Fullscreen HTML5Med Res (??MB)HTML4

Part 1 by Richard Palmer WiFi DDS Function Generator A signal generator is one of the most useful instruments on the test bench. This flexible and easy-to-build generator provides two wide-range, low-distortion outputs. It can be controlled from its LCD touch screen or remotely by a computer, tablet or smartphone using a WiFi connection. W hile sinewave signals are commonly used for testing audio circuits, triangle, square, pulse and ramp signals have many test applications in the power, digital and linear arenas. Some commercial function generators include many rarely-used waveforms, such as Lorenz pulses and even heartbeats, but we have chosen to avoid excessive complexity in this project. Subwoofers and their crossovers pose challenges for many audio tools that can only generate signals from 20Hz to 20kHz. This unit’s outputs are DC-coupled, so it can produce very low (sub 1Hz) frequencies with very low distortion. Out-of-band frequency testing can be important in other aspects of audio design, as signals beyond the audio range may impact signals at audible frequencies through intermodulation and other effects. We have extended sinewave generation up to 70kHz to support such testing and included an intermodulation distortion (IMD) test signal. 40 Silicon Chip Tone burst and pulse testing can help identify ringing and other circuit misbehaviour. Both frequency and amplitude sweeps are supported for sine, square, and triangle waves. Additionally, duty cycle sweeps are provided for square and triangle waves, plus amplitude sweeps for pulse and step waveforms. Bursts are available for all waveforms. For sine, triangle and square waves, channel B can mirror channel A with a phase shift of 0-360°. For most other waveforms, channel B can mirror channel A, either in phase or 180° out of phase. The output terminals are floating with respect to mains Earth to avoid hum-inducing Earth loops. Like my previous lab projects, it can be remotely controlled via a web browser, see: • Programmable Hybrid Lab Supply (May & June 2021; siliconchip. au/Series/364) • WiFi Programmable DC Load (September-October 2022; siliconchip. au/Series/388) Australia's electronics magazine • Automated Test Bench/Swiss Army Knife (April 2023; siliconchip. au/Article/15736) Also like those other projects, it has a set of SCPI commands that allows it to be part of an automated test setup controlled by a program such as TestController (siliconchip.au/link/abev). I described how to use that software in an article in April 2023 (siliconchip. au/Article/15740). Automatic testing of a vast range of equipment is possible using combinations of the four instruments in this series! Performance Noise and distortion are important performance measures for sinewave signals. As shown in Figs.1 & 2, for a 2V peak-to-peak signal from 20Hz to 2kHz, distortion was below 0.0025%, rising to only 0.0055% at 10kHz. Accurate measurement beyond 25kHz was not possible due to the predominant 3rd harmonic moving beyond the Nyquist limit of my audio analyser, which samples at 192kHz. siliconchip.com.au Features & Specifications » Seven waveform options: sine, square, pulse, triangle, stepped, IM (intermodulation) & white noise » Two independent or linked channels with adjustable phase difference » Maximum output level: 11.5V peakto-peak » Sine frequency range: 0.01Hz to 70kHz » Sine frequency response: ±0.1dB, 1Hz to 50kHz (see Fig.3) » Sine distortion: <0.003% to 5kHz, <0.007% to 20kHz (see Fig.2) » Signal-to-noise ratio: >80dB » Square, triangle and pulse frequency range: 0.01Hz to 20kHz » Intermodulation (IM), white noise and step functions » Comprehensive sweeps and tone bursts » Control via 3.5in LCD touchscreen and WiFi (web browser or SCPI) » TestController definitions for automation » Optional laser-cut bench stand The number of sample points at 10kHz is only nineteen, reducing as the frequency rises. At 50kHz, there are only four sample points per cycle, with the DAC’s interpolation filter raising the number to eight. It is no wonder that it is impossible to maintain extremely low distortion with very few samples per cycle. Measured distortion below 20Hz rises to 0.0055% at 7Hz. The actual distortion could be lower, as my test equipment is AC-coupled and the input filter’s drop-off degrades measurements below 20Hz. The background noise is -89dB (unweighted) below a 2V peak-to-peak output level in the 5Hz-20kHz band, mainly comprising 50Hz mains hum and its harmonics. That figure should be better at higher output levels as the signal will be greater but the noise/hum level will remain similar. The sinewave frequency response varies by less than ±0.1dB across the entire range (see Fig.3). Further exercising the DAC and filter with an SMTE/DIN standard intermodulation signal, a 4:1 mix of 