ultra low frequency low pass filters.

Out = Out + (In-Out) / K

If that settles too sloppily for your taste, just do a 2nd or 4th order version of same.

John

On an ultra-cheap processor that's

Out = Out + ((In - Out) >> S);

Just make sure that your variables have at least 16+S bits.

size in X7R).

Why limit yourself to ceramic caps? [Metalized] polyester and polypropylenes are available up to 10uF and there's no problem at all with using multi-megohm resistors if you want. FET op-amps (like you'd use with high impedances like this) have a bigger offset voltage but you seem vague as to what your offset requirements are. (You're also vague about the temperature requirements, starting with 125C but then complaining about Polymimide PCB's at 160C).

But your option 2:

seems like a no-brainer winner, because your acquisition rate is so glacially slow and you're going to A-D the result anyway.

Tim.

Right; I usually do it that way. And I usually throw away a couple more bits so I don't have to worry about 2's comp rollovers and stuff, which always hurts my head.

What's interesting is trying to do a higher-order filter using only shifts to make the coefficients. You get some strange unorthodox transfer functions.

John

As I've po Using integrators of the form 1/(p+1) eliminates problems with capacitor dissipation factor at low frequencies. (p = s/(2*pi*Fo), normalized Laplace variable)

However this means using a state-variable-like circuit structure.

See "StateVariableFilter(P+1).pdf" on the S.E.D/Schematics page of my website for details.

...Jim Thompson

Hey all,

When designing sub 1Hz filters for a very slow moving signal that can have noise on it from 1Hz and upward, what is the safest way to achieve low DC offset and only mild shifting of the corner frequency? Let say the temperature range is from 0 to 125C.

The signal is a 0 - 5V analog signal analog signal being read by a 16 bit a2d.

looking at two way of doing this.

2) DSP filter, the antialias filters will have similar but reduced affects of the straight analog option with teh only advantage being the lack of crossover point drift. The increased performance of the "brickwall" filters will not have a big benefit.

Any other circuit topologies or tricks to making this thing stable?

Since you didn't say you wanted your result in digital form, I will assume that you really need an analog output.

One of the problems with the ADC-DSP-DAC method is that the DC offset in the converters can add up to a fair amount. A trick for dealing with this is to make an integrator with a very low gain and feed it the difference between the input and the output. So long as the integrator's gain is way down before the phase shift of the filter gets big, the exact gain of the integrator doesn't matter, you can use it to tune out the offset. This lets you get away with inexact parts and yet have an exact cut off.

A DSP or even a micro controller making an IIR filter can have a very sharp cut off without too much cost in hardware. You may want to follow the DAC with an analog filter and dither it to increase the effective number of bits. Since the frequencies you want to do are very low, 24 bits worth of DAC can be bought so you should be able to make 26 or 28 bits worth of results.

You alerady jave excellent responses concerning filtering. The PCB material you mentioned can be replaced with much less expensive PCB material made by Matsushita; see the PCB reference on the

Thanks guys,

As I learned more about filters I found a IIR filter that would do the trick, was stable and simple to impelemnt. I could use a PIC if I wanted to since the signal is so SLOW.

Thanks for all the imput.