speaker, or 4x driving a 4 ohm.. If you tweak the overall gain, the two input resistors on the input side of the opamp need to be the same value, and the two feedback resistors on the output side need to be the same value. 

Have fun...

JR

Note: you will be dropping a bunch of power in the 1 ohm series resistor, but it is not that uncommon for modest sized power amps to use .1 to .33 ohm degeneration resistors in the output stage.

 

[PRR]

I suppose you have read High Performance Loudspeakers by Martin Colloms? ABE.com has older copies priced about what IMHO it is worth ($8-$30). There's a new edition, I have not seen it.

> operate in Class A ... - ideally, readily available ICs should be used

That size total power-out, in A, will need a heatsink bigger than your cabinet.

The usual "audio ICs" are all AB. And several of them handle it very well. Crossover is a known solvable problem. And often easier to manage than the Gm-doubling and HEATsinks needed in A.

> ...Class A to avoid ... need for elaborate phase inverters

I do not understand.... unless you mean "single-ended"?

> current amplifiers ... circumvent the dominant source of non-linear distortion in direct radiator applications for monitoring systems - voice coil resistance changes

I don't accept that at the half-cycle level.

Current-source does "cure" longer term thermal compression; by supplying more power to a hot voice coil. Which gets hotter. The acoustic output stays similar.... until the coil melts. With volt-drive, the electrical power and acoustic output drops, which is unfortunate, true.

Wouldn't it be nicer if we had a coil so big it didn't get hot? (Yeah, the impossible dream... there are users who can melt anything, meanwhile the big coil interferes with other acoustic goals.)

But also (unless the speaker is John's tiny 4-inch) the listeners' ear sensitivity drops during prolonged high-output sound. Shouldn't we compensate for this by INcreasing electric and acoustic level in prolonged loud passages?

25W-50W lightly-clipped speech/music in decent size drivers should not give huge thermal compression. There are a variety of drivers (mostly woofy) which will take an apparent 500 Watts for hours. "Apparent" because thermal resistance means they are really sucking 250 Watts after the first few minutes. But 25W-50W amplifier on clean speech/music is 2W-5W heating power. Anything much bigger than an FE-103 should take that with less than 1dB thermal compression.

I won't even discuss "minimum phase" on loudspeakers, but yes of course you can EQ-away the bass bump. Note that a peak plus a dip is NOT flat: the steady-state is flat but it still rings. Slow to build and to decay.

> The most extreme VC res variations are 1:2 factor (from 0°C to 300°C)

Oh, I bet we could find ceramic binders to keep a copper coil together at much higher temperatures. Might need to water-cool the magnet below Curie temperature. The history of the internal combustion engine is a lot about heroic heat and keeping it together. We used to pour a bucket of water over the engine every hour, now we have constantly pumped water, and some folks are trying to make ceramic pistons to run even hotter. Loudspeaker designers are chickens who stalled as soon as the (already poor) efficiency fell off.

> If a speaker was heavily electrically damped (very low VC res)

Then it would be as much fun as an electrostatic speaker: pure capacitance (above resonance). The impedance would fall and fall, the loadline become circular and trespass the transistor SOA. The VC resistance spoils our damping, ruins efficiency, and adds thermal compression.... but the alternative (zero resistance) is worse.

Nelson Pass has posted of experiments with current-drive. Works for him. There is also that large class of Guitar Amps with near-naked Pentode outputs and damping-factor of 1 to 0.1. Pass's work could become mainstream.... but I don't see it happening.

 

[JohnRoberts]

Coincidentally there is a discussion right now on another forum about (big dog) SR loudspeakers losing sensitivity over time, and the speculation there is that VC heat is trashing the neo magnets.  OTOH it may just be some flaky magnets. Stranger things have happened with the funny new improved magnetic compounds.

JR

 

[abbey road d enfer]

> If a speaker was heavily electrically damped (very low VC res)

Then it would be as much fun as an electrostatic speaker: pure capacitance (above resonance). The impedance would fall and fall, the loadline become circular and trespass the transistor SOA. The VC resistance spoils our damping, ruins efficiency, and adds thermal compression.... but the alternative (zero resistance) is worse.

 

It would not be capacitive, it would be inductive, which is a manageable load. But its frequency response would not be flat.

 

[abbey road d enfer]

Coincidentally there is a discussion right now on another forum about (big dog) SR loudspeakers losing sensitivity over time, and the speculation there is that VC heat is trashing the neo magnets.  OTOH it may just be some flaky magnets. Stranger things have happened with the funny new improved magnetic compounds.

JR

Some neo magnets have a very low Curie point (80°C). If special care is not applied to make sure the magnetic pieces are sufficiently cooled, they may well become permanently demag. This has been the main concern when designers have introduced Nd in high-power loudspeakers; today, most of the serious manufacturers have mastered this issue.
Were there any specific brands/models quoted?

 

[tv]

http://www.national.com/an/AN/AN-1515.pdf

fig 5. and later


http://www.westebbe.com/Vox/AD15_30Servicemanual.pdf

find the LM1875 output stage ... a circuit funkier than bootsy

 

[JohnRoberts]

Some neo magnets have a very low Curie point (80°C). If special care is not applied to make sure the magnetic pieces are sufficiently cooled, they may well become permanently demag. This has been the main concern when designers have introduced Nd in high-power loudspeakers; today, most of the serious manufacturers have mastered this issue.
Were there any specific brands/models quoted?

I think they were talking about some big JBLs...  I repeat this is speculation. Most of these high power speakers try to use VC movement to force air over the motor to improve cooling and power output. The speculation was about what happens when the music, and airflow stops? Certainly beyond this discussion, involving relative thermal masses, differential passive heat flow, yadda yadda...

JR.

 

[xmvlk]

Hi folks,

I'm looking for a current amplifier circuit or topology which is simple to integrate.

I do mean, that circuit like Howland and so on {circuit based on NIC with grounden norator} are
not suitable for power circuits. There is sensing resistor and feedback floating over full supply swing.

May be more effective is circuit based on OPA660 supertransistor OTA with output in collector, input
in base and output in collector. There is no problems with integration, because OTA is transilinear circuit
{BTW this circuit is not realisable by discreetes}
BTW I love current amp sound on woofers, because their resonance is not damped.It results on linear
distortion - bas boost, which is rather pleasant in closed cabinet loudspeakers.

Yes, it is problems, that we do not have power OTAs, but Cheenese peoples may fix this in short time.

 

[JohnRoberts]

Yes, it is problems, that we do not have power OTAs, but Cheenese peoples may fix this in short time.

There is no reason why you couldn't take an IC OTA front end and buffer the output with a pair of current amplifying current mirrors from each supply. This would be a lot like designing a power amp output stage from scratch, so perhaps beyond entry level DIY.

JR

 

[tv]

Marshall, Vox etc use such circuits in their _built_to_be_fu..in'_abused_ guitar amps (discrete and chipamps) so I don't see any reason why they wouldn't be suitable...

--- since it dawned on me that the OP seeks a current-drive amp and not an amp that would be merely a current amplifier (i.e. a voltage follower in class-A .. )