60Hz and 7kHz, the artefacts are 100dB below the 60Hz peak and 88dB below the 7kHz component, as shown in Fig.4. siliconchip.com.au Fig.1: distortion at 1kHz measures 0.0026%, with noise nearly 90dB below the fundamental. Fig.2: distortion remains below 0.003% between 10Hz and 2.5kHz, rising to 0.006% at 25kHz, the limit of our testing capability. Fig.3: the DDS frequency response is flat within 0.1dB across the whole range. Australia's electronics magazine May 2024  41 Fig.4: in the intermodulation distortion (IMD) test, the sidebands, separated from the fundamental by 60Hz, are more than 100dB below the -3.18dBV 60Hz signal, which is off-screen for clarity. The sinewave frequency is very accurate and stable, being locked to the Pico’s crystal oscillator. As the phase accumulator has a finite number of bits, there will always be a small positive frequency error due to truncation from floating point to integer values when the per-sample phase increment is calculated. However, this frequency error is less than 0.012% across the audio band. Pulse waveforms have a resolution of 5µs due to the 192kHz DAC sampling rate. The maximum square wave frequency has been limited to 20kHz. While it is possible to create pulses with a shorter duration, the filtered output becomes increasingly rounded beyond that point (Scope 1). This is due to the DAC’s internal digital filter and the analog filter’s 100kHz corner frequency. The maximum slew time of rising and falling edges is 5µs (Scope 2), largely independent of voltage. This is mainly due to the DAC’s digital filtering and interpolation. For square, triangle and pulse waveforms, the available frequencies become more granular as the frequency Scope 1: by 16kHz, the square wave output starts to become rounded. 42 Silicon Chip increases, as each phase of the waveform – rise, high, fall and low – is limited to an integral number of samples. At 1kHz, the total number of samples in a cycle will be 192; the next available step, at 191 cycles, will give a frequency of 1.0052kHz. A 10kHz square wave will have 20 samples, slightly higher than the perfect value of 19.2, resulting in a frequency of 9.6kHz. The next available step, with 18 samples, is 10.66kHz. The step from 19.2kHz to 21.33kHz is twice as large. Sinewave generation is not greatly affected by the small number of samples at high frequencies due to the DAC’s oversampling at 384kHz, which fills in the missing samples quite effectively up to 60kHz, as seen in Scope 3. Closer to the Nyquist limit of 96kHz, artefacts from the DAC’s digital filter start destabilising the waveform (Scope 4). Signal generation Traditional analog sinewave generators, such as the Wien bridge, can produce excellent noise and distortion figures but lack flexibility. Generating more complex waveforms, such as tone bursts, frequency sweeps, triangular waves and dual-tone intermodulation (IMD) signals rapidly increases analog circuit complexity. Direct digital synthesis (DDS) provides the required flexibility and can be implemented with simple circuitry at the expense of somewhat more complex software. Generating a high-quality sinewave Scope 2: interposing a single sample midway between the high and low levels largely cures signal overshoot. Australia's electronics magazine siliconchip.com.au Most of the components and connectors, including the two main modules, mount on the side of the PCB shown at right. Sockets are recommended for the microcontroller and DAC modules, while IC sockets are optional. Either of the two commonly available PCM5102A modules can be accommodated on the PCB. The photo on the left shows the LCD screen, function LEDs and user controls, which are mounted on the rear of the main PCB. poses a significant CPU speed challenge. Calculating a separate sin(x) value for each sample is relatively time-consuming, requiring several double-precision floating-point calculations. At a 192kHz sampling rate, two new samples (for both channels) are needed every 5µs. Calculating a pair of values using sin(x) on the Pico, which lacks floating point hardware, takes 16µs, making direct calculation impractical. The calculations must take less than 1µs, leaving sufficient time for other processing and housekeeping tasks such as triggering, burst and sweep management. To achieve this, a 4096-entry sine lookup table (LUT) is pre-calculated. It contains one whole cycle of a sinewave in integer format. The code steps through the table, selecting the correct value for each sample. So that any frequency can be selected, not just those that correspond to exact entries in the table, the required location in the table is calculated using a 20-bit phase accumulator with a 12-bit upper part to index into the lookup table and an 8-bit lower part used to interpolate between the two nearest table entries. While there are