Fullscreen HTML5Med Res (??MB)HTML4

SmartProbe Project by Andrew Levido The SmartProbe is an extremely handy little device for making voltage & continuity measurements. It won’t replace your multimeter, but it is designed to be the first piece of test equipment you reach for when debugging or repairing a circuit. I have made the SmartProbe very small, measuring just 60mm long, 30mm wide and 15mm thick. That’s about the size of a box of matches, for anyone old enough to remember one! A probe fine enough for modern surface-mount circuits (or through-hole parts) is fitted to one end of the case, while a short flying ground lead emerges from the other. It is equipped with a 128 × 64 pixel OLED display and an audio transducer to show voltage measurements and give feedback, respectively. There are no buttons or switches. The SmartProbe switches itself on when you pick it up, and off again when it senses no movement for a few seconds. You switch between voltage measurement and continuity modes by tapping it with your index finger. The display automatically flips rightway-up however you hold it. The unit is powered by a single CR2032 coin cell that should last many months with typical use. ±0.5%, and the input impedance is around 1MW on both ranges. In the continuity mode, the SmartProbe sources a low current (around 1mA) and displays the voltage drop seen across the probes, just as your multimeter does on the diode test range. The source voltage is 3.3V, enough to forward-bias typical diodes, transistor junctions and most LEDs. If the resistance between the probes is greater than about 60kW, the display shows “OPEN”. There is an audio indication of continuity if the measured drop is less than about 1V. The continuity beep responds within a few milliseconds, which is essential for a good user experience. The SmartProbe is not suitable for use with high-voltage or mains- ­ owered circuits. While it has a degree p of input protection, it does not have the insulation or overload ratings of a good multimeter. I considered adding AC voltage or frequency measurement capabilities, but elected to keep it really small & simple. Ultra-low power design The SmartProbe operates in either voltage measurement or continuity mode. In the former, it can measure voltages up to ±50V, switching automatically between two ranges. For input voltages below ±6V, it has a resolution of 5mV or better; for higher voltages, the resolution is 50mV. Its absolute accuracy is within » Compact size (60 × 30 × 15mm) and lightweight (24g) » Measures voltage or continuity » 128×64 pixel OLED screen » Internal buzzer for continuity checking » Measures up to ±50V » Can also test diodes/LEDs and measure forward voltage » Single fine-tipped probe with a ground clip » Powered by an internal CR2032 coin cell One of the main aims and challenges of this design was to keep the power consumption when the device is ‘off’ to a level that would give meaningful battery life, while still being able to sense movement and wake up. A CR2032 cell has a capacity of about 235mAh while discharging from 3V (fully charged) to an end voltage of 2V. I set myself the goal of aiming for a shelf life of one year (8760 hours), meaning less than 27µA of idle consumption. As we shall soon see, that goal was more than met. Before diving into the circuit, it is helpful to look at the block diagram of the front end (Fig.1). This shows the device in voltage measurement mode, with just one voltage range for simplicity. We want to convert a bipolar voltage up to ±50V, appearing between the probe and clip, to a unipolar voltage between 0V and 3.3V (V1) suitable for the analog-to-digital converter (ADC). We do this by fixing the bottom (ground clip) end of the Ra/Rb voltage divider to half the supply rail, ie, around 1.65V. Voltage V1 is therefore siliconchip.com.au Australia's electronics magazine July 2025  33 Specifications Features & Specifications Fig.1: one end of the input voltage divider is fixed to ½ of the supply voltage to provide an offset so that bipolar (±) input voltages can be measured with a unipolar ADC. The offset is later subtracted by the firmware. an attenuated and buffered version of the input voltage offset by ½Vcc, as described by the equation V1 = (Vin × Rb) ÷ (Ra + Rb) + Voffset. If we also buffer and convert the offset voltage, Voffset, we can subtract it from the converted version of V1 in firmware. The resulting digital code will be a signed value proportional to Vin. The full circuit (Fig.2) shows that there are actually two dividers and associated buffers in the SmartProbe, one for each input voltage range. For the high-voltage range, it uses a 2MW/51kW divider, while the 2MW/680kW divider is for the low-voltage range. The output of each is buffered by IC1d and IC1c, respectively, and fed to its own ADC input channel on microcontroller IC2. I chose the divider resistor values such that the voltage span seen by the ADC inputs is around 3V centred on around 1.65V (0.15V to 3.15V). This allowed me to stay away from the very ends of the ADC range and avoid a potential source of errors. I am using ±0.1% tolerance resistors here, as these are critical to achieving the required resolution and accuracy. The op amps are low-cost zerodrift (auto-zero) op amps. These have a worst case offset voltage of ±10µV with just 50nV/°C drift. Since they are connected as unity-gain buffers, there is no appreciable gain error. The offset voltage is also buffered (by IC1a) and fed to another ADC channel. All three of these input buffers are identical, including protection diode pairs (D1 through D3), and a small amount of low-pass filtering on the inputs (using 1nF/100nF capacitors) and outputs (1.5kW/1nF). Most of the filtering of these signals occurs in software, as discussed below. 