6
\$\begingroup\$

This is from datasheet of TI part ADS7041:

enter image description here

The 200ohm resistor and capacitor form a low pass anti-aliasing filter. On the left of this, we have the input driver buffer op-amp. How do we know if we need to use a buffer op-amp between the voltage source and the ADC?

I assume that we are putting the buffer so that the ADC does not load the voltage source which can cause distortion. But then, the 200ohm resistor in series with the input impedance of the ADC is going to cause signal distortion anyway. I am quite confused.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Possibly a duplicate of Why is a Buffer+RC-combination recommended for driving a SAR ADC analog input?. \$\endgroup\$
    – davidcary
    Commented Aug 3, 2023 at 16:33
  • \$\begingroup\$ What is the sample rate? 200 Ω into 1.5 nF has a knee at past 100 kHz somewhere. I doubt the aim is to have an anti-aliasing filter. You could have a two-pole filter with an op-amp in the picture (e.g. Sallen-Key). It would be foolish to just have a RC after the op-amp. \$\endgroup\$
    – Kaz
    Commented Aug 4, 2023 at 6:01
  • \$\begingroup\$ @Kaz: The sample rate is marked in the diagram: 1 Msps. For that rate, a lowpass filter with a corner frequency near 100 kHz will provide some anti-aliasing (a higher order filter would be better, of course). \$\endgroup\$
    – Ben Voigt
    Commented Aug 4, 2023 at 14:58

4 Answers 4

10
\$\begingroup\$

If the ADC requires a low impedance to take a sample quickly, then a source with high impedance cannot drive the ADC input.

For example if you monitor a 9V battery voltage with a 5V MCU such as AVR, and in order to not waste battery too much, you divide it with two 1 megaohm resistors down to half for the ADC input. It has a source impedance of 500k.

The AVR ADC input requires impedance less than 10k to work properly. The source impedance is too high to charge the ADC sampling capacitor quickly enough during the sampling phase and the sampling capacitor is not fully charged before the sampling phase is over and conversion of the sampled voltage starts.

Then you need a buffer op-amp, which has very high input impedance so it does not load the weak source, and can provide very low output impedance to ADC input that requires a strong source.

So the 200 ohms is not a problem, it is basically nothing compared to the required 10k, and sometimes it is even required.

Some op-amps are not stable with a capacitive load on output, and they don't work well when the empty sampling capacitor is periodically connected and disconnected directly to op-amp output for sampling.

In these cases the series resistance makes the op-amp stable when empty sampling capacitance is switched to op-amp output resistor.

And sometimes, like pictured, there is a simple RC filter before the ADC, which makes the AC impedance for the ADC input lower than the impedance of the resistor, and it may also function as simple anti-aliasing filter for the ADC input to prevent aliasing when sampled signal contains too high frequencies to sample the waveform properly.

Sometimes you don't need the op-amp, as you can also just connect the resistor divider or other high impedance source directly to the ADC input - and put a large enough capacitor on the ADC input to bring down AC impedance. This will make the voltage to stay relatively unchanged even when ADC takes a single sample of the voltage, but because after taking each sample, it takes long to charge the capacitor back to nominal voltage, it will limit the sampling rate how often the voltage can be sampled before the DC impedance starts to limit the sampling rate.

So if you need to sample a high impedance source at high sampling rate, the op-amp is necesary.

\$\endgroup\$
6
  • \$\begingroup\$ During my search I found that some ADCs have buffers built in, some have anti-aliasing filter built in and it can be controlled by writing to specific registers in the ADC, some even have amplifers built in for which the gain can be set. There are also dual ADCs (wonder who needs them) The variety available is truly mind blowing and mind boggling. I am in state of shock. Do you know of any resource that summarizes what exists out there and how to know which one to use? \$\endgroup\$
    – quantum231
    Commented Aug 3, 2023 at 0:07
  • 2
    \$\begingroup\$ I wish EEVBlog had also done a video on top 5 jellybean ADC and DAC. \$\endgroup\$
    – quantum231
    Commented Aug 3, 2023 at 0:07
  • 2
    \$\begingroup\$ @quantum231 There are DACs and ADCs specialized for all kinds of purposes, such as data acquisition, audio, video, and even different subcategories of video. You use a specific ADC for specific purpose, or a generic ADC for generic purpose. Dual ADC means it can sample 2 channels at a given time, as normal ADC can sample 1 channel at a time, even if it had 16 multiplexed input channels. \$\endgroup\$
    – Justme
    Commented Aug 3, 2023 at 5:28
  • 1
    \$\begingroup\$ @quantum231 Dual ADCs are simply for anyone who needs two inputs measuring at the same time. :) If they don't need to be sampled at exactly the same time, you can also get ADCs which have one ADC with an analogue multiplexer selecting one of multiple inputs, and you can read each input consecutively. For your other question, basically you decide what to use by what your design needs. How accurate, how fast, and how you need your micro to talk to it, are the three main things to think about. \$\endgroup\$
    – Graham
    Commented Aug 4, 2023 at 12:57
  • 2
    \$\begingroup\$ @gyuunyuu Audio can normally tolerate higher non-linearity, but also normally needs to run pretty fast. So we tend to separate audio out as a separate use case from instrumentation-type applications. \$\endgroup\$
    – Graham
    Commented Aug 4, 2023 at 12:59
9
\$\begingroup\$

