LMF100 as a variable fitler - trying to smooth the output

[atferrari]

A square wave ranging from 10 to 1000 Hz is filtered with an LMF100 (switched capacitors filter - bandpass) to obtain a sine wave.

Using fCLK/f0 = 100:1, fCLK goes from 1000 Hz to 100 KHz.

Basically it works, but, to smooth the step-like output, I intend to use a single lowpass RC cell as the datasheet reccomends.

Problem is that an RC filter able to stop the 1000 Hz (fCLK) at the lower range is affecting my output at the upper end (1000 Hz as well).

I feel I am going in circles loosing the advantages of a continuously variable filter as implemented.

Still in square one and perplex, wondering how the pros do it.

BTW my idea is not to start switching things along the range. Oh no!

Ideas anyone?

 

[ronsimpson]

Sign wave output is 10hz to 1000hz.
The clock is 1k to 100khz.
I have not used a band pass filter to make a sign wave from a square wave. I have used low pass switched cap filters to do this.
One problem is that the clock at low frequencies is the same as the output at high frequencies.
Add a RC low pass filter. If you add a 2khz low pass filter to the output, (filter that passes 1000hz output but blocks the clock) it will work well at high frequencies. As your output nears 20hz the clock is at 2khz and the new low pass filter will not remove the clock.
I see you don't want to have three low pass filters to remove the clock. I would add a R and C filter and switch the C at three points.

Another idea is to make the first filter using the LMF100 in band pass or low pass mode to get the sign wave. Then add a second LMF100 running at 10x or 100x faster to remove the first clock. Now add the RC filter to remove the last clock.

Look at the MAX297. It is also used to do this job. Simple 8 pin IC.

 

[atferrari]

Maybe I have to switch caps

In spite I do not want to do it, more and more I think of it, I know that I should have to switch caps in and out.

I recall switching caps for an 8038 (?) years ago using just NPN transistors between the caps legs and GND.

Do you think that would be correct, strictly speaking?

KISS principle seems gone with the wind here...

 

[atferrari]

Problem solved

¡Gracias Ron!

Problem solved... with the KISS principle still intact.

I implemented a single-pole RC filter for the upper end with:
R = 39 K
C = 0,0033 uF (permanent)

Later, I programmed the micro in charge to switch, via NPN transistors, additional caps at two convenient frequencies, in parallel with the "permanent" one.

At 307 Hz C = 0,01 uF (added in parallel to the above)

At 36 Hz, C = 0,1 uF (added in parallel to both above).

Today is a nice day! A saga of false starts, wrong designs, lack of adequate testing and ignorance of basic concepts, lasting several years, is over.

 

[alec_t]

Hola, Agustin.
Always good to hear a success story!