34 Silicon Chip When the input voltage is outside the ±6V range, the output of the low-voltage sensing circuit (IC1c) saturates, and the digital code associated with this input moves outside the expected range. The firmware automatically switches to using the high-voltage input in this case. If the high-voltage input approaches saturation, a voltage over-range warning message is displayed. The bottom ends of the dividers are fed by a current-limited buffer, IC1b. The 10W resistor provides a bit of protection to the op amp, since its output would otherwise be connected directly to the ground clip and therefore exposed to the outside world. The input of the buffer is connected to a 100kW/100kW voltage divider fed from one of the microcontroller’s GPIO pins (PA06, pin 12). This pin is configured as a digital output. If it is high, the buffer input is half of the supply voltage, as in Fig.1. If the output is low, the bottom end of the divider is effectively connected to 0V – a state which comes in handy for continuity mode. Continuity measurement So far, we have ignored the network consisting of Mosfets Q1-Q3 and the associated passive components. These form an analog switch that is off in voltage mode. In continuity mode, the offset at the bottom of the voltage dividers is set to zero, as mentioned above, and the analog switch is on. This connects the 3.3V supply to the input probe via the 3kW resistor. A voltage of approximately 3.3V is therefore present across the probes when they are open circuit. This voltage drops as the impedance between the probe falls, ultimately to zero if the probe is shorted to the Australia's electronics magazine clip. If a diode junction is connected across the input, with its anode to the probe, the forward drop of the diode will appear across the input. In continuity mode, the voltage is read by the ADC in the same way as already described, except the offset voltage will be close to zero. The equation shown in Fig.1 will still hold, but we will no longer be able to read negative voltages; that doesn’t matter in continuity test mode. You will notice that the output of buffer IC1c connects directly to the PA03 pin of the microcontroller, as well as to the ADC input via the RC filter. PA03 is internally connected to a fast comparator that drives the continuity beep tone. The analog switch The analog switch deserves a closer look. We require a switch with a very high impedance when off, so that no appreciable current flows through the 3kW resistor when measuring voltages. The switch must withstand ±50V when open, but have relatively low on-resistance when closed. I could not find a suitable off-theshelf analog switch because the voltage requirements are relatively high, so I built my own using two P-channel Mosfets with an N-channel Mosfet to drive them. The switch is open when Q3 is off, and the gates of Q1 & Q2 are held at their source potential by the 100kW resistor. Since the gate-source voltage is zero, both Mosfets will be off. You can see what happens when an external voltage is applied by referring to the left and middle diagrams in Fig.3. If the drain of Q1 is at +50V, its source will also be at this potential due to the conduction of its body diode. Both Mosfet’s gates and sources will therefore be at +50V, so they will remain off. Q2 will therefore block any current flow, since its body diode is reverse-biased. If the drain of Q1 is at -50V, the body diode of Q2 is forward-biased, leaving the sources and gates of both Mosfets at 3.3V. Q1 now blocks any current flow. When Q3 is switched on, the gates of both Mosfets gates will be at 0V. The drain of Q2 is fixed at 3.3V, so the body diode initially conducts, causing the sources of both Mosfets to rise almost to this value. The resulting -3.3V gatesource potential is enough to switch both Mosfets on, effectively shorting siliconchip.com.au Fig.2: the SmartProbe circuit is fairly straightforward, except for a few tricks related to achieving ultra-low power consumption that are detailed in the text. out their respective body diodes and switching the analog switch fully on. The ZXMP6A17E6 Mosfets I chose have a maximum Vds of -60V and an Rds(on) of less than 0.5W with a Vgs of 2.5V (interestingly, they have six pins, but the three additional ones are just extra drain connections). The total on-resistance of the analog switch will therefore be around 1W. The P-channel Mosfets see a worstcase voltage of -53.3V. siliconchip.com.au The BSS138K (Q3), with a maximum Vds of 50V, places the upper limit on switch voltage. This is a bit of a soft limit, since the 100kW resistor limits the avalanche current if Q3 were to break down. Nevertheless, this switch is what limits the nominal maximum voltage for the SmartProbe. Digital circuity The microcontroller is a 32-bit STM32L031F6 low-power Arm Cortex Australia's electronics magazine M0+ from ST Microelectronics. This has 32kiB of flash, 8kiB of RAM and comes in a 20-pin TSSOP package. Importantly, it operates at any voltage between 1.8V and 3.6V, and can be put into various low-power modes where its current draw reduces to single-digit microamp levels while still retaining RAM contents. The display connects to the microcontroller via an I2C interface. This is a 128 × 64 pixel white OLED screen that July 2025  35 Both sides of the SmartProbe PCB. The flat flex cable on the back of the OLED is soldered to the PCB and the screen is then held down with double-sided tape. measures just 34 × 22 × 1.5mm. These displays are readily