AnalogTalk

We are regaled with the wisdom of using smart power in tech pubs. It’s one of the buzz words of our day, with no one wanting to robe themselves in anything other than the green trappings of environmentally better, efficient smart power. It seems geekdom has decree pronouncing anything other than praises is sheer sacrilege. Having always believed that technology must be approached with an irreverence that breaks down barriers and builds bridges to new vistas, … here I go.

Without a doubt placing the computational power to provide power factor correction and pulse modulated passing of just the right amount of energy to the next stage in an AC/DC power supply for a server, computer or another power spiking and hunger device is indeed smart. And further downstream the same argument may be made for a point of load DC/DC stage, less the power factor correction. But as the current demand lowers so does the budget that can be allocated to a device that MIPS its way to responsive PWMing.

Lower voltage and current become the estate of state machines and other more energy efficient controllers and switching regulators. But these devices can wreak havoc in stages that deal with the processing of small signals like various sensors or wireless reception may require. Here a simple regulator may be needed to supply power and isolate from upstream noise.

I think it was about 1979 when I first tripped across the circuit for a regulator. Having used a resistor and a zener diode to regulate and also trying a regulator to regulate, I became duly impressed how much more sophisticated, dare I say ‘smart’ a regulator was. But to see the current mirrors, negative feedback to control thermal run way and the passing on of ripple, and generally the effective use of transistors to control transistors, suddenly I understood why 60 cycle seems so depressed at the output and the supply voltage so stiff. Yes, I totally geeked about how cool it was and even found fellow geeks to share the elation.

But isolation is not the only benefit a low drop out regulator or LDO can bring to the design. Having ascended the power transfer pyramid from the broad base of high power to less and less we find ourselves in an altitude of very thin current. Here too smart power finds a home if the application in question can have discontinuous operation. XLP PICmicro is famous for sleeping most of the time and thus reducing energy demand.  But here too live the creatures of the dumb power domain.

In many battery applications the drop out from source to supply is so small greater efficiency can be achieved with an LDO. So while great strides in increased power efficiency can be made with smart power in many applications, there are still many applications where using so called smart power is not so smart.

With the push for smaller yet simpler electronic devices in consumer and industrial markets, primary cell batteries (typically AA, AAA or sometimes AAAA), continue to be the go-to power source in many applications. Cheaper than Li-ion rechargeable batteries, perceived safer by many than most Li-ion type chemistry and cheap to implement by manufacturers and cheap to replace by consumer, primary batteries are an ideal power source for user friendly and low cost applications. Continued decrease in electronic devices’ power consumption makes primary batteries good candidates even for long lasting applications.

A single AA alkaline battery can have a capacity of 2000-2800 mAh (1.5V to 0.8V discharge), while a low power MCU (PIC16LF1824 as an example) can have supply current range from 210-1800 uA in active mode, down to 40 nA in sleep mode. That means that theoretically, a single battery can power this MCU anywhere from 1000 hours to 10,000+ hours without even engaging sleep mode. With intermittent use of sleep modes, that time can be extended by an order of magnitude. Of course, the challenge here is that while the alkaline AA battery has voltage range 0.8V to 1.5V, the PIC16LF1824 has a typical input range from 1.8V to 3.6V. Common solution is to use two AA batteries for 1.6V to 3V input, but that increases the size of the application and the operational costs to the consumer. In applications where size is important, it is possible to limit usage to just one primary AA battery by utilizing a boost regulator like the MCP1640 to provide the necessary operating voltage. A solution using a boost regulator can occupy less than 90 mm² on the application board, considerably less than the 8.1 cm³ volume needed for another AA battery.

Implementing a boost regulator solution in an application using AA battery is very simple, but also allows for a higher degree of efficiency optimization than does a typical dual AA battery setup. Below are a few tips and tricks on how to get the most out of the boosted AA battery application.

