LM3886 AS POWER BUFFER

[Stee]

look this project
have a special addition
a pseudo bridge with dummy load

seemingly useless
works better then classical balanced connection

 

[Michael Bean]

Have you looked at the LM3886 data sheet? It requires a minimum gain of 10 (20dB), and that output arrangement isn't a bridged connection, it's paralleled and it won't work right with only one resistor.

Mike

 

[Arty]

aye, that is not going to work.
ever.

you would at least need a poweropamp (is that even a correct term??) that is unity gain stable...

 

[woody]

Apex has such a thing but I don't think you want to know how much it cost !


https://www.apexanalog.com/apex-products/pa50/

 

[AudioLapDance]

See the unity gain chipamps here:




http://www.diyaudio.com/forums/chip-amps/250263-improving-lm3886-amplifier-21.html#post4167380

http://www.diyaudio.com/forums/chip-amps/250263-improving-lm3886-amplifier-22.html#post4167711



Or better perhaps, the TDA7293/4 has a pre gain stage that then connects, through external connection, to the power buffer stage. I wanted to try this but I remember some issues ...

Cheers

 

[tomchr]

The circuit in Post #5 reduces the overall loop gain to make the amplifier stable. While that certainly is a valid approach, you'll get much better performance by using noise gain compensation. NG compensation allows you to maximize the loop gain at LF while rolling it off beyond some cutoff frequency. All that's required is an RC circuit between the inverting and non-inverting inputs of the opamp/LM3886.

~Tom

 

[Stee]

All that's required is an RC circuit between the inverting and non-inverting inputs of the opamp/LM3886.
Thanks to all friends

 

[Stee]

That can work

maybe

 

[Arty]

mmm please, tell me again why is this good in the first place ?

 

[tomchr]

Um... You may want to re-think that 15 ohm on the output.

Here's a good app note on op-amp compensation: http://www.ti.com/lit/an/sboa015/sboa015.pdf

~Tom

 

[nigelwright7557]

Bridged is much better to parallel.
Put output1 to one side of speaker and output 2 to other side of speaker.
No 15 ohm resistor required.
Have both LM3886 none inverting.

 

[tomchr]

Bridged is much better to parallel.

I'd say each has its own advantages. Bridge offers more power, but each amp half needs to be designed for half the load impedance (--> bigger heat sink, higher output current needed). Parallel offers twice the output current and the ability to drive low load impedances, but does not increase the output power. If you want the benefits of both, go with a bridge+parallel solution...

Tradeoffs, tradeoffs...

~Tom

 

[nigelwright7557]

Great care with parallel has to be taken to match output resistors.
Also the same goes for gain resistors.
If there is even a small difference the amplifiers can start fighting each other.
If somehow one amp gets a glitch and the other doesn't they can again fight each other.

 

[benb]

Bridged is much better to parallel.
Put output1 to one side of speaker and output 2 to other side of speaker.
No 15 ohm resistor required.
Have both LM3886 none inverting.

I'm thinking the OP needs more education than a simple bridged vs. parallel discussion.

 

[Arty]

i think bridge and parallel will both give the same power, just to different loads.

but with prallel config i think it is harder to work with, requires more precision and better matching.

 

[jcx]

given that single chip amps are designed for pretty standard consumer loudspeakers you really need to parallel when you bridge
unless you are driving your own creation with say individually amped drivers then you can match chip amp I, V capability more closely to the better known load, say 4 Ohm woofer vs 16 Ohm CD

 

[Arty]

not really, an 8 ohm speaker is safe to drive from bridged lm3886, at+-28 V rails.
no different than 4 ohm load with a single lm3886 and +-28 v rails.
16 ohm is absolutely okay with bridged config, at +-35 v rails.

yeeehaa starts when you have a 2 ohm speaker...
then paralell will be a logical choice, question is why would you have a 2 ohm speaker at all.

OR, if you plan on using +-35 rails and drive a 4 ohm load. then parlelling is a good idea.

or if you need more juice, then go BPA. witch is btw totally pointless with lm3886.
simply choose a different IC that suits the job better.

 

[jcx]

single drivers aren't so bad as loads - but dynamic systems aren't limited to simple I=V/R

certain patterns of signal - essentially "pumping" a resonance then reversing the phase can draw more current than the voice coil R would lead you to believe - 2x is a minimum factor for dynamic current peaking

that makes a 3886 "4 Ohm" drive spec really mean its just barely competent for a 8 Ohm nominal dynamic driver directly driven


multiway speakers can have poor crossover designs that make the problem much worse Self, Crodell's Books both cite large worst case overcurrent demands of >5x the loudspeaker's "nameplate" impedance

 

[AndrewT]

............multiway speakers can have poor crossover designs that make the problem much worse Self, Crodell's Books both cite large worst case overcurrent demands of >5x the loudspeaker's "nameplate" impedance

This Forum has a post with a link to a (three different) speakers test that showed more than 5times the current passing to the reactive speaker load than what the rated speaker impedance would indicate for a resistor/dummy load.

Reading that test report made me change MY design procedure from allowing double the output current to triple the output current.
i.e. a 100W into 8ohms meeds to drive 40Vpk and 5Apk into a 8r0 resistive load AND must be able to drive 15Apk into a reactive load, without the PSU or the amplifier collapsing.

 

[palstanturhin]

i.e. a 100W into 8ohms meeds to drive 40Vpk and 5Apk into a 8r0 resistive load AND must be able to drive 15Apk into a reactive load, without the PSU or the amplifier collapsing.

How long would this peak power requirement take (current pulse width)?