available for just a few dollars each from AliExpress. They have a SH1106 control chip onboard, and can be controlled via an 8-bit parallel or SPI/I2C serial bus. The display only needs four 100nF capacitors and one resistor added to create the necessary internal voltages (all shown to the right of DISP1). The OLED current is set by the resistor. I have used 510kW, which gives a current of about 10µA per pixel. This is nice and bright, but means that the display could draw as much as 82mA (128 × 64 × 10µA) if all pixels were on. In reality, the measured current stays under 20mA or so, since we never light more than about 25% of the pixels simultaneously. Nevertheless, the display is the main current consumer in the circuit and has to be completely shut down when the SmartProbe is inactive. To help ensure it comes up reliably when awakened, I have wired its hardware reset pin to a microcontroller GPIO pin (PC15). The audio transducer, MB1, is a small magnetic beeper that requires an AC signal to operate. This means it can deliver a variety of tones if driven at different frequencies. I have used one of the micro’s PWM outputs (routed to the PA05 pin) to drive it via Mosfet Q4. Using a PWM output allows me to set the tone by varying the PWM carrier frequency and manage the volume, and more critically, the current consumption, by limiting the duty cycle. Even so, it presents a significant (relatively speaking) load on the supply. For this reason, I used it only when I think it is really necessary. Accelerometer Because the accelerometer is the only component in this circuit that is always operational, I needed to choose it carefully. I selected the LIS2DW12 three-axis “digital motion sensor” for this application because it is specifically geared toward ultra-low-power applications. This chip (IC3) contains a three-axis MEMS accelerometer and a heap of signal processing hardware that can be configured to detect device orientation, free-fall events and tap or double-­tap events on any axis. It can detect activity and put itself into a lowpower state when it senses inactivity, waking itself up again autonomously. I took advantage of the orientation function to flip the display the right way up, the single tap function to change modes and the activity/inactivity function to turn the SmartProbe on and off. The device supports a range of sampling rates and draws anywhere between 500nA and 90µA when operating. Lower data rates result in lower operating currents, but some features we need won’t work at the very lowest data rates. For this reason, we run the accelerometer at 400 samples per second when active, dropping to 200 samples per second when ‘off’. This gives us an ‘off’ power consumption of somewhere around 12-20µA. The data sheet says the consumption will be 12µA in the mode and data rate we use, but that is specified at 1.8V and 25°C. The data provides no help in understanding what the consumption will be at higher temperatures or with voltages up to 3.0V, as it will experience in our circuit. No data usually means you can safely assume it will be worse. My measurements show consumption closer to 16-20µA with our battery voltage range and my (unairconditioned) room temperature. While that is higher than the published figures, I think it is still pretty amazing performance considering the chip is taking three 14-bit accelerometer samples every 5ms and pushing them through a fairly complex digital signal processing chain. Hardware-wise, the LIS2DW12 is very nice; it has just 12 pins, requires no external components and costs just $2.50 in single quantities. The downside is that it is only available in a 2 × 2mm leadless package. Fortunately, it does not have a thermal pad, so it is easier to hand-solder than some chips I have come across. Fig.3: the left and middle diagrams show the voltages on the analog switch in the off state with +50V and -50V applied to the input, respectively. The rightmost diagram shows them when the switch is on. 36 Silicon Chip Australia's electronics magazine siliconchip.com.au It is a complex device from a firmware perspective, but I took the time to write a (reasonably) comprehensive driver, since I intend to use this chip again. Power supply The power supply scheme is straightforward, as shown in Fig.4. There are two power rails: Vbat, which is derived from a lithium coin cell and is always available, plus a 3.3V (3V3) rail that is only available when the SmartProbe is on. The 3.3V rail is derived from the coin cell via a boost converter based around IC4, a TPS61033. The boost converter is enabled by the PWR_ON signal from the microcontroller, so the display, analog front end and beeper are only powered when the SmartProbe is on. The shutdown leakage current of the TPS61033 is specified at 0.1µA, so well within our meagre power budget. The Vbat supply is the diode-OR combination of the coin cell voltage and the 3.3V supply. The upshot of this is that Vbat will be approximately 3.3V while the device is on, but will fall to very near the battery voltage when asleep. Both the accelerometer and microcontroller will happily operate at any voltage between 3.6V and 1.8V, so this is not a problem for them. I used a pair of schottky diodes to combine the supplies to minimise the forward drop, keeping it to only about 0.2V at the low currents drawn in standby mode. Firmware The firmware architecture is shown in Fig.5. The software consists of a main loop, shown at the top, and four asynchronous tasks triggered by interrupts, shown at the bottom. Some of these asynchronous tasks communicate with the main loop via a few shared data registers and flags. When an