Lower MCU Operating Voltage

A simple trick to achieve lower overall power consumption when using a single cell primary battery with a boost regulator is to use the lowest possible voltage at which the MCU (or the entire application) is capable of working at.  As the current consumption of the MCU is determined by the input voltage (thank you, Mr. Ohm), that means that a factor of X drop in voltage will mean approximately X² decrease in power consumption. For example, dropping input voltage of PIC16LF1824 from 3V to 1.8V (40% decrease) also leads to a decrease in operating current from 210 uA to 118 uA  (43% decrease*). As the result, at 1.8V, the overall power consumption is about 1/3 of the power consumption at 3V.

This approach allows going from a two cell power source to a single cell plus a boost regulator, yet keeping the same run time for the application and is possible specifically because the original input source is lower than the operating voltage.  This would be impossible when using two AA batteries, where the input voltage is predetermined by their state, typically 0.8V to 1.5V each, and cannot be changed without some sort of voltage regulator.

*Example for 4 MHz speed, Medium power mode

Dual Operating Voltage

In applications where 3V or higher is necessary for active periods of time but not for standby, such as remote controls or security sensors, using a single cell with boost regulator allows to provide higher voltage when necessary and lower voltage the rest of the time. That can be accomplished through an external switch changing between two resistor dividers for the necessary voltages

Increasing Value of Feedback Resistors

When boost regulators with external resistor dividers for setting voltages are used, such as the MCP1640, using higher value resistors (on the MΩ scale) helps decrease leakage currents. The few µA that this approach saves, can make quite a big difference for ultra low power applications, considering that some MCUs can use as little as 40 nA in deep sleep mode. However, one should be careful with using this approach in high humidity conditions, as that can affect overall stability and power consumption of the boost regulator.

Selecting Quality Passive Components

Choosing the right components around the boost regulator can greatly increase (or decrease) overall efficiency and be the difference between a barely acceptable replacement for a two-cell application and a great one. For example R_DC (DC series resistance) of the inductors can affect the efficiency by as much as 5%. DC equivalent resistance (ESR) difference between tantalum and ceramic or aluminum electrolytic capacitors can chew up as much as 10% of efficiency while also reducing maximum output current capability.

A study based on searching through Facebook has found a greater number of company founders and CEOs to have an advanced engineering background compared to MBAs. Additionally, the average age has dropped since the last study. Of course this is a single source for sampling, but good news for budding engineers.

Culling through 36 million U.S. and Canadian professional profiles on Facebook, the company found that 3,337 company founders and chief executives have an advanced engineering background, compared with 1,016 MBAs.

Article

Nickel based batteries have a long history in portable products since 70s. Fully discharging the NiCd (Nickel Cadmium) batteries before recharging it is always stated as a requirement in the user’s manual.  Although you can leave the device operating continuously until the battery is run down there are also charging stations that have discharge function to drain the batteries before charging. This exercise will prevent the crystalline formation in a cell which reduces its capacity. When NiMH (Nickel-Metal Hydride) batteries became available in the early 90s for portable products, consumers were still taught to do the same.

Modern Li-Ion (Lithium-Ion) batteries do not require this exercise to prolong its life due to the nature of its design. However, most consumers, unaware, still fully drain their mobile devices before recharge without knowing it affect on the life of a Li-Ion battery. When a battery’s capacity decreases to 80% of its original capacity, it is considered to be at its end of its life. A battery has two types of life – calendar life and cycle life. A typical calendar life of a Li-Ion battery is about two years. It will still work, but can’t restore much energy after each charge and the impedance will increase as well. The cycle life for a Li-Ion battery depends on its grade and manufacturer. It includes charge and discharge from 500 cycles to 1000 cycles. Each fully charge or discharge will consume a cycle life from a battery. Thus, shadow charge, or “topping off” the Li-Ion battery before it is significantly drained  can help to reduce the cycles.

Today’s system design also allows a device to operate from the power source to reduce the usage of a battery. Thus, plug in your mobile device when you can, so you can have a better user experience.

