LM3886F, discreet with MOSFET outputs

[audio_tony]

I use Tina quite extensively myself. I do quite often find the Fourier analysis to be somewhat inaccurate when compared to actual 'real 'life' THD measurements.

It does provide a fair indication of what to expect though.

I look forward to seeing your final build with measurements.

 

[Hayk]

You need to adjust the time just where the phase of the fundamental is 90°.
I checked many times with LTSPICE, it is always pessimistic but it gives directive if better or worst.
I'll make on LTspice tomorrow.
If you are seriously interested with this project, I can put it in priority, it can be ready at the end of next month.

 

[Hayk]

Ltspice model. I didn't find A970 working model. It measures 0.007% with 5401.
With Tina model it is 0.006%

 

[Hayk]

I added offset adjust.

 

[Hayk]

Now I finished with a good sounding amp, I will make it good measuring one.
I will implement an error corrector shown bellow.

 

[Hayk]

There is a crossover problem which is limiting the THD. It occurs on the falling.
This is what the differential sees.

 

[Hayk]

Problem resolved. I eliminated the diode between VAS and lower mosfet. I don't know for what really it was there for.

 

[solderdude]

Problem resolved. I eliminated the diode between VAS and lower mosfet. I don't know for what really it was there for.

In the orihinal version it is for current limiting. In your design it won't work properly.

 

[Hayk]

It has to do with thermal drift. Now I need to add a thermistor parallel to the VAS emitter resistor to decrease the voltage by -4mv/°C.

 

[Hayk]

I still have a bottleneck limiting to 0.0015%. Both LTSPICE and Tina shows the same. With Tina with such thing, the generator gets distortion.

 

[Hayk]

I will leave it for now until I get better view, where is the bottleneck.

 

[Hayk]

The spectrum of input and output, only noise.
I will implement the EC and let the user decide to wire it.

 

[Hayk]

This is the EC. I made it with 3 transistors. Look how the error voltage decreases with EC by 13 times. The distortion is at minimum of measurement.0.0015%.
For now the operating point is adjusted, I will try to make it self adjusting.

 

[Hayk]

To see the distortion, I subtracted the output from the input to magnify the THD.
Here it is magnified 17000 times. The real THD is 0.00015%. The EC will not do much.

 

[Hayk]

The Lm3886 has input buffers, I also added but jfet ones. The distortion and noise are lower. For 22.7v output, the error of 1.4mv has 1.5% THD. This is 0.00009%. It liberates the offset problems for input impedance.

 

[Hayk]

Taking final shape. I added adjustments.

 

[Hayk]

With jfet inputs, it is easy to make differential. It has a strap that needs to be selected either single or differential input. The gain remains the same in both cases. I checked the distortion for single, it is the same as before.

 

[solderdude]

Maybe it is time for a new name ... doesn't look anything like the LM3886 chip anymore

 

[Hayk]

The structure is the same, the external components of application are different.
What do you propose?

 

[Hayk]

TI called amplifiers with such output stage ,Overture series. Overture is a trademark. Maybe I call, Fet Overture or F.Overture.
Overture, in French means opening, and Fête, pronounced FET, means festive. I can call it Ouverture en FET.