interrupt occurs, the processor stops what it is doing and starts running code from the appropriate interrupt service routine (ISR). When the code starts from reset, the microcontroller core and onboard peripherals are initialised. This includes such things as setting up the microcontroller clocks, the I2C peripheral, PWM, timers and the like. This only needs to be done once, because when the microcontroller is stopped, the RAM contents and register data are retained. After this, the external peripherals are initialised via their drivers. This involves enabling the boost converter, initialising the accelerometer and the OLED display. This step is repeated each time the SmartProbe awakens because the display driver’s configuration registers are lost when the 3V3 rail is disabled. We also reconfigure the accelerometer, even though it is never shut down completely. We do, however, disable some of its functions before putting the microcontroller into sleep mode, so it is easier just to reprogram it completely when it wakes up. Once everything is initialised, we enter the main loop proper. Here, we check if the accelerometer has detected a period of inactivity and put itself into low-power mode. If it has, we proceed to put the SmartProbe into ‘off’ mode, as described below. If it is still active, we check if fresh data is available from the ADC sampling task. If not, we loop back and repeat the cycle, checking continuously for inactivity and new data. When fresh ADC data is available (approximately twice per second), the display is updated according to the operating mode and taking into account the orientation of the SmartProbe. This routine also takes care of the auto-ranging and over-voltage detection. ADC sampling The ADC sampling routine operates independently of the main loop, triggered every 500ms by a repeating timer. The analog-to-digital converter (ADC) peripheral within this microcontroller is extremely flexible. It is a Fig.4: the 3.3V power rail is derived from the coin cell voltage via a boost converter. This is disabled when the SmartProbe is ‘off’ but the microcontroller and accelerometer remain powered by the Vbat rail. siliconchip.com.au Australia's electronics magazine Fig.5: the firmware consists of a main loop and four asynchronous tasks, as described in the text. They communicate with each other via a few shared data and status registers (not shown). 12-bit successive approximation converter with an input multiplexer that can handle up to 18 input channels. Sixteen of the channels can be connected to input pins (not all of which are available on the 20-pin version of the chip), while two can be connected to internal sources. One of these is a 1.2V bandgap voltage reference. The ADC also includes a zero-calibration feature, which we use each time the SmartProbe becomes active. The ADC can be configured to July 2025  37 Table 1 – STM32L031F6P6 power saving modes Power Mode Core Peripherals RAM/Registers Wake-up Run On On Retained Not Applicable Sleep Off On Retained Any Peripheral Stop Off Off Retained External interrupts, low-power peripherals, RTC, watchdog Standby Off Off Lost RTC, watchdog, wake-up pin automatically scan and convert channels in a sequence. It also has a hardware oversampling capability. If this is enabled, the ADC will take a series of samples and average the results for each channel in the sequence. I used these features to create a sampling regime that eliminates a lot of the mains-frequency interference that would otherwise make achieving stable readings difficult. If we take many samples of a signal over an integral number of mains cycles and average them, any mains frequency component will average to zero. This is because the average of any sinusoidal signal over a full cycle is zero. So, if we make the ADC sample and average each input over one or more 20ms intervals, any 50Hz component will be eliminated. In the SmartProbe, the clock division options available to us mean that we can’t quite do this perfectly. The best we can manage is to take the 256 samples of each input over a period of 59.05ms – very close to three mains cycles. This means the mains cancellation will not be perfect, but we should still reduce it by 34dB (50 times) or thereabouts, which helps a lot. This oversampling and averaging also serves as a simple low-pass filter, helping to smooth out any small perturbations in the voltage being measured. The ADC is therefore set up to convert four inputs in sequence: the high and low range voltage measurement inputs, the offset voltage and the internal reference. Each is sampled 256 times, and the results averaged twice per second. Once configured, all this happens more-or-less automatically. An interrupt is triggered at the end of each conversion sequence, at which point we need to translate the averaged ADC readings into absolute voltages that we can display. The ADC output is ratiometric with the Vbat power rail – that means its output code is a measure of the input voltage as a fraction of Vbat. Vbat is nominally 3.3V when the 38 Silicon Chip SmartProbe is on, but this voltage is not regulated to the extent that would allow conversion to absolute voltages at the level of accuracy we want. Fortunately, there is a way around this, using the internal bandgap reference. This has a nominal output of 1.2V and pretty good stability. When the chip is manufactured, the value of this internal reference is measured by the ADC while the supply voltage is fixed at a fairly precise 3.00V, and the resulting code burned into non-­ volatile memory. The firmware uses this code, together with the real-time measurement of the internal