The smoke detector in your home goes off.  It is a loud and annoying pulsating tone that is specified by UL 217.  The loudness is specified in terms of the A-weighted sound pressure level which approximates the frequency response of the human ear.  The minimum sound pressure level as a continuous tone is specified as 85 db(A) at 10 ft.  The pulse rate of the tone is also specified and defined as a three pulse temporal pattern.  The frequency of the pulsating tone is not specified but the A-weighted sound pressure level provides some practical guidance.  The frequency of the pulsating tone should be in the range of 1000 Hz to 5000 Hz.  A horn frequency of about 3100 Hz is usually chosen.   Frequencies outside this range would require more energy to generate the required sound pressure.

UL217 does not specify how the sound is to be generated.  In battery powered and battery backed up smoke detectors power consumption is a concern and the 85 db(A) specification must be met at the low battery voltage. The device usually chosen for residential smoke detectors is the piezoelectric horn.   Another consideration is size and weight.  The piezoelectric horn consists of a piezoelectric diaphragm mounted over a resonant cavity and the associated electronics. The piezoelectric diaphragm and resonant cavity are small and light compared to a conventional speaker.  A two wire piezoelectric horn can be operated as a conventional speaker driven by a square wave.  Maximum sound pressure is achieved when the driving frequency matches the resonant frequency of the piezoelectric element.  Matching an oscillator frequency to the resonant frequency in manufacturing is difficult.  A three wire piezoelectric horn provides a feedback signal which can be used to create a self resonating horn.  The third wire provides a feedback signal to a push pull circuit which allows the piezoelectric element to operate at its resonant frequency.  This circuit configuration is generally referred to as a self driven horn.

The sound pressure is also a function of the applied voltage and 7 volts is usually the minimum voltage that can achieve the necessary sound pressure.  An ionization smoke detector needs a 9 volt battery to power both the ionization chamber and the piezoelectric horn.  Photoelectric smoke detectors only need 9 volts to power the piezoelectric horn.  In these 9 volt applications the RE46C100 and RE46C101 provide the necessary circuitry to drive the piezoelectric horn.  The RE46C101 provides an LED driver since smoke detector applications are also required to provide a visual indication of operation.

Photoelectric smoke detectors can be built using low voltage components but driving the piezoelectric horn with sufficient sound pressure is the challenge.  The RE46C117 horn driver contains a boost regulator that can power the piezoelectric horn on demand with an input voltage down to 2 volts.  The boost regulator provides a nominal 4 volt output when the horn is not on and a nominal 10 V output when the horn is on.  The RE46C107 is similar but adds a linear regulator with a 3.0 V or 3.3 V output and a LED driver.  This horn driver provides power for the piezoelectric horn and other circuitry such as a microcontroller.  The RE46C107 minimizes the effects of a discharged battery which is important for reliable smoke detector operation.

Horn drivers have evolved from providing the basic capability of driving piezoelectric horns that meet UL 217 or EN 14604 to a component that provides additional necessary functionality in low voltage detector applications.   This additional capability makes the horn driver a useful companion IC to a variety of circuit approaches used in detector applications.

For those accustomed to measuring signals in the lower realms of signal bandwidths, moving to signals much above DC can involve hardware specifications other than the familiar INL, DNL and output noise.  When discussing AC versus DC specifications, scenes of old rock n’ roll come to mind blurring any further understanding of potential new knowledge.  However with a little sampling of antiphonal baroque through stereo headsets, the mind clears and AC specifications of analog-to-digital converters can settle in with the DC.

So what is this new alphabet soup of AC specifications (also called dynamic specifications), and what is the purpose of using them?  Specifically will be discussed SNR, SFDR, THD, SINAD, ENOB and IMD.  For measuring signals with a higher bandwidth, it is important to find a way to characterize the frequency components of the measured signal and the analog-to-digital converters measuring them.  Most DC specifications will not capture these characteristics.  Let’s look at these specifications one by one.

