LM3886 as transconductance amplifier?

[squadra]

About 2 years ago I built an amplifier using the LM3886 (non inverting).
It sounds quite OK and is in use driving a pair of monitors on a mixing console.
But I have some pcb's left for another pair and am thinking about the following experiment:
- Remove the feedback resistors
- Connect the loudspeaker from the output to the - input
- Connect a 0.33E resistor from - input to ground ( 3 * 1E parallel)
This looks to me as a very effective way of creating a V to I converter, Ispeaker equals Vin/0.33

The loudspeaker's impedance is 8 Ohm and doesn't reach 3 Ohm, so the LM3886 appears to operate in it's stable region.
Since I love my speakers and don't want to fry them , I was wondering if there are reasons not to try this.

[leadbelly]

Originally posted by squadra
I was wondering if there are reasons not to try this.

You're leaving out way too many things needed to make the chip operate properly. Read up on very basic op amp applications, like the 1st chapter of the Analog Devices handbook. It explains things like voltage gain and noise gain and the like.

[KSTR]

In this config you need to compensate for any driver/cable inductivity, otherwise it may become unstable very quickly. But basicly it's working fine, I've used it several (although not for driving speakers).

If we assume the driver inductance as perfect, you get an impedance rise of +6dB/oct from the Re-Le intersect point. Say you have 6Ohms DCR and 1mH of coil inductace, you need to compensate that with a series RC in parallel to the speaker, with R=Re and C=Le/Re². Then you have true ohmic impedance of Re and it will be stable. There are other ways to stabilze it, see ref.#2 This is partly because the impractial values for the compensation cap: 1m/36=27uF, a rather big cap. If it is too small, you get impedance peaking, if it is too big, a drop, and both might easily be in audio band, in this example right around 1kHz.

A resistive bypass (on the order of 100R...1k) is also a good thing to have as a fallback, you don't need pefect V-I behaviour to current-drive a speaker.

See also:
http://sound.westhost.com/lr-passive.htm#s3.0
(and note that in reality the Le is not perfect, no true coil behaviour).

http://apex.cirrus.com/en/pubs/white...e_circuits.pdf
(Chapter 7: Floating Loads. You could also try a Howland pump. Slightly mistuned Howlands have been used. ie in Mauro Penasa's RevC, albeit with a voltage feedbacl loop around it to make it a voltage output again).

http://www.essex.ac.uk/dces/research...0amplifier.pdf
(here, some minor problems of this circuit are discussed, right at the beginning, and a better circuit is proposed).

- Klaus
EDIT: Error in formula

[gaetan8888]

Hello

There is also this quite simple version of an Howland pump lm3886 amp;

http://www.cabezal.com/~inguz/amp/inguz_amplifier.pdf

Gaetan

[squadra]

Thanks for the info.
My monitor is a 2 way system and I have no idea how to compensate for that... looks like my easy solution is a no-go
I have to check what changes i need to make to the pcb to go for the howland current pump.