I have a working model on the bench, the simulator guides me what can be done but not what should be done.
When something goes good in simulation, it can only be simulation fantasy, when it goes bad it can only be the circuit.
Which parameter of the DS is not accurately modeled?
Look at at Neochrome website, there is only simulation.
The load current do you mean be infected by the power supply noise?
I am reviewing from the start the stability of the 2 loops, something much better to come.
When something goes good in simulation, it can only be simulation fantasy, when it goes bad it can only be the circuit.
Which parameter of the DS is not accurately modeled?
Look at at Neochrome website, there is only simulation.
The load current do you mean be infected by the power supply noise?
I am reviewing from the start the stability of the 2 loops, something much better to come.
I found a nice way of replacing the AD829 with NE5534 in error subtracting circuit. I reduce the gain of the opamp by using the compensation path which gives an internal VAS feedback. I get 70db NFB @20khz.
C1 is for the sake of simulating correctly. I just adjusted R8 to 33k for perfect square wave.
C1 is for the sake of simulating correctly. I just adjusted R8 to 33k for perfect square wave.
Attachments
Post in thread 'LM3886 Stereo Amplifier and Power Supply Design' https://www.diyaudio.com/community/...r-and-power-supply-design.409952/post-7622730
Estimating the performance of the error corrector and the error subtractor.
The 3886 alone has a noise at 10mw 0.03% with 8 ohm, hence with 4 ohm it gets the same at 20mw because the output voltage becomes the same as 8ohm 10mw. It is nearly the same THD+N for 1w 8ohm and 2w 4ohm. In other words, the noise is the main value in THD+N at this range.
With Modulus-86, the story is rather strange.
The noise is reduced to 0.006% for both 4ohm and 8ohm which is 5 times less for 8 ohms than normal 3886. But strangely, the noise is 1.4 times lower if 4 ohms load as it has the same percentage, doubtful isn't it? On the other hand the max power for 4 and 8 ohm are nearly the same, is there 4-8 ohms switch?
If I compare on simulator what I get, the LM3886 alone has 3.5mv output noise for 20h-20khz BW. With error subtractor it is 80uv, with error corrector, it is 24uv 1k and 50uv 10k resistor for volume control.
This means I get 42 times less noise with error subtractor and up to152 times with error corrector.
The 3886 alone has a noise at 10mw 0.03% with 8 ohm, hence with 4 ohm it gets the same at 20mw because the output voltage becomes the same as 8ohm 10mw. It is nearly the same THD+N for 1w 8ohm and 2w 4ohm. In other words, the noise is the main value in THD+N at this range.
With Modulus-86, the story is rather strange.
The noise is reduced to 0.006% for both 4ohm and 8ohm which is 5 times less for 8 ohms than normal 3886. But strangely, the noise is 1.4 times lower if 4 ohms load as it has the same percentage, doubtful isn't it? On the other hand the max power for 4 and 8 ohm are nearly the same, is there 4-8 ohms switch?
If I compare on simulator what I get, the LM3886 alone has 3.5mv output noise for 20h-20khz BW. With error subtractor it is 80uv, with error corrector, it is 24uv 1k and 50uv 10k resistor for volume control.
This means I get 42 times less noise with error subtractor and up to152 times with error corrector.
Attachments
Last edited:
The ZD-50 circuit uses stabilized power. When unstabilized, the parameters are worse.