reference, to calculate the Vbat voltage on each measurement cycle. Knowing the supply voltage allows us to determine the absolute value of the input voltages. It is then only a matter of subtracting the offset voltage from each of the input voltages to determine the voltage across the lower resistor in each divider as a signed integer in units of millivolts. These are later adjusted for the divider attenuation in the display update routine. When new values are calculated, they are stored in shared memory, and the flag set to let the main loop know that new data is available. Continuity mode & beeper The accelerometer is set up to detect single-tap events in the ‘vertical’ axis of the smart probe (ie, the top or bottom surface when looking at the display) and assert the interrupt line. The associated interrupt service routine responds by switching modes: opening or closing the analog switch and setting the offset voltage appropriately. A short tone sounds when changing modes – a higher frequency when switching to voltage mode, and a lower frequency when switching to continuity mode. The beeper driver makes use of the M0+ core’s dedicated tick timer, which is usually set to provide a system tick interrupt every millisecond. A tone is initiated by calling a driver function Australia's electronics magazine specifying the desired frequency and duration. The function starts the tone playing at the appropriate frequency, then returns. The tone is automatically terminated after the requisite number of 1ms ticks elapse. It is important (to me anyway) that the continuity beeper has a very fast response. The ADC samples are only updated every 500ms, which is fine for the display, but way too slow for the beep. Therefore, I used a comparator, as mentioned earlier. One of the two onboard comparators is configured to compare the low-range input voltage with a fixed internal voltage set to ¼ of the internal 1.2V reference. If the voltage at the input pin falls below 0.6V, a flag is set to indicate continuity. If it is above this level, the flag is cleared. The comparator output flag is sampled every system tick and the continuity beep is sounded if it is set. Low power operation We mentioned above that the accelerometer is configured to detect a period of inactivity and autonomously put itself into a low-power mode. When the microcontroller detects this has occurred, the rest of the circuit must be shut down until the accelerometer indicates activity has resumed. We have seen that the accelerometer consumes up to 20µA in its low-power mode, and we have a total design target consumption of 27µA or less. This leaves us with just a few microamps for everything else, including the microcontroller. The Cortex M0+ architecture supports a variety of low-power modes with differing levels of power consumption. The trade-off for lower power is longer wake-up times and more limited wake-up functionality. Table 1 shows a (very) simplified chart of the available modes and their key differences. Every one of the modes shown in the table has several variations, and it is entirely possible for a lower power mode to consume more than a higher power mode siliconchip.com.au depending on the exact configuration. “Run” is the normal operating mode of the microcontroller and has the maximum power consumption; the core and all peripherals are operating. The easiest way to reduce power in this mode (or any mode where the clocks are operating) is to reduce the frequency of the system clock below its maximum of 32MHz. I use a 3MHz system clock in the SmartProbe for this reason. In “Sleep” mode, the core clock is disabled, but the peripherals remain fully operational. RAM and register contents are preserved. This allows for a very fast wake-up (in the order of 0.35µs) but comes at the cost of around 1mA current consumption at 16MHz. As the peripherals continue to operate, pretty much any of them can wake the processor up. “Stop” mode, on the other hand, has the potential to reduce power consumption to the sub-microamp level. This is the mode I used for the SmartProbe when it’s ‘off’. Here, the core and most peripheral clocks are halted; only the real-time clock and watchdog timer continue to run if they are enabled (which they aren’t). Volatile memory is retained. Several possible sources can wake the microcontroller from stop mode, including interrupts triggered by the states of external pins changing, which is what we use. The final low-power mode is “Standby”. In this mode, almost everything is powered off, including the RAM and almost all the registers. Only a very limited selection of wake-up sources is available, and the wake-up time is the longest. Putting the processor into stop mode is in itself not enough to get the current consumption down to the very low levels we require. There is quite a lot to do both within the microcontroller and in the external circuit before executing the instruction that halts the processor. Externally, we shut the OLED display off via an I2C command, shut down the beeper PWM and turn off the boost converter. Internally, we stop the ADC conversion process and the timer that triggers it, disable the bandgap reference and the buffers that feed its output to the ADC and the comparator. We also limit the functionality of the accelerometer to just detecting activity or inactivity. siliconchip.com.au A close-up of the probe we used for our SmartProbe. We also put most of the I/O pins into analog input mode so they look like high-impedance inputs, minimising any leakage currents that might otherwise occur. The data sheet suggests that the digital input schmitt trigger buffers can be a