Combing amplifier circuit with low-pass filter using LM358 frequency response messed up

Thread Starter

William Gomes

Joined Jul 10, 2019
6
Hello, I have a problem and until now I was not capable to solve it.
I designed two circuits. The first one is to receive the audio signal from a mic and amplify the mic. And the second one is to filter the high frequencies using a low-pass filter. When I simulated them individually the frequency response work as expected, but when I simulated together the frequency response is a huge mess.

separeted.png

And the last one occurs when they are combined.

together.png

Can someone give me a clue, I can't understand why that happens.
 

Attachments

crutschow

Joined Mar 14, 2008
34,280
Here's my take on the circuit:
I used the similar LM324 models I had.

Look closely at the value of R8 in your combined schematic to see the problem.

I changed the circuit so both op amps use the same offset virtual ground.
This eliminates the need for C2 and R5.

Note that if you use 1VAC for the source voltage, the output is then given in dB gain, since the dB values are referenced to 1Vac.
There's no concern about overdriving the circuit, since the AC analysis uses linear models that are insensitive to signal levels.

1597701801674.png
 
Last edited:

Irving

Joined Jan 30, 2016
3,843
In your original circuit the input to the filter is 2.37k. In the combined cct its 2.37ohm? So 60nF C4 is having big impact on input stage...

edit: crutschow types faster than me!
 

Thread Starter

William Gomes

Joined Jul 10, 2019
6
Here's my take on the circuit:
I used the similar LM324 models I had.

Look closely at the value of R8 in your combined schematic to see the problem.

I changed the circuit so both op amps use the same offset virtual ground.
This eliminates the need for C2 and R5.

Note that if you use 1VAC for the source voltage, the output is then given in dB gain, since the dB values are referenced to 1Vac.
There's no concern about overdriving the circuit, since the AC analysis uses linear models that are insensitive to signal levels.

View attachment 214996
I really appreciate your help, now I understand the problem and the improvements done. And thanks for the tip for the source voltage, I will use for other simulations.
 

Audioguru again

Joined Oct 21, 2019
6,671
NOBODY uses a lousy old LM358 dual or LM324 quad opamp for a mic preamp because:
1) It is very noisy (hisssss).
2) It has a slew rate so low that it has trouble producing frequencies above 2kHz.
3) It produces crossover distortion.

Why is your filter cutting all the important high frequency consonant sounds in speech? The conversation will be, "What did you say? What did you say?" over and over.
 

ci139

Joined Jul 11, 2016
1,898
using unlabelled nodes may be confusing . . . as you change your circuit the node-numbers may be re assigned so your graph goes for not the initial target anymore -- put the node label on what you want to plot or -- confirm each time your plot corresponds to grid
 

Thread Starter

William Gomes

Joined Jul 10, 2019
6
NOBODY uses a lousy old LM358 dual or LM324 quad opamp for a mic preamp because:
1) It is very noisy (hisssss).
2) It has a slew rate so low that it has trouble producing frequencies above 2kHz.
3) It produces crossover distortion.

Why is your filter cutting all the important high frequency consonant sounds in speech? The conversation will be, "What did you say? What did you say?" over and over.
Thank you for your response. I'm using that particular one because I want something low-cost. Do you recommended something else? And I would like to cut 6kHz above because I'm working with 4 array mic and I have to split 48kHz sample rate into those 4 mic. That's why, but I do not need something really good, the purpose it's for embedded system
 

Thread Starter

William Gomes

Joined Jul 10, 2019
6
using unlabelled nodes may be confusing . . . as you change your circuit the node-numbers may be re assigned so your graph goes for not the initial target anymore -- put the node label on what you want to plot or -- confirm each time your plot corresponds to grid
Thanks for the tip
 

crutschow

Joined Mar 14, 2008
34,280
I think AG has previously recommended the TLO7x series, The MC3317x series, and the MC3407x series for low noise audio applications.
 
Top