Composite amplifier: LM3886 + LME49710

[U130421]

I don't have any experience with 24 bit , but with 16 . To get lowest closed loop distortion you must output the sine wave with the highest level to be described by 24 bit , then attenuate by resistors to get -3db FS . May be there is a minimum to search for .

 

[zarandok]

The Xonar D2X is a relative good card (I've it too) with PCM1796 dacs. But: the analog on-board regulator is a 7805 (!!!), and there are some lower quality ceramic caps in the signal path wher they shouldn't be... I've updated them to Elna Silmic and TPA7S4700...

 

[zarandok]

Just let me please one queston what I do not understand:
The R23, R25... resistors with 1k (parallel with + input of LM3886) seems to be very low value for me!!! With simulations I got the best parameters with ca. 10k by this position!
Anyway, in the case of 1k you have a huge signal voltage drop..

 

[jeanlemotan]

For measuring a bridged amp, you need a differential input.

Tom

After the 2nd round of resistor smoke, I figured this out
I'll build a differential probe with a LME49710, 9V batteries and hand picked resistors and that's it.

 

[jeanlemotan]

Just let me please one queston what I do not understand:
The R23, R25... resistors with 1k (parallel with + input of LM3886) seems to be very low value for me!!! With simulations I got the best parameters with ca. 10k by this position!
Anyway, in the case of 1k you have a huge signal voltage drop..

Those resistors together with R6 and R31 form 2 voltage dividers (R6+R25 and R31+R23) that is meant to allow the LME49720 to clip before the 3886. Like this I get much better clipping behavior than if the 3886 would clip.
There was some discussion about in earlier posts (Composite amplifier: LM3886 + LME49710)

 

[zarandok]

Thanks. What is the simulated THD+N of the whole cicruit? Could you show me a step response with about 500ns/div resolution?


Tom is using by this positions 1k-330R. Pretty low...

 

[jeanlemotan]

Thanks. What is the simulated THD+N of the whole cicruit? Could you show me a step response with about 500ns/div resolution?


Tom is using by this positions 1k-330R. Pretty low...

The first post has the simulated step response and THD+N. Note that the TINA model doesn't include the LM3886 THD so the simulation for that is not relevant.
I got a good sound card to measure this but I'm still learning about how to calibrate the whole thing in REW...

 

[jeanlemotan]

Just finished analyzing some NE5532 opamps that I got in a preamp. They seem to be fake, LM358 in disguise...
Here's the thread here: Fake NE5532

 

[zarandok]

Ok, really. But I can not see the simulated THD values still...

 

[1audio]

I don't have any experience with 24 bit , but with 16 . To get lowest closed loop distortion you must output the sine wave with the highest level to be described by 24 bit , then attenuate by resistors to get -3db FS . May be there is a minimum to search for .

Actually all the DAC's and ADC's I have tested are better at a reduced level. Typically the best distortion is at -15 dB. This brings a penalty in SNR but for distortion testing a good FFT program will get you past that limitation. The best I have used for this is: DiAna, a software Distortion Analyzer DiAna Index of /diana/

 

[seanny]

Have you built your differential probe?
Is there a lot of distortion added to test, or your software can calibrate with your probe to zero out distortion?

 

[sid12345678910]

Did you manage to get some measurements of the amp?

Also keep up with the YT channel very interesting!

Thanks for doing all this!

 

[dannybeckett]

bump

 

[simonra]

Apologies for resurrecting an old thread but I was looking at the simulation schematic back in post 18 and wondered whether the loop is broken in the wrong location. Shouldn't Vo moved to the output of U8 followed by L1 and the injected signal?

 

[DaveFred]

Any update on how the project is coming?