source of leakage – hence using the analog input mode. In fact, leakage current is our number one enemy in ultra-low power circuits such as this, and it can come from some quite obscure sources. For example, consider the 4.7kW I2C pullup resistors. These are pulled up by one of the microcontroller’s GPIO pins instead of being directly connected to the power rail (Vbat) as one might do normally. Fig.6 shows why we have to do this. The I2C bus connects to the micro and the accelerometer, which remain powered up in stop mode, but also to the display, which does not. The display driver’s I2C inputs are internally protected by diodes connected as shown in the figure. In normal operation, these prevent the input pin rising more than about 0.6V above the 3.3V supply or falling more than 0.6V below ground. However, when the boost converter is off, the display driver’s positive power rail is at 0V, allowing a leakage current path from Vbat to ground via the I2C pullups, as shown by the red paths. If Vbat was at 2.5V and the protection diode forward drop was 0.5V, there would be more than 850µA leakage in total. This is 30 times more current than our target, so clearly not acceptable! Powering the pullups from a GPIO pin that can be put into a high-­impedance state eliminates this problem. Once all the internal and external extraneous current consumers are dealt with, we are ready to stop the microcontroller core and peripherals. We just have to set or clear a few bits in various control registers to ensure the core enters the correct low-power mode and that it wakes up due to the right stimulus with the right clock source. We have to disable all interrupt sources except the external interrupt pin associated with the accelerometer, and globally disable interrupts. Finally, we can execute a “Wait for Interrupt” (WFI) instruction that stops the core until an interrupt is received. It might seem odd that we globally disable interrupts if we want to wake up due to an interrupt, but the way it works is that the peripheral’s interrupt flag does the waking (in this case the external interrupt) regardless of the state of the global interrupt enable flag. By disabling global interrupts, we ensure that when the processor wakes up, it continues executing code where it left off and not in an ISR. On resumption, we have to undo all the work we did before entering low-power mode, restoring the I/O pin states, enabling the boost converter and the internal regulators and buffers we disabled. We then reinitialise the drivers to start the ADC, the display and all the rest of the stuff necessary to resume operation. Construction Fig.6: if the I2C bus was pulled up to Vbat, significant leakage currents would flow through the pullup resistors and display driver chip’s input protection diodes when the 3V3 rail is off. Australia's electronics magazine The SmartProbe is built on a small double-sided printed circuit board coded P9054-04 that measures 54.5 × 29.5mm. Both sides of the board are fairly tightly packed with surface-mounting parts, although most are M2012/0805-sized (2.0 × 1.2mm), so pretty easy to handle. The overlay diagrams, Figs.7 & 8, show where everything goes. Start assembly by mounting the trickiest part, the accelerometer (IC3). If you are using solder wire, you will need to apply some flux paste and a July 2025  39 thin layer of solder to the pads first. Try to get roughly the same amount of solder on each pad – if there is too little on one or two, you risk an open-circuit connection. If you elect to use solder paste, try to get a more-or-less even smear across the footprint. You can then reflow the chip using something like a hot air wand, making sure to get it the right way round. Despite its small size, I found this package fairly forgiving when it came to assembly. Next, solder in the boost converter chip (REG4) and then the optional programming connector, CON3, if you will use it (it isn’t necessary if you get a pre-programmed chip from our Online Shop). They should solder in fairly easily using a fine tip soldering iron, some good flux and a bit of solder wick to clean up any bridges. Mount the rest of the components on the top side of the board, except the coin cell holder and the probe connector (CON1), in the order you prefer. I tend to install the finer-pitch and smaller parts first, working my way up to the larger ones. Flip the board over and install the six passives on the back side, plus the audio transducer. Finally, add the coin cell holder and the probe connector (CON1) to the top side. It is a good idea to check your work and clean up the board with a bit of isopropyl alcohol or another solvent at this point, before installing the display. The display’s flat flex cable is soldered directly to the row of pads on the back of the board. Make sure the pin 1 designator on the flex is aligned with the small dot on the board. This should correspond with the display being face-up when folded back on itself. It should be face-down when the flat flex is straight, and it should extend over the side of the board nearest the row of pads. Align the flat flex so that about 1mm of the PCB pads are visible, and secure it in place temporarily with a couple of bits of adhesive tape. Double-check everything lines up and tack the display in place by soldering a couple of the pins. I suggest using two of the signal pins for this, rather than the pins connected to the ground plane, as they require a lot more heat. If everything still looks good, go ahead and solder all the pins, taking care not to create any short circuits. You can remove the tape and clean up with a little solvent if you need