1
\$\begingroup\$

We have several types of data acquisition devices where I work. They're used primarily to measure low frequency voltages (<100HZ, +/-10V). One device has a built in filter that does an excellent job. I asked the manufacturer for details about the filter. They would not elaborate on the specific design of the filter, or where to purchase one. Here's the way they describe the filter in an email:

"the filter before the ADC is a 2nd order RC filter with 90dB at approximately 128kHz"

The user manual describes it simply as a 50/60 Hz filter.

To my surprise I'm having difficulty finding such a filter using Google. Can anyone tell me how to construct such a filter or where to purchase them? Ideally it would be a module with BNC connectors so I could simply attach it inline with the signal cable to our other data acquisition devices.

As a mechanical engineer I have little knowledge about filters. However I did try connecting a 2000uF capacitor across the test leads. To my surprise this actually helps a lot to remove noise. It is an electrolytic type which I understand is polarized and only to be used with positive signals. I haven't found any non-polarized types of capacitors with that much capacitance. But 20000uF seems like way too much. I want to find a more appropriate solution.

How do I design an appropriate filter for my application?

\$\endgroup\$
  • \$\begingroup\$ Signals and noise depends greatly on cable quality and ground connections and proximity to transient noise current. Can you use Coax earth-grounded at ADC? the manual link might be helpful if you follow cable advice. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 28 '18 at 17:21
  • \$\begingroup\$ What is a) your budget, and b) your sample frequency? Is this an anti-alias filter? Do you need many matched filters? \$\endgroup\$ – Scott Seidman Jun 28 '18 at 17:39
  • \$\begingroup\$ You havn't said what you are trying to filter out. 50/60Hz is usually removed by careful choice of sample rate to cancel it out, differential inputs and twisted/shielded cables, not by filtering. Frequency of signal and interference, source impedance/type, and voltage level are what we need to know. \$\endgroup\$ – Henry Crun Jun 28 '18 at 19:32
2
\$\begingroup\$

If it's a 2nd order filter it's roll-off slope is 40 dB per decade so if you have 90 dB of attenuation at 128 kHz there should be 50 dB of attenuation at 12.8 kHz and 10 dB of attenuation at 1.28 kHz.

If you did a little math to predict where the cut-off frequency is you'd find that a 2nd order filter with a cut-off frequency of 720 Hz fits the bill. This is because you know that the cut-off is one-quarter of a frequency decade below 1.28 kHz. So, to get it in real numbers you take the antilog of 0.25 and this equals 1.778. Divide that into 1.28 kHz and the cut-off frequency for the filter will be 720 Hz.

720 Hz is the 1st vital ingredient that you need to be able to design your filter.

If you were making it from an RLC circuit like this: -

enter image description here

You would choose L and C to give this response: -

enter image description here

Interactive tool source.

I've just fiddled with a few values to get Fc = approximately 720 Hz and altered the resistor value to give a damping ratio of about 0.866. This last bit is important because it's highly likely that the type of filter used is a Bessel filter and it has very flat phase delay - see the step response curve below the bode plot - there is an overshoot of about 0.4% and this makes it a good all-round filter for the front-end of an ADC that is used for measurements.

So armed with the cut-off frequency and damping ratio you can choose an active op-amp sallen key architecture like this: -

enter image description here

See this link for a tool where you can enter values of R's and C's and get values for cut-off frequency and damping ratio. If you chose R1 = 14000 ohm, R2 = 7000 ohm, C1 = 27 nF and C2 = 18 nF you'd just about get what you want: -

enter image description here

  • fc = 729.2710256902[Hz]
  • ζ = 0.86602540378444

I did try connecting a 2000uF capacitor across the test leads. To my surprise this actually helps a lot to remove noise.

If you invested in a non-polarized capacitor of 27 uF and an inductor of 1800 uH (as per the bode plot detail on component values) you would get a decent filter that broadly has the requirements as I see them. But, I'd go for a sallen key type.

\$\endgroup\$
  • \$\begingroup\$ Andy, Thanks for your excellent response. You are obviously extremely competent is this area. Unfortunately, I'm at the opposite end of the competency spectrum. I'm considering how to implement your concepts but I've never purchased components like these before. Electronics web sites I've visited have an overwhelming number of choices related to these components. Is guess any 1800uH inductor and 27uF capacitor will do? \$\endgroup\$ – David Jul 2 '18 at 15:20
  • \$\begingroup\$ If you pick an 1800 uH choke/inductor and leave a link I'll take a look at it. Ditto the 27 uF but it might be easier to find a 22 uF and parallel it with a 4.7 uF. \$\endgroup\$ – Andy aka Jul 2 '18 at 17:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.