many ways to calculate a value partway between two points on a curved line, the quickest is a linear approximation, which treats each segment as a straight line. While it might seem that this would lead to significant inaccuracies, the substantial size of the lookup table ensures that the error, which is greatest at the point of maximum curvature on Scope 3: the raw 20kHz sinewave output of the DAC (orange trace) has steps at 384kHz due to the DAC oversampling rate. The output of the filter circuit (green trace) smooths the steps out. siliconchip.com.au the sinewave, remains below 0.0005%. That is four times lower than the 0.002% distortion figure specified for the DAC. As the Pico does not have integer division arithmetic hardware, we have avoided using division during the calculation of individual samples by the simple expedient of precalculating the inverse of any required divisors using fixed point (with 10 bits for the fractional part) and multiplying instead. That is around 23 times faster than using actual division. The intermodulation mode sums the A and B channel sinewave calculations, sending them to the channel A output. Channel B may still be used for other waveforms while this is happening. White noise is implemented as a Scope 4: by 70kHz, the DAC output has become quite ragged, and the filtered output shows visible distortion near the peaks. The scalloping of the raw signal is due to the DAC’s internal digital filter. Australia's electronics magazine May 2024  43 sequence of pseudorandom values calculated to be within the desired voltage range. The square, triangle, pulse and step waveforms all use the same pulse-­ generating algorithm internally. Pulse signals are straightforward to generate, being a sequence of linear ramps between set points. For steep slopes, the value increments by an appropriate value for each sample, while for shallower slopes, the value is incremented after several cycles. When the slope approaches one increment per sample, the range of slopes becomes very limited. In this case, increments are calculated for blocks of 50 samples, increasing both slope granularity and accuracy. Channel A provides sweep and burst capability across a range of waveforms. For burst waveforms, the output alternates between the idle value and active signal, each with the desired 44 Silicon Chip number of cycles. For sinewaves, the idle value is the DC offset, while it is the low voltage set point for square, triangle, pulse and step waveforms. For IMD and white noise, it is 0V. For IM and white noise, the on and off periods are calculated in milliseconds, and the changes are made immediately after a zero-crossing to minimise transients. For all waveforms, the output is switched on or off immediately after a zero crossing to minimise transients. Sweeps are generated as a series of stepped values between two end points. Amplitude, frequency or duty cycle can be swept as appropriate to the waveform. Sweeps may be linear or logarithmic. At each sweep step, the new waveform value starts from the last output value (see Scope 5) to minimise transients. Negative values are problematic for logarithmic sweeps, as the logarithm of negative numbers is undefined. Australia's electronics magazine Where this is detected, a value of 0.01 is substituted for the offending setting. Circuit operation As shown in the full circuit diagram, Fig.5, we use an audio DAC (digital-to-analog converter, MOD1) driven by a microprocessor to generate the signals. Following the DAC is a filtering buffer amplifier (IC1), which reduces out-of-band frequencies and increases the available output voltage. The Raspberry Pi Pico W microcontroller at the heart of this project features two CPU cores and comprehensive WiFi and Bluetooth capabilities. In this application, one core is dedicated to signal generation while the other manages WiFi, the LCD screen, EEPROM, switches, the rotary encoder and housekeeping tasks. While the ESP32 used in previous instruments in this series has sub- siliconchip.com.au stantially faster arithmetic hardware that would be useful for DDS signal generation, the simplicity of uploading programs and the low cost of the Pico W made it a better choice for this project. A PCM5102A 24-bit stereo DAC chip translates the values calculated by the Pico into voltages, with a maximum output of 2.1V RMS (5.9V peakto-peak). Thanks to a charge pump that’s internal to the PCM5102 chip, it can produce negative and positive voltages from a single supply, allowing the design to be direct-coupled, which is critical for low-frequency signal generation. Modules containing the PCM5102A are readily available online, premounted with support components on a small PCB, avoiding the need to solder an SMD component with finely spaced pins. The circuit of our recommended module is shown in