In order to look at noise over the frequency input of the A/D converter, the specification Signal to Noise Ratio (SNR) is very useful.  In order to make this measurement a signal is put through the A/D converter and is called the fundamental.  SNR measures the difference between the noise power and the level of the fundamental.  It shows the usable range of the A/D converter without interference to the input signal from noise inherent to the A/D converter itself.  Note in the graph below that the noise floor is lower than the noise power.

In addition to noise, it is also of interest to look at distortion components that might be inherent in the A/D converter which were not originally present in the input signal.  Spurious Free Dynamic Range (SFDR) gives the range from the fundamental (input signal) to the highest distortion component.  This is not necessarily the 2^nd harmonic, though this is most commonly the highest distortion component.  It could also come from idle tones from the A/D converter or some other generated tone.

A measure of the sum total power from all distortion components is called Total Harmonic Distortion (THD).  As shown in the graph below, this level is higher than the largest peak distortion component as it is a sum of all peaks.

These specifications all give measure to the noise and distortion components that are contributed by the A/D converter to the digital output values which are not originally present at the original analog input signals.  This characterizes the ability of the A/D converter to accurately measure the signal of interest.  In order to get an overall value of the total accuracy of the A/D converter over the desired bandwidth of interest, the combination of noise and distortion is found in the specification called Signal to Noise and Distortion, or Signal In Noise And Distortion (SINAD).

These AC specifications (SNR, SFDR, THD, SINAD) are typically all given in terms of decibels (dB) which can be specified relative to the fundamental or to full scale input.  Since SINAD captures all noise and distortion power components it can be converted to another specification called Effective Number Of Bits (ENOB).  The conversion takes place using the following equation:

ENOB = (SINAD – 1.76)/6.02

All the previous specifications were given where there was a single frequency sinusoidal input signal to the A/D converter.  Another specification that measures additional distortion components that arise from having multiple input signals to the A/D converter is called Intermodulation Distortion (IMD).  It is typically defined as the ratio between the power of the 2^nd and 3^rd order intermodulation products relative to the power of one of the original input signals.

In conclusion, these additional specifications give valuable information for evaluating the performance of A/D converters.  When measuring higher frequency signals, the best way to evaluate the A/D converter to measure them is to compare these AC specifications to ensure that minimal noise and distortion will be added to the original signal at the digital output value.

One of my colleagues was recently performing some calculations on a signal conditioning circuit that connected to a thermocouple. One of the calculations that he was doing involved the leakage current at the input pins of an auto-zero operational amplifier (in order to calculate the resulting voltage error). This resulted in a conversation that I felt was worthy of sharing with other engineers.

The two specifications that were brought into question were input bias current and input offset current. Let’s take a moment to address the actual definition of these two specifications. Input bias current is defined as the average of the currents into the two input terminals of an amplifier. Recall that convention dictates that for the input leakage, a current into the device is positive, and current out of the device is negative (except for the output pin). Input offset current is defined as the difference between the currents into the two input terminals of an amplifier.

Here it is in equation form. The two physical currents into an op amp’s inputs are:

IB_N = current into the non-inverting input

IB_I = current into the inverting input

From them, we calculate the bias and offset currents respectively:

I_B = (IB_N + IB_I)/2

I_OS = IB_N – IB_I

Rearranging gives:

IB_N = I_B + I_OS/2

IB_I = I_B – I_OS/2

The question originally came up when working with the MCP6V06 auto-zero operational amplifier. The datasheet for this device specifies a typical bias current of +6 pA at room temperature, but a typical  input offset current of -85 pA at room temperature. Without looking at the specific definition of these two specifications, these numbers may seem incorrect, but they are indeed true.

