How to calculate settling time of LPF?

Thread Starter

picstudent

Joined Feb 3, 2009
91
I need to monitor 12 nos of LM35 outputs in sequence. So I am trying to read the ADC in a round Robbin fashion.

I have LM35 outputs directly ( a RC LPF is there anyway!) goes to multiplexer, CD4065. then it goes to A Opamp based sallen key butterworth 2 pole filer and to PIC ADC. I plan to use this Opamp LPF because LM35 is mounted using 1.5 meters length shileded cable..

But the settling time of the filter will cause my sampling to be done at slower rate. High speed is not required anyway. How to calculate the settling time of a LPF filter to reach a optimum time?

thanks
 

t_n_k

Joined Mar 6, 2009
5,455
Have you set a Q value or damping factor for the filter design? Is it an under or over-damped response?

Response to what?

Perhaps you maybe consider the filter response to a step input - probably the worst case - and use a defined settling criterion as your benchmark waiting time.

What is the settling criterion? Not sure ... probably subject to your requirements and possibly a compromise.

Maybe the time it takes from the aforementioned step change to the point when the output disturbance is within 1% of the steady state output.
 

hgmjr

Joined Jan 28, 2005
9,027
If you have the luxury of time and if the expected "rate of change" of temperature on any given channel is not great, you could consider taking mulitple samples from a given channel until the "sample to sample" change on a given channel is within some pre-defined delta temperature. Once the temperature settled within the error band then you could use it as the temperature reading from that channel.

hgmjr
 

Thread Starter

picstudent

Joined Feb 3, 2009
91
..until the "sample to sample" change on a given channel is within some pre-defined delta temperature. Once the temperature settled within the error band then you could use it as the temperature reading from that channel.
hgmjr
Got the idea, but I feel it will make things or code more complicated for me.


But if I get an idea how much time it will take a 20hz or 100Hz LPF to settle, I may be able to draw a clearer picture of its pros and cons in my case.

Thanks for the response
 
I know this sounds odd, but many lowpass filters don't like DC on their inputs. You should get away with it with a unity gain Sallen & Key lowpass, but of you're using the variant with the damping set by the 2 Rs between the op-amp output and the inverting input then any DC on the input is going to be multiplied by (1+Rbottom/Rtop).

If the filter can cope with a DC bias on the input, then it's a question of checking the step response. Luckily these (normalised) responses are widely published - here's a great tome from Walt Jung (link) - check page 5.28

You might be better off with a Bessel response (damping factor ζ = 0.8659, compared to the Butterworth ζ of 0.7071) as, counter-intuitively, it has a faster step response than the Butterwoth, and no overshoot.
 
Top