to, but keep it away from the display itself. If you intend to program the microcontroller in-circuit, do that now. You will need to make a suitable adaptor to connect the programmer to the flat flex connector. See the details in the accompanying panel. You will have to insert a coin cell (or otherwise power the microcontroller) while it is being programmed. If your microcontroller came preprogrammed, you should skip this step. You can now perform a quick test by inserting a coin cell into the holder. You should hear the start-up beep, and the display should come to life in voltage mode with a reading close to 0.000V. The least significant digit (millivolts) and the sign may move around a little, but the rest of the digits should be zero. If you do nothing, the probe should Figs.7 & 8: follow these diagrams to populate the PCB. Start with the accelerometer (IC3), as it is the fiddliest part to mount. 40 Silicon Chip Australia's electronics magazine siliconchip.com.au switch itself off after five or six seconds, and it should wake up again you give it a bit of a jiggle. This is enough to test the basic functionality. If everything is OK, you can remove the coin cell and fix the display in place with a small piece of double-sided foam tape. Fold it over and align the edges of the display glass with the outer rectangle on the PCB silkscreen. Debugging If there is a problem, do some troubleshooting before fixing the display in place. If the unit appears dead, first measure the 3V3 rail to make sure it is working when the unit is awake. If 3.3V is not present, check that the battery voltage is getting to it and the boost converter enable pin is high. The latter is a sign that the micro is at least trying to start the converter and means the problem is in the boost converter itself or there is a short on the 3V3 rail. If the 3V3 rail is fine and you hear the start-up beep, but the screen remains blank, the problem is probably in the I2C bus or with the display. Check the associated components and the soldering of the display connector. It is possible to get shorts under the flat flex that you can’t necessarily see. Use a multimeter to look for (unwanted) shorts between adjacent pins if you suspect this may be the case. The firmware does have a fair bit of error-detection built in. If an error occurs, the beeper emits a short, low tone (it is fairly quiet, so listen carefully) and displays a small fault icon on the screen. If this says “ACC!”, the code encountered a problem communicating with the accelerometer. If it says “DIS!”, the problem is with the display communication, and if it says “SYS!” the problem is a processor exception, so probably related to corrupt code. Mechanical assembly Once everything is working as expected, you can prepare the case. Mark out and cut the opening for the display and the sound hole according to Fig.9. The display aperture can be cut by drilling a series of holes inside the marking and finishing up to the line with files. I used a file to put a chamfer around the display hole – the exact dimensions are not critical as this is purely cosmetic. Drill the holes in each end of the siliconchip.com.au Parts List – SmartProbe 1 Hammond 1551JBK 60 × 35 × 15mm black ABS enclosure [Altronics H9003] 1 double-sided PCB coded P9054-04, 54.5 × 29.5mm 1 BC-2600 SMD CR2032 cell holder (BAT1) 1 CR2032 3V lithium coin cell (BAT1) 1 Wago 2946-2060-471/998-404 single entry SMD terminal block (CON1) 1 Molex 503480-0800 8-pin flat flex connector, 0.5mm pitch (CON3; optional, for ICSP) 1 X30654 128×64-pixel 3.3V graphic OLED screen with SH1106 controller (DISP1) [AliExpress 32890183042] 1 WLPH201610M1R0PP or equivalent 1µH 2A+ inductor, SMD M2012/0805 size (L1) 1 CMT-0525-75-SMT-TR SMD magnetic transducer buzzer (MB1) 1 6mm-long No.6 self-tapping screw 1 size 4 straw sewing needle 1 alligator clip 1 length of medium-duty hookup wire 1 short length of 2mm diameter heatshrink tubing 1 small piece of double-sided foam-cored tape Semiconductors 1 TP5534-SR quad chopper-stabilised op amp, SOIC-14 (IC1) 1 32-bit STM32L031F6P6 microcontroller programmed with 0411025A.HEX, TSSOP-20 (IC2) 1 LIS2DW12 ultra-low-power 3-axis accelerometer, LGA-12 2 × 2mm (IC3) 1 TPS61033DRLR adjustable boost regulator, SOT-583 (REG4) 2 ZXMP6A17E6TA 60V 2.7A logic-level P-channel Mosfets, SOT-23-6 (Q1, Q2) 2 BSS138K 50V 220mA N-channel logic-level Mosfets, SOT-23 (Q3, Q4) 4 BAV99 ultrafast dual series diodes, SOT-23 (D1-D4) 1 SDM40E20LC 20V 400mA dual common-cathode schottky diode, SOT-23 (D5) Capacitors (all SMD M2012/0805 size 50V X7R ceramic unless noted) 1 33μF 16V D-case tantalum 4 10μF 16V 8 100nF 1 10nF 5 1nF 1 100pF C0G/NP0 Resistors (all SMD M2012/0805 size, 1% unless noted) 2 2MW ±0.1% 1 1MW 1 680kW ±0.1% 1 510kW 1 220kW 5 100kW 1 51kW ±0.1% 2 4.7kW 1 3kW 3 1.5kW 1 1kW 1 10W Optional SWD flat flex adaptor 1 double-sided PCB coded P9045-A, 35 × 25mm 1 Molex 503480-0800 8-pin flat flex connector, 0.5mm pitch (CON1) 1 CNC Tech 3220-10-0300-00 10-pin, 1.27mm-pitch box header (CON2) 1 Molex 0150200079 8-way, 0.5mm pitch 100mm flat flex cable Australia's electronics magazine July 2025  41 SWD Programming Adaptor There was no space on the main PCB for the standard ST Micro SWD/JTAG programming header, which is a 2×5pin miniature shrouded box header (1.27mm pin pitch). Thus, a more compact 8-way FFC connector was used. This small PCB adapts that to the more standard connector so that a programmer can be connected via a ribbon cable with an IDC plug. Its circuit is shown in Fig.a. Because pins 7 & 8 are not connected to the main board, only the single-wire debug (SWD) protocol is supported, not