RC Filter delay with PWM to DC

[brownt]

Hi, I am using Autocorrelation in an embedded system to determine audio frequencies 80 - 1400Hz.

After correlation I am using PWM to represent that frequency then passing it through an RC filter, and using the resulting DC as an input to an A/D on another system/enclosure via a 15cm cable.

I am using 1Kohm and1uF for the filter, which according to my calculations should settle at 90% within 1mS. However, it seems to be taking much longer. I can change it to 100uF which should be 100mS settle time, but it doesn't seem to make any difference. It is all part of an audio system and there is a noticeable audible delay.

I can visually see the slow DC change on an oscilloscope compared to a much faster visual change on the PWM pin.

And I wonder if the impedance of the A/D is a significant part of the equation that I am missing, or something else.

thanks

 

[KlausST]

Hi,

I am using 1Kohm and1uF for the filter, which according to my calculations should settle at 90% within 1mS

review your calculations.
1kOhm and 1uF form afilter with tau of 1ms.
But one tau is only 63%.
You need 2.3 tau = 2.3 x 1ms to get 90%

100uF will make it much slower. I assume this is not what you want.

If you need low ripple and faster resoponse you need to go for a higher order filter.
Example: (raw estimation, not calculated)
you have a 10kHz PWM and want a ripple of less than 1%
Then for a first order filter you need a cut off frequency of about 10kHz / 100 = 100Hz. Tau = 1.6 ms.

Now with a second order filter you need to set cutoff frequency to about 10kHz / (100^(1/2)) = 1kHz. (where the "2" represents the filter order)
Tau is about 2 x 0.16ms = 0.32ms.

Expect a 5 times faster settling response.

In detail settling time depends on filter characteristic.

I recommend to use some simulation software to simulate your circuit.
Even Excel could help.


Klaus

 

[brownt]

thanks for that. So instead of 2.3mS, the settling time with be 1/5th of that. hmmm, I guess the problem is somewhere else then as 1 - 20mS is not significant for an audible delay. Back to the software. thanks

 

[KlausST]

Hi,

Now I'm a bit confused.
You say audible delay... but I refered to "PWM to DC".
But "DC" is not audible. Maybe I misunderstood the problem.

Maybe post a sketch of the signal flow.

Klaus

 

[brownt]

The DC is fed to the A/D on another system. The digital is used to manipulate some audio. There are various delays within the system. But the biggest one, now I know is software. There is a little in the RC filter, and some in the remote system, but mostly there is about 100mS in software as the autocorrelation is executed.

thanks

 

[KlausST]

Hi,

instead of "mS" you should write "ms",
because
* an "s" means seconds
* an "S" means Siemens = A/V= 1/Ohms

Refer to:

Klaus