Forgive me if I'm repeating things you already know. The ADC has an internal capacitor that charges to the sampling input voltage, disconnects from the input, connects to the measurement circuit, and reads the voltage. If your input is high impedance, a buffer amp can charge the sampling capacitor faster and essentially eliminate the sag on the sampling line as it charges.

The 200 ohm resistor is too low to affect things here. Microchip suggests using an internal impedance of 2k ohms and a sample/hold capacitance of 15pF in the absence of additional information from the manufacturer. If your signal source alone is unable to charge the S/H capacitor to within 1/2 of the LSB value of the desired final voltage in the sampling time, then a buffer amp would be indicated.

\$\endgroup\$
2
  • \$\begingroup\$ You wrote "If your signal source alone is unable to charge the S/H capacitor to within 1/2 of the LSB value of the desired final voltage in the sampling time", how does a person prove if this is or is not the case? \$\endgroup\$
    – gyuunyuu
    Commented Aug 3, 2023 at 11:18
  • 1
    \$\begingroup\$ @gyuunyuu By simulating or doing the basic math of capacitor charging from a voltage source with known impedance. The linked Microchip appnote gives the formulas. \$\endgroup\$
    – jpa
    Commented Aug 3, 2023 at 18:01
4
\$\begingroup\$

You can see from the datasheet the (typical) degradation of performance with increasing series resistance:

enter image description here

If your sample rate is low and your source impedance is relatively low you may be able to get away without an amplifier, as described in the datasheet.

If your sample rate is very high and you need the best performance a very low value of resistor and a good amplifier with adequate stability under the challenging conditions of capacitive loading would more likely be required unless your source has extremely low source impedance.

For example, at 10% of maximum sample rate (100kSPS) 10kΩ does not degrade performance too much.

\$\endgroup\$
3
  • \$\begingroup\$ I see, all along I had thought that we just pick up an ADC and stick it into the circuit. Now I can see that there are all these other details. So besides anti-aliasing filter and a possible buffer between the voltage source and the ADC, is there something else that we might need? Also, why not integrate the anti-aliasing filter into the op-amp to create an active filter? \$\endgroup\$
    – gyuunyuu
    Commented Aug 3, 2023 at 11:17
  • 1
    \$\begingroup\$ @gyuunyuu That is not infrequently done, especially at lower bandwidths, however note that requires an even higher performance op-amp. You really have to run the numbers. \$\endgroup\$ Commented Aug 3, 2023 at 11:41
  • \$\begingroup\$ If we make a sallen-key low pass filter using that op-amp, what type of challenges will that create? Close loop gain being too small in the op-amp or something else? \$\endgroup\$
    – quantum231
    Commented Aug 4, 2023 at 21:08
1
\$\begingroup\$

Many microcontrollers' ADCs, when they take a sample, will behave as though they take a small cap, charge it to some potential, connect it to the input, and then some time later sample the actual voltage on the pin. While I wouldn't think it would cost much to force the cap to be charged to either the positive or negative reference input under programmer control, different controllers seem to do things differently without offering programmers any choice in the matter.

If the input has minimal capacitance and series impedance, momentarily connecting the precharged capacitance will disturb the input voltage, but the disturbance will settle out before the actual measurement is taken. If the input has large external capacitance connected with minimal impedance between it and the chip, the disturbance will be limited by the ratio of the internal cap and the external cap. If the cap is small enough to be disturbed, but large enough to create a significant time constant, that may significantly degrade measurements.

Here, the cap is large enough not to be significantly disturbed by a single measurement, but the RC time constant is long enough that the disturbance from one measurement may not fully settle before the next measurement is taken. The bigger the series resistance, the more each measurement will be affected by unsettled effects from earlier measurements.

\$\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.