Unlike traditional op amp input stages, these auto-zero operational amplifiers have switches at the input that add a current flow path, through parasitic switched capacitances. It turns out that the current flows through the switches from one input pin to the other. So for the MCP6V06, IB_N is approximately -37 pA and IB_I is approximately +49 pA. This means that the non-inverting input is sourcing current, while the inverting input is sinking current.

Even though the average input current is relatively small (6 pA), the offset current is actually quite large. As noted earlier, this is a function of the self-correcting architecture of the auto-zero amplifier, and is uniquely different from the characteristics of a traditional operational amplifier.

Related Links

MCP6V06

There are few specifications that must be carefully considered, such as, conversion accuracy and noise performance.  Many of the ADC errors can be improved by taking the average of multiple samples to precisely determine temperature.

RA Tolerance and Measurement Accuracy

The variation in RA characteristics introduces temperature accuracy error. A 1% tolerance in RA produces a 20°C error and a 0.1% tolerance produces a 2°C error.

For lower tolerance resistors, RA must be calibrated for precision temperature measurements. In order to precisely calibrate RA, a calibration resistor can be used in place of the RTD, such as 100Ω 0.1% tolerance resistor and the equation for “ADC RESOLUTION and ADC Code Relationship” can be rearranged to determine RA.

RTD Temperature Calculation

RTDs are significantly non-linear. Depending on the RTD type and specification, the resistor to temperature conversion equations have been defined and standardized. The equation for the PT100 RTD can be found at American Society for Testing and Materials (ASTM) [1] specification number E1137E. The figure below shows the error that occurs by ignoring the 2nd and higher power errors from RTD.

FIGURE: RTD to Temperature Conversion Error

Power Supply Noise

Another source of error is the system power supply. Most power supplies for portable systems use switching regulators which generates high frequency glitches at the switching frequency of typically 100 kHz. Other sources of noise include digital switching from system processor or system oscillator. This high frequency noise can couple throughout system and directly influence the measurement accuracy. Therefore, high performance sensor applications require analog filters.

The power supply voltage, VDD, connected to the input of the LDO must be filtered using Resistor Capacitor network (RC network) with low corner frequency, approximately 1 kHz. The filtered voltage can be set to a desired level using a low dropout linear regulator (LDO). Refer to the LDO datasheet for dropout voltage specification when setting the LDO output voltage. The two RC filters provide 40 dB per decade rolloff.

FIGURE: RTD Biasing Circuit

Note that the RC filter is applied before the LDO. Typically, the Power Supply Rejection Ratio (PRSS) of an LDO is ~0 dB at higher frequencies. Therefore, It is necessary to filter the input voltage to prevent the noise coupling through the LDO to the ADC and RTD. In addition, when designing PCB layout, avoid placing digital signal traces in close proximity to the RTD biasing circuit.

Effect of RTD Self-heat Due to Power Dissipation

When biasing RTD, self heat due to power dissipation can compromise system accuracy. The effect of Self heat can be reduced by reducing the biasing current magnitude. The current magnitude needs to be sufficiently low to reduce self-heat while providing adequate voltage range and measurement resolution. Ideally, the added temperature due to self-heat must be lower than the temperature measurement resolution, TRES

To determine error due to self-heat, refer to the RTD datasheet for Self-heat coefficient specification in degree Celsius per milli-watt (°C/mW). This coefficient is used to convert heat due to power dissipation to temperature. For example, a small surface mount PT100 RTD with 0.2°C/mW self-heat coefficient would dissipate 0.002°C with 300 μA bias current at 0°C (100Ω), and 0.006°C at high temperature (350Ω). In this case, the maximum heat dissipated due to selfheat is less then 0.008°C TRES. Therefore, error due to self heat is not measurable.

EQUATION: RTD POWER

This approach was validated using Microchip’s MCP3551 ADC device. The ratiomatric solution was used with a calibrated RTD simulator to generate the data as shown in the table below.

