1
\$\begingroup\$

I would like to filter signal on ADC input. My question is if I can add before input ordinary RC filter. I suppose that this is not good idea. I have to limit number of operational amplifiers on the board so it can't be isolated by buffer amplifier.

If resistance is too big the internal (S&H) capacitor will not fully recharge. Capacitor from RC filter will be parallel with capacitor from S&H and it will change the capacitance of sampling capacitor? Am I right or not ?

cheers

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Usually, ADC manufacturers suggest a typical interface connection in the datasheets, or related application notes. \$\endgroup\$
    – Nazar
    Commented Nov 16, 2016 at 15:15
  • \$\begingroup\$ You are probably not right. There is a lot of supposition here. What a2d are you talking about, what clock frequency, and what single frequencies do you want to pass? \$\endgroup\$
    – owg60
    Commented Nov 16, 2016 at 15:22
  • 1
    \$\begingroup\$ If you have enough frequency 'space' between signal and sampling rate, an R/C can surely be used. For the R, look into the datasheet to find the max. impedance the ADC can accept on its input and/or make sure that the C is much bigger than the S&H capacitor (which may be some pF only). \$\endgroup\$
    – JimmyB
    Commented Nov 16, 2016 at 16:18

4 Answers 4

2
\$\begingroup\$

When not sampling, the filter capacitor will charge to the filtered vin value. Now when you sample, charge will move from the filter cap to the hold cap. The resistor is there but doesn't have much to do with it. If the filter cap is large compared to the hold capacitor, then the voltage will change very little and the hold capacitor will charge very quickly.

\$\endgroup\$
1
\$\begingroup\$

Generally a C to ground at the input of a sampling converter is a good idea, as it provides a reservoir for the charge needed by the converter. Check the data sheet for what value capacitors are OK. Once the C is providing the charge, a series R is usually OK. Again check the converter data sheet for recommendations on the value of the R.

\$\endgroup\$
2
  • \$\begingroup\$ I was trying to find this data in datasheet but I found only information how to calculate protection resistor. Maybe I don't understand this datasheet, could you tell me which data indicates to this? MAX11046 [link]datasheets.maximintegrated.com/en/ds/MAX11044-MAX11056.pdf \$\endgroup\$
    – e2p
    Commented Nov 17, 2016 at 15:32
  • 1
    \$\begingroup\$ @e2p That is a tame converter, the inputs are easy to drive. The analogue inputs go to T/H amplifiers, not directly to naked ADCs. My comments were addressed to the latter. The specified input capacitance 15pF is very low, and the 1uA input leakage current is the only limitation on the size of the R. You have no worries. \$\endgroup\$
    – Neil_UK
    Commented Nov 17, 2016 at 16:33
1
\$\begingroup\$

A simple RC filter will get the job done but you'll be sacrificing a lot of bandwidth. The maximum frequency you can sample is 1/2 sampling frequency so with 48kHz sampling frequency you'd be limited to 24kHz or less. To achieve that you actually need to set filtering corner frequency at a lower frequency to have adequate attenuation at the 1/2 sample frequency.

You only get -20dB/decade of attenuation with RC filter so to get -20dB attenuation at that 24kHz frequency, your corner frequency should be 2.4kHz so you're filtering out quite a lot.

You'd really do better with 2nd or 3rd order opamp filter which requires you to use one of your opamps. See some examples here: http://www.edn.com/design/analog/4363970/Design-second-and-third-order-Sallen-Key-filters-with-one-op-amp

Here's a neat tool for designing the 3rd order sallen key filter. Just give it the cherbychew corner frequency and off you go. You probably want to define C1=C2 based on some convenient component value. http://sim.okawa-denshi.jp/en/Sallen3tool.php

With regards to the capacitor, what will actually happen (with some ADC types e.g. sample and hold input stage) is that the charge from the external capacitance transfers to the internal capacitance. So you in fact want to have enough capacitance to keep this glitch < 1/2 ADC unit.

\$\endgroup\$
9
  • \$\begingroup\$ kS/s not kbit/s but +1 for reminding about aliasing filters. \$\endgroup\$
    – Andy aka
    Commented Nov 16, 2016 at 14:55
  • \$\begingroup\$ Thanks Barleyman, unfortunatelly as I wrote I coulnd't use more opamps and build active filter. Sacrificing bandwidth is not a problem (sampling frequency is 250 kS/s and signal is maximum 15kHz ) \$\endgroup\$
    – e2p
    Commented Nov 16, 2016 at 15:02
  • \$\begingroup\$ @user128885 I usually have some extras since they come in pairs.. There's even a song about how to connect your unused opamps. \$\endgroup\$
    – Barleyman
    Commented Nov 16, 2016 at 15:36
  • 1
    \$\begingroup\$ @user128885 If your sample frequency is 250kHz, you need to filter >125kHz, meaning you should have your corner frequency @ 12.5kHz which is going to eat at your signal. -3dB at corner frequency means halving the signal! \$\endgroup\$
    – Barleyman
    Commented Nov 16, 2016 at 20:08
  • 2
    \$\begingroup\$ @e2p As pointed out a couple of times, there's a charge transfer going on so you want adequate capacitor size. Maxim doesn't actually tell you what's the size of the SAH sample capacitor but we can presume it's smaller than the 15pF input capacitance. Going by healthy three orders of magnitude, with 15nF you should see (almost) no sampling glitch. If you have a decent oscilloscope, use small ~22pF RC filter cap and you should see the sampling glitch. \$\endgroup\$
    – Barleyman
    Commented Nov 17, 2016 at 18:03
1
\$\begingroup\$

In this case, we need to talk about impedance.

The ADC will have specified a certain maximum impedance for its input signal. That number already factors in the S&H capacitor, so in the simple case you have just an R/C of the signal impedance/line resistance and the S&H C. You are allowed to increase the signal's impedance up to the maximum, so you can just add a resistor in line to get close to that value.

Notice that a C to ground decreases the impedance, so it only makes sense if your signal impedance (after the R) is bigger than allowed. If it is, then you need to select (for a given R) a C to get to within the allowed impedance range for the frequency of the signal.

Hence, the numbers for R and C to use can be figured out from the required impedance of the ADC and the signal's frequency alone.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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