New LM3886

[mravlca]

Hi all!

I've decided to go a little bit different on this one, because I've heard one LM3886 that is built on this topology and it sounds awesome. So can you tell me if this will work or if I can ran into some problems. There is also current source on the output, made with LM317, now rated at about 430mA, but it will be lowered to 100mA.

Regards Aleš

[Tekko]

I don't see the point of the LM317 current source other than making a huge dc offset on the output.

[mravlca]

Originally Posted by Tekko

I don't see the point of the LM317 current source other than making a huge dc offset on the output.

Why would it make huge DC offset? Anyway can I omit feedback capacitor if I am going to use input cap?

Regards Aleš

[ampimp]

The input cap blocks dc comming from your source. The feedback cap blocks dc created in the amplifier.

[epitaxial]

What does it use for a power supply?

[AndrewT]

post1 shows a schematic with 10off GND.

Do not connect these without thinking about what each does and what signal each passes.

[ashok]

Originally Posted by mravlca

Why would it make huge DC offset? Anyway can I omit feedback capacitor if I am going to use input cap?

Regards Aleš

The LM317 circuit is apparently to drive the output stage into class A at low levels. The LM3886 sounds quite clean at low levels and so you might skip that part or make it on a smaller ( daughter) pcb and removeable if you find it doesn't really make a diff.
Wonder what the idea was as the current will only affect one output transistor.

I remember trying the amp without the feedback cap. DC offset went up to about 0.1V or thereabout. But it worked fine ....better sound without the cap. I did use a good film cap at the input.

Your feedback resistor ( 1 K ) is much lower than the input resistor ( 22K). Assuming equal bias currents in the input stage you might get a greater differential in the voltage drop between these two resistors and cause a higher offset at the output. Do check it with 1K and 22K . ( 47 ohms will also have to change to about 1 K to keep the same gain ).

[mravlca]

Originally Posted by ashok

The LM317 circuit is apparently to drive the output stage into class A at low levels. The LM3886 sounds quite clean at low levels and so you might skip that part or make it on a smaller ( daughter) pcb and removeable if you find it doesn't really make a diff.
Wonder what the idea was as the current will only affect one output transistor.

I remember trying the amp without the feedback cap. DC offset went up to about 0.1V or thereabout. But it worked fine ....better sound without the cap. I did use a good film cap at the input.

Your feedback resistor ( 1 K ) is much lower than the input resistor ( 22K). Assuming equal bias currents in the input stage you might get a greater differential in the voltage drop between these two resistors and cause a higher offset at the output. Do check it with 1K and 22K . ( 47 ohms will also have to change to about 1 K to keep the same gain ).

I've used LM317 on the output so that the circuit stays simple and you can use it with different voltages without changing anything, if I will have any problems with LM317 then I will try with transistor, BD139 maybe. The feedback resistor is low because, as said before, I've heard one with this kind of values and it was really neutral and of the best I've heard. For the start here is the first PCB drawing I made today with input cap. If anyone has any idea on PCB improvment, hit me, before I make one in couple of days.

@AndrewT: Maybe I should put resistor between signal GND and GND, but I will see if the amp will be dead quiet, then I will leave it this way.

Regards Aleš

[Gopher]

Maybe you should read Doug Self's work on distortion in amplifiers where he shows that a richly biased Class AB output stage like you are making here actually has higher distortion than a plain old Class B output stage due to gm doubling effects.

By biasing the chip like this you intuitively think that you're getting a bit of Class A performance, but you're not really.

[abraxalito]

I'm not convinced that resistive loading the output gives rise to gm doubling. That occurs when the bias current is fed down through both transistors in the output stage. Here, only one transistor sees the extra current, not both. The output stage of an LM3886 is quasi-comp anyway, can gm doubling apply to such stages?

Your point about higher distortion may well stand nevertheless, its unwise to increase the loading of a chip amp, they're already pretty heavily challenged in the thermal dept.