Lm3886 current feedback

[digi01]

I WILL TRY ANOTHER WAY TO CONFIGURE LM3886.
THE LM3886 DATASHEET only recommend how to configure lm3886 in no-inverting & inverting mode that is voltage feedback.i wonder why nsc have not suggestion current mode.

anyway,the sch file attached.i will try it weekend i hope great.

ZANG

[sawreyrw]

The gain of this circuit will be approximately 1+Rspeaker/R5. Is that what you want?

[digi01]

yes,my bookshelf is 4ohm.

by the way,the R4 can be ignored.it is going to extremity simple.

[janneman]

Originally Posted by digi01

yes,my bookshelf is 4ohm.

by the way,the R4 can be ignored.it is going to extremity simple.

Your bookshelf is nominally 4 ohms. This value will vary wildly over the audio band, perhaps from 2 ohms to 20 ohms. That means your gain, and thus your final frequency response, will also vary wildly over the audio band. I guess you don't want that?

In most designs with this kind of current drive the speaker is a wideband driver without xover that has a relatively flat impedance response, or with more complex speakers some kind of impedance or other equalisation is used.

jd

[teemuk]

...Or guitar amps where commonly mix of both voltage and current feedback is used and the unlinear frequency response is sought after because it's closer to characteristics of generic tube amps with high output Z.

[KSTR]

This type of hookup can be a hell to get stable, with an amp like the LM3886 which needs noise gain > 20dB.

For example any capacitance of the speaker wire would pose a problem here, or a shorted speaker wire

Normally full current-type output (not current-feedback, but we have discussed this naming conventions before ) is not really needed, a high enough Zout is sufficient and more stable to implement, see ESP project 56, mixed-mode feedback:
http://sound.westhost.com/project56.htm
Still it would need some compensation for the LM3886 to remain stable under all conditions.

If true current output is wanted, there are better circuits which are *not* based on schemes where the load alters the operating conditions for the opamp.

- Klaus

[relder]

I've been experimenting with the circuit below. The pot makes the output impedance adjustable. Thus far it seems to work OK. No oscillations visible on the scope. Measuring speaker distortion, it's kind of cool to adjust the output impedance knob and watch distortion go down as you bias it toward a current source. I have one built, I'll be making an active 3-way, so only 3 more to go (2 LM3886 for the woofers, 2 LM4780 for the mid/tweet)

[CLS]

I'm glad I found this and no need to start a new thread - after all LM3886 threads are already too many....

The diagram in the 1st post look strange to me. Don't know how it turned out since there's no update over a year.

I try mixed mode (both voltage and current) feedback on this chip and it works That's according to the ESP site (mentioned above in #6 post by KSTR).

As I brought up in another thread, I need about 20 Ohm output impedance to make the system Q= 1.1 to my OB bass (with a woofer of 0.29 Qts).

So, my setup is:

Rf1 = 91k
Ri = 1k
Ci = 220uF
(look figure 1 of the datasheet for their positions)

and Rsense = 0.22 R (in series with the load, to pickup current signal)

I built this prototype with a mindset of building a (breadboarded) tube amp which I'm more familiar with. I use unregulated PSU with good old CLC filter which is seldom seen in the world of chip amps.

The power trafo I have on hand is 28V output, so it's approaching the limit of the chip. I got +/- 41.5V after the filter, just +/- 0.5V shy of the line. The load is a woofer for under 240Hz, the measured impedacnce is ranging from 8.2 Ohm to 200 Ohm (fs peak), so I guess I'm OK with that voltage.

There're so many beautifully and tidily made chip amps around here, just can not do that, I'm almost too embarrassed to post the pics of mine. Ah~ what the heck, never mind, here it is.


Topless MDF chassis with plywood front panel:

Overkill power supply - my style Unmatched power filter chokes because those are what I have. With very little DCR, they don't affect the rail balance, at least at idle. Perfect +/- 41.5V on both channel


The chips run hotter than I expected, so I added some more heatsinks on the back:

However they help very little. The temperature is about high 40s degree C. Very warm to touch, obviously higher than body temperature but OK to touch it continously. Eventually I added a small fan to cool it down to my like. I think wood chassis is very much inferior to metal in loosing heat. (my other tube amp suffers from this, too)


The boards (bought from local web auction seller).

You may see the 'add-on' parts -- standing but tilted 2W resistor is the Rsense, and the 'floating' cap is the Ci.

Overall a messy mockup but it worked fine in the first 2 hours test (and the bass performance is very good). No audible hum or noise, turn on and off quietly. 9mV output offset. (all these are much better than my old Hafler)


Look much nicer if step back a little

(The middle one in the rack)

[CLS]

Those chokes are no good in this low quiescient current situation. They slow down and round off the sound I took them out after 2 days listening.

Without the chokes, power rails are now +/- 41.7V. The sound is more lively and agile.

[AndrewT]

3886 with +-41.7Vdc is just inside spec limit.
What do your rails rise to when mains is at maximum supply tolerance?