TABLE: RATIOMETRIC TEST RESULTS USING AN RTD SIMULATOR

CONCLUSION

This discussed an RTD application which uses a ratiometric relation between the ADC LSB quanta and the RTD temperature coefficient. This was achieved using low tolerance resistor and a reference voltage to bias the RTD and ADC and measure temperature ratiometrically with 0.01°C temperature resolution from -200°C to 800°C temperature range. A 0.1°C accuracy can be achieved using a single point calibration. This approach eliminates the need for a high-performance RTD systems that require constant current source and complex instrumentation systems. This technique provides a low cost, high performance, plug and play solution for all RTDs.

Today RTDs (Resistive Temperature Detector) are the benchmark for high accuracy temperature measurement and are a key component in many high performance thermal management applications. Today we will discuss how to use a high resolution Delta-Sigma Analog-to- Digital converter, and two resistors to measure RTD resistance ratiometrically.  We will demonstrate how a ±0.1°C accuracy and ±0.01°C measurement resolution can be achieved across the RTD temperature range of -200°C to +800°C with a single point calibration.

A high resolution Delta-Sigma ADC can serve well for high performance thermal management applications. Traditionally RTDs are biased with a constant current source. The voltage drop across the RTD is conditioned using an Instrumentation Amplifier circuit.  The instrumentation amplifier can be built using discrete operational amplifiers with multiple resistors/capacitors or there are many stand-alone instrumentation amplifier offered in the industry today. Using the analog instrumentation amplifier technique requires a low noise and stable system to calibrate and accurately measure temperature. It also requires an operator for optimization on the production floor.

Utilizing a Delta-Sigma ADC solution, the RTD is directly connected to the ADC (Microchip’s MCP3551 family of 22 bit Delta-Sigma ADCs) and a single low tolerance resistor is used to bias the RTD from the ADC reference voltage (see figure below) and accurately measure temperature ratiometrically. A low drop out linear regulator (LDO) is used to provide a reference voltage.

SOLUTION

This solution uses a common reference voltage to bias the RTD and the ADC which provides a ratio-metric relation between the ADC resolution and the RTD temperature resolution. Only one biasing resistor, RA, is needed to set the measurement resolution ratio.

RTD RESISTANCE

For  example, a 2V ADC reference voltage (VREF) results in a 1 μV/LSb (Least Significant Bit) resolution.

Setting RA = RB = 6.8 kΩ provides 111.6 μV/°C temperature coefficient (PT100 RTD with 0.385Ω/°C temperature coefficient). This provides 0.008°C/LSb temperature measurement resolution for the entire range of 20Ω to 320Ω or -200°C to +800°C. A single point calibration with a 0.1% 100Ω resistor provides ±0.1°C accuracy (as shown in the above figure). This approach provides a plug-and-play solution with minimum adjustment. However, the system accuracy depends on several factors such as the RTD type, biasing circuit tolerance and stability, error due to power dissipation or self-heat, and RTD non-linearity.

Ratiometric Measurement

The key feature of a ratiometric measurement technique is that the temperature accuracy does not depend on an accurate reference voltage. The ADC reference voltage varies with respect to change in RTD resistance due to the voltage divider relation. This measurement maintains constant resolution. It eliminates the need for a constant biasing current source or a voltage source, which can be costly, while providing a highly accurate temperature measurement solution. The figure below shows circuit block diagram with the ADC reference.

EQUATION: REFERENCE VOLTAGE

Figure: RTD Biasing Circuit

EQUATION: VOLTAGE ACROSS RTD

EQUATION: ADC RESOLUTION and ADC Code Relationship

When RA = RB = 6800Ω, the bias current is ~290 μA. This provides < 0.01°C/LSb temperature resolution. As the RTD resistance varies due to temperature, the IBIAS

(biasing current) varies and temperature resolution remains below 0.01°C/LSb as shown in the figure below.

FIGURE: TRES vs. RTD Resistance

There are few specifications that must be carefully considered, such as, conversion accuracy and noise performance which will discuss in Part II.

Related Links

MCP3551