DACs and ADCs, or there and back again

A look at how digital-to-analog and analog-to-digital converters work - from resistor ladders to delta-sigma modulation.

In one of the earlier articles, I remarked that microcontrollers are eating the world. Even for rudimentary tasks, such as blinking a LED, a microcontroller is now cheaper and simpler to use than an oscillator circuit built with discrete components — or implemented using the once-ubiquitous 555 timer chip.

But in this increasingly software-defined world, zeroes and ones don’t always cut it. Image sensors record light intensity as a range of analog values; a speaker playing back music must move its diaphragm to positions other than “fully in” and “fully out”. In the end, almost every non-trivial digital circuit needs specialized digital-to-analog and analog-to-digital converters to interface with the physical world. These devices are often baked onto the die of the microcontroller, but they are still worth learning about.

Simple digital-to-analog converters (DACs)

The conversion of digital signals to analog usually boils down to taking a binary number of a certain bit length and then mapping subsequent integer values to a range of quantized output voltages. For example, for a 4-bit DAC, there are 16 possible output voltages, so its model behavior could be:

0000 (0)  = 0 V
0001 (1)  = 1/15 Vdd
0010 (2)  = 2/15 Vdd
0011 (3)  = 3/15 Vdd
...
1111 (15) = Vdd

About the simplest practical way to implement such a conversion is a resistor-based binary-weighted DAC:

A simple binary-weighted DAC.

It should be evident that if the binary input is 0000, the analog output is 0 V; conversely, if the input is 1111, the output must be Vdd. For inputs in between, we should get a resistor-weighted average, with each bit having half the influence of its more significant predecessor. This intuitively aligns with how binary numbers work.

We can analyze the behavior a bit more rigorously, too. Let’s see what happens it the input value is 0001. In this scenario, the top three resistors (bits #1, #2, and #3) are connected to the ground in parallel, so we can consider them equivalent to a single resistance:

\(R_{eq} = { 1 \over { 1 \over 1k \Omega } + {1 \over 2k \Omega } + {1 \over 4k \Omega } } = { 4 k \Omega \over 7 } \approx 571 \ \Omega\)

The bottom resistor for the least significant bit (LSB) connected to Vdd. In effect, the circuit can be thought of as a pair of series resistors forming a voltage divider between Vdd and ground. The output of this divider is:

\(V_{out} = { 571 \ \Omega \over 8 \ k\Omega + 571 \ \Omega } \cdot V_{dd} \approx 0.0666 \ V_{dd} \approx {1 \over 15} V_{dd}\)

Analogously, if the input is 1110 (decimal 14), we get Vout ≈ 14/15 Vdd. That tracks with the behavior we were hoping to see.

The most significant issue with this DAC architecture is that the required resistor values quickly get impractical. To avoid high idle currents, the resistance attached to the most significant bit (MSB) can’t be too low; 1 kΩ is a sensible starting point. But then, for a basic 16-bit DAC, this puts the LSB resistor at 1 kΩ * 2¹⁵ ≈ 32 MΩ; for 24-bit resolution, we’d need tens of gigaohms. Precise resistances of this magnitude are difficult to manufacture on the die of an integrated circuit — doubly so if they need to have the same temperature coefficients.

A clever workaround for this issue is the R-2R DAC architecture:

A typical R-2R DAC layout.

This circuit is less intuitive than its predecessor, but it works in a similar way. To decipher the design, let’s start at the bottom node — the two horizontally-placed resistors for bit #0. They supply equal currents to the upper stages, and so we can treat them as functionally equivalent to a single 1 kΩ that’s connected to a new, synthetic input voltage. This new voltage sits at 0 V if LSB = 0 and at Vdd/2 if LSB = 1; in other words, it’s equal to 50% of the input value for bit #0.

If we make this substitution, we end up with the circuit shown below on the left. Further, because the circuit now features two 1 kΩ resistors in series with the bottom input, it’s equivalent to the 2 kΩ single-resistor variant on the right:

Analysis of the R-2R DAC, part 1.

Let’s have a look at bit #1 on this new schematic. The node now features a 2 kΩ resistor connected to the corresponding binary input, plus a 2 kΩ resistor feeding from the previous stage. No matter what’s going on above, we can substitute these two with a single 1 kΩ hooked up to a yet another synthetic input signal:

Analysis of the R-2R DAC, part 2.

This process can continue; it should be clear that after the final iteration, we’re bound to end up with an output voltage that’s 50% of bit #3, 25% of bit #2, 12.5% of bit #1, and 6.25% of bit #0.

(This doesn’t sum up to 100% because some of the upper range is lost to the pull-down resistor at the bottom of the ladder.)

Oversampling DACs

Although the designs discussed above are simple and sweet, they also pose challenges with linearity at higher resolutions — especially past 10-12 bits. Resistors can be easily made with accuracy as good as 0.1%, but in a 16-bit DAC, the influence of the LSB is supposed to be just 0.003% of MSB. If the MSB resistor deviates by 0.1% from the intended value, this is more than enough to throw the whole scheme badly out of whack.

This problem led to the development of the so-called oversampling averaging DACs. Such devices output lower-resolution but alternating signals at a high frequency. Then, a lowpass filter on the output averages the values to produce a greater range of slower-changing intermediate voltages.