JTAG. Importantly, note that the pinout of CON1 is reversed compared to CON3 in Fig.2. That’s because the FFC is inserted flat, such that the connections are reversed between those two connectors. The adaptor is built on a double-­ sided PCB that’s coded P9045-A and measures 35 × 25mm. The assembly should be straightforward, referring to the overlay diagram, Fig.b. The TP-X pin is provided to allow a custom debugging signal to be generated from the microcontroller. When using it, make sure that the FFC cable is inserted with the correct orientation between the two ends. The easiest way to check this is to verify continuity between the grounds of the SWD Adaptor board and target board before connecting the programmer. Fig.a: this SWD adaptor connects to the main circuit (Fig.2) via a flat flex cable (FFC) and allows a standard ST Micro programmer to connect via CON2. The FFC is connected such that it reverses the connections, making CON1’s pinout correspond to CON3 in Fig.2. A close-up of the Adaptor’s flex cable is shown above and the finished PCB below. We have used some tape on both ends to provide extra rigidity. eye off your needle if it is wider than about 1mm; otherwise, it may be too big for the connector. The connector I used for the probe has a spring operation. You press down on the small divot at the top to insert or remove the conductor. When released, the connector firmly grips the probe. That, plus the close fit of the hole in the case, is enough to hold the probe solidly in place. Finally, you need to clip out one of the two PCB mount bosses inside the case as it interferes with the display. The one to remove is diagonally opposite the sound hole. The PCB is secured to the case with a #6 × 6mm self-tapping screw – see Figs.10 & 11. You should solder the ground lead to the PCB before you finally mount the board. It solders in from the top (battery) side, then loops down and up again through the two strain relief holes to emerge on the same side. Thread the wire through the hole in the case before installing the board into the case. I used a ground lead with a connector on the far end, so I had to thread the near end through the case before soldering it. Remember to secure the lid to the case with the two screws provided. Coin cells are extremely dangerous for children, and it is mandatory that they are only accessible with the use of a tool. You should also be very careful about where you store them. Keep them in their special child-­resistant packaging until they are required and always well and truly away from inquisitive little hands. Using the SmartProbe Fig.b: the SWD adaptor PCB has just two components, the connectors, plus three test points. case to suit the diameter of the ground lead and probe you choose (the dimensions shown are what we used). Try to make the probe hole a close, but not tight, fit with the probe, including its insulating sleeve. If it is too loose, the probe may wobble around. I made my probe from a sewing needle. These are nice and sharp, so good for probing surface-mount parts, and 42 Silicon Chip fairly hard, so they last a while. I used a size 4 straw-type needle, but any needle with a diameter between 0.5mm and 1.0mm should do. I covered the needle in heatshrink tubing, leaving it about 5mm short of either end. The eye of my needle was a similar diameter to the shaft, so I could insert it through the case and into the connector as-is. You may have to cut the Australia's electronics magazine There is no need to calibrate the SmartProbe, but you can check its operation fairly easily with basic test equipment. The absolute accuracy is measured by setting a bench supply to some voltage near the middle of each range (I used 2V and 30V) and comparing the SmartProbe reading to that from a known good meter, preferably one with five or more digits. The three units I built were all well within ±25mV on the 5V scale and 250mV on the 50V scale (0.5% of full scale in each case). You can get an idea of its precision by taking a series of readings (the more the better) and looking at the variation between them. Successive readings should not differ more than about siliconchip.com.au ±5mV and ±50mV on the two ranges, respectively. Keep the multimeter connected when you do this, to make sure the measured voltage does not change. I measured the current consumption of the three units I built. The maximum current when on was between 15mA and 25mA, depending on the cell voltage. This means the average battery life when on will be around 12 hours. When off, the current consumption was always less than 25µA, corresponding to a shelf life of about 387 days. Of course, neither of these is a realistic scenario. However, if we assumed an on-time of 30 minutes per week (remember it switches off after five or six seconds of inaction, so this is 30 minutes of actual measurement time), a fresh cell should last a little over 4 months. The SmartProbe won’t replace my multimeters, but if it becomes the first piece of test equipment that I reach for, SC I will consider that a success! Fig.9: drill the holes in the ends of the case to sizes that suit your probe and ground wire. You want a close (but not tight) fit, so the probe is held firmly in place. The programming adaptor connected to the prototype SmartProbe. Figs.10 & 11: the PCB is secured in the case with a #6 × 6mm self-tapping screw. The probe is inserted through the hole in the case and into its connector after the PCB is mounted. Ensure the lid of the case is secured with the supplied screws to comply with the safety requirements for coin cells. siliconchip.com.au Australia's electronics magazine July 2025  43