Skip to content

Analog-to-Digital Converters

ADC Input Impedance: Why the Source Sees a Dynamic Load

The input is a switch and a capacitor, and every sample takes a bite out of your signal.

  • Lesson 5 of 5
  • Advanced

5 min read

A datasheet says the input impedance is 500 kΩ. You conclude, reasonably, that your amplifier is barely loaded and move on.

There is no 500 kΩ resistor anywhere inside that part. There is a switch and a capacitor, and the number you read is an average over a sampling period. It changes when you change the sample rate, and the instantaneous behaviour, which is what your amplifier has to survive, looks nothing like a resistor.

Quick answer. A SAR converter's input is a capacitor connected through a switch. Each time the switch closes, that capacitor takes whatever charge it needs from your circuit, in a gulp. The average current is the charge times the sampling rate, and dividing the input voltage by that average current is where the equivalent resistance in the datasheet comes from. It is an average, not a component.

The charge, not the current

The amplifier drives an isolation resistor and a filter capacitor, with its feedback closed before the resistor. Inside the converter a switch connects the sampling capacitor, which takes its charge from the filter capacitor in an instant.
Figure 1. The amplifier only replaces the average; the capacitor supplies the gulp.

Suppose the sampling capacitor is 20 pF and the input sits at 2.38 V, the full scale of the front end this course carries. At the moment the switch closes, the charge that has to move is:

Q = C_sh · V_in = 20 pF × 2.38 V = 47.6 pC

Forty-seven picocoulombs is not very much. What makes it awkward is the time it has to move in, which is the acquisition window, a few hundred nanoseconds. A gulp of that size in that time is a current spike of tens of milliamps for a few nanoseconds, and that is what your amplifier is asked to deliver.

This is also the mechanism behind the phrase charge kickback: the same switching event pushes a spike back out into whatever is connected, which is why an unbuffered sensor connected straight to an ADC input can read differently depending on the sample rate.

Where the equivalent resistance comes from

Average the gulps and you get a current:

I_avg = C_sh · V_in · f_s

At 100 kSPS:

I_avg = 47.6 pC × 100 000 = 4.76 µA

And the equivalent resistance the datasheet quotes is simply the input voltage divided by that:

R_eq = 1 / (f_s · C_sh) = 1 / (100 000 × 20 pF) = 500 kΩ

Now change nothing but the sample rate:

Sample rateAverage currentEquivalent resistance
10 kSPS0.48 µA5 MΩ
100 kSPS4.76 µA500 kΩ
1 MSPS47.6 µA50 kΩ

The part did not change. The number did, by two orders of magnitude, because it was never a property of the part on its own.

The error this actually causes

The average current is small, but it has to come through whatever sits between your amplifier and the converter. If that is a series resistor of value R, the current develops a voltage across it:

error = I_avg · R

That voltage is a DC error, present on every reading, and it scales with the input voltage because the charge does. In other words it is a gain error, and it is invisible to an offset calibration.

At 100 kSPS with a 1 kΩ series resistor:

error = 4.76 µA × 1 kΩ = 4.76 mV = 5.9 codes

Almost six codes of gain error, caused by a resistor somebody added as a filter. This is the single most common way a correctly designed signal chain loses accuracy at the last centimetre, and it is the constraint that dominates the next lesson.

Why a bigger capacitor is the obvious answer and not a complete one

The usual fix is a local capacitor at the ADC pin that supplies the gulp, so the amplifier only has to supply the average. Sizing it from the droop:

ΔV = Q / C_filt
C_filtDroop per sampleIn codes
1 nF47.6 mV59
10 nF4.76 mV5.9
100 nF0.48 mV0.59

Reading that table, 100 nF looks obviously right. It is not, and the reason is the subject of lesson 6: the droop does not have to be small, it has to recover before the next sample, and a large capacitor behind a series resistor recovers slowly. The two requirements pull in opposite directions and the design is the compromise between them.

Things this explains

  • Readings that change with sample rate. The average current scales with

it, so any series resistance turns the rate into a gain term.

  • A multiplexed channel reading high after a high channel. The array keeps

a residue of the previous sample if the acquisition window was short.

  • A high-impedance sensor that works on a meter and not on an ADC. A meter

draws nanoamps continuously; an ADC input draws a burst. Source impedance that is fine for one is not for the other.

  • A design that worked at 10 kSPS and failed at 500 kSPS with no other

change.

What comes next

Lesson 6 turns all of this into a design: three constraints, a worked choice of resistor and capacitor for a real converter, and which one of the three actually binds.

Sources and further reading

Where to go next

¿Te ha resultado útil?

Reddit LinkedIn

The XKEMA newsletter

Hardware engineering worth keeping

New courses, engineering notes and hardware guides, whenever there is something genuinely worth reading. Nothing else.

We will email you to confirm. You can unsubscribe from any issue.

XKEMA, in one sentence

Engineering knowledge management software for companies that develop electronic products: the product structure, its bills of materials, its documentation and its revisions, all in one place.