To illustrate, averaging four subsequent one-bit DAC outputs can add three intermediate voltages in between what the DAC can natively produce — an effective gain of two bits:

average(0, 0, 0, 0) = 0
average(0, 0, 0, 1) = 0.25
average(0, 1, 0, 1) = 0.5
average(0, 1, 1, 1) = 0.75
average(1, 1, 1, 1) = 1

There is a price to pay, of course; for one, some of the high-frequency noise inevitably gets past the filter. That said, as a whole, the approach is remarkably robust. Indeed, many DACs used in consumer audio employ single-bit pulse trains at hundreds of kilohertz to produce claimed output resolutions of up to 24 bits. In practice, the noise floor makes that number fairly meaningless; on the flip side, the linearity of 1-bit DACs is superb, because accurate timing is much easier than manufacturing ultra-precise resistors.

Classical analog-to-digital converters (ADCs)

Compared to going the other way round, converting analog voltages to binary numbers is a fairly involved affair. About the only practical way to obtain accurate and instant readings is to use one voltage comparator (an open-loop op-amp) per every quantization level desired, for example:

The architecture of a flash ADC.

Such “flash” ADCs are sometimes used in specialized applications where speed is paramount, but the size of the circuit grows exponentially with the number of bits — and with it, the chip’s power consumption, input capacitance, and so forth. For these reasons, they’re not made for resolutions higher than 4-8 bits.

A more common architecture is to use a single comparator combined with a reference voltage that changes over time in some predictable way; a rudimentary example may be a capacitor that is being charged through a resistor. The time elapsed from the beginning of the charging process to the comparator triggering can be used to infer the unknown input voltage.

In practice, because the charging curve of a capacitor is nonlinear, the reference signal is more commonly provided by an integrator circuit:

An op-amp integrator.

An integrator is a basic op-amp circuit with an interesting tweak: a capacitor goes where a feedback resistor would normally be. If the voltage on the inverting (Vin-) input becomes higher than on the non-inverting one (Vin+), the output immediately swings negative, allowing the capacitor to charge via R. The process is current-limited in proportion to the supply voltage because excessive current would cause Vin- to drop below Vin+. Constant-current charging means that until saturation, the voltage across the capacitor increases linearly, and the output voltage keeps sliding down to match. In other words, if supplied with a square waveform centered around Vref, the integrator will output a nearly-perfect triangle wave — a superbly linear reference for an ADC!

The time elapsed between the beginning of the charging process and the triggering of the comparator depends not only on the input voltage under test, but also on the slope of the triangle wave; that slope in turn depends on R and C. It follows that to improve accuracy, an ADC should instead measure the duty cycle of the comparator output signal across several cycles of the triangle wave. A 25% duty cycle means that the compared voltage sits at 75% of Vdd, no matter the value of R and C.

Slope-based ADCs are very accurate and exhibit low noise, but they tend to be painfully slow. One way to improve their performance is a bit of digital trickery, known under the trade term of a successive approximation register (SAR). In essence, the ADC uses an internal DAC to generate reference voltages, and then implements the equivalent of a binary search algorithm that should be familiar to all computer science buffs. It starts by comparing the unknown input voltage to Vdd/2. If the comparator indicates the input is higher, the ADC can rule out the entire lower half of the range, and performs the next test at the midpoint of what’s left (3/4 Vdd). The successive halvings of the search space get to the exact value in just a couple of steps. The price to pay is some loss of precision due to DAC linearity errors, along with a bit of digital switching noise.

An advanced (“pipelined”) ADC may combine a couple of these techniques in a single package; for example, it might use a “partial” flash ADC architecture to instantly resolve a couple of bits, and then execute a sequence steps to scale the sampled input voltage and resolve a couple more.

Delta-sigma ADCs

So far, so good — but the most interesting trick in the ADC playbook is high-frequency interpolation, commonly using what’s known as delta-sigma modulation. And it’s done in a pretty wacky way.

In its most basic variant, a 1-bit delta-sigma ADC rapidly outputs a train of logic “0”s or “1”s via a comparator stage. The initial output of the ADC — whatever it might be — is then used as a part of an unusual feedback loop that computes the difference between the output signal and the input. In most circumstances, the analog input doesn’t match neither of the two possible digital output voltages, so the differential amplifier at the front of the delta-sigma ADC outputs a large momentary positive or negative error value:

A simplified 1-bit delta-sigma ADC architecture.

These momentary errors are then fed into an integrator. This component, discussed in detail earlier on in the article, essentially sums the errors over time. It converges on the midpoint output voltage if all the positive and negative errors cancel out, or ramps away from the midpoint if they don’t.

If the integrator-measured average error becomes positive (i.e., the ADC is outputting too many zeroes), this prompts the comparator at the output stage to start producing “1”s. Conversely, if the net error is negative (too many ones), the output stage starts producing “0”s.

Although this may seem like a fairly insane approach to making precise signal measurements, the 1-bit stream produced by the comparator can be then digitally processed to compute the duty cycle of the high-frequency pulses — and thus infer the analog input value with superb linearity, because there is no need to produce any sort of an analog reference for measurements. Of course, to achieve reasonable precision, the clock used to operate the ADC must be much faster than the desired analog sample rate.

The same “delta-sigma” moniker is also used to refer to a subclass of oversampling interpolating DACs that employ similar pulse modulation, as discussed earlier. That said, they’re not nearly as clever: the bulk of the pulse modulation happens in the digital realm, with no cute analog feedback loops to speak of.

If you liked this article, please subscribe! Unlike most other social media, Substack is not a walled garden and not an addictive doomscrolling experience. It’s just a way to stay in touch with the writers you like.

For more articles on electronics, click here.