Fig.6. The top of the Function Generator has four RCA sockets for the two output channels plus the trigger input and output, as well as the coaxial power connector. It is shown here on the optional stand. A key design consideration was the DDS sampling rate. At 96kHz, two sinewaves can be comfortably synthesised on the Pico. However, the practical sinewave frequencies would be limited to around 35kHz and square waves to 10kHz before significant distortion. Simultaneously synthesising two sinewaves with a 192kHz sampling rate is difficult with the Pico running at its standard 133MHz clock speed. However, stable synthesis is achieved with mild over-clocking to 240MHz. That might seem extreme, but the maximum supported clock rate is Fig.5: the Pico W microcontroller module streams serial digital audio data to the PCM5102 DAC module. Its analog outputs are filtered and amplified by dual op amp IC1 and fed to the outputs at CON4 & CON5. The Pico W also manages the LCD touchscreen, control switches and LEDs. It provides web services via WiFi for remote control as well. siliconchip.com.au Australia's electronics magazine May 2024  45 The WiFi DDS Function Generator resting in the optional acrylic stand. 300MHz, and RP2040 processors have been clocked to 436MHz and beyond without damage. I ran some tests and after several hours of operation at 240MHz, the RP2040 chip was only a few degrees hotter than it was running at 133MHz. The DAC’s output signals are amplified and filtered to reduce out-ofband components using a dual lownoise NE5532 op amp IC, increasing the available output to 4.1V RMS (11.5V peak-to-peak), which should be sufficient for most applications. The DC-coupled filter allows voltage offsets to be applied to the signals, which is particularly useful for square, pulse, stepped and triangle waveforms. For the LCD screen, 3.5in touchscreens are only slightly more expensive than the 2.8in variety and provide 50% more screen area with 100% more pixels. While creating all sorts of waveform options in software is relatively straightforward, every option requires an on-screen parameter. To keep the Fig.6: the circuit of the PCM5102-based DAC module. Using it means we don’t have to solder the SMD IC. It provides two regulators and some other necessary components, including numerous bypass and filter capacitors. 46 Silicon Chip Australia's electronics magazine siliconchip.com.au size and separation of the on-screen buttons and text reasonable, we have limited the main elements of the touchscreen layout to seven lines of editable parameters and a row of buttons across the top and bottom of the screen. A 24C256 I2C EEPROM chip stores parameters when the circuit is powered off. While the Pico has onboard flash memory that could be used to store the values, it has a write endurance measured in tens of thousands of cycles, rather than the millions of a true EEPROM. With parameters saved every 30 seconds, the flash memory could wear out after less than 100 hours of use. The four pushbutton switches are debounced in software and use the Pico’s internal pull-up resistors to sense contact closure. The rotary encoder creates significant switching noise, so it requires additional components to function correctly; two RC low-pass filters remove most of the noise while passing up to 100 pulses per second. Power supply Powering the unit posed several challenges. The Pico generates significant noise in the 2-12kHz band from its onboard switch-mode voltage regulator. This is reflected back into the 5V supply if there is any significant resistance in the path between the supply and the unit, for instance, when powering it via a USB cable. A 5V supply is required for the Pico, LCD screen and PCM5102A module, while 3.3V is needed for the trigger input protection and rotary encoder circuitry. The op amp runs from a ±9V split supply; the supply rails for the buffer amplifier must be several volts greater than the desired maximum output and referenced to the analog rather than digital ground plane. Mixed-mode circuits should have separate ground planes for the analog and digital sections. These should be joined at only one point, preferably under the DAC chip. The PCM5102A module provides this feature, while a 10W resistor between the two ground planes ensures they remain closely coupled if the unit is tested without the DAC module in place. The device is powered from an external 12V DC source, with 9V linear regulator REG1 stabilising the supply for the linear electronics, plus 5V siliconchip.com.au Parts List – WiFi DDS Function Generator 1 double-sided PCB coded 04104241, 149 × 108mm 1 Raspberry Pi Pico W microcontroller programmed with 04104241A.uf2 (MOD1) 1 micro Type-B to Type-A USB cable (for programming the Pico W) 1 PCM5102A DAC module (MOD2) 1 200 × 114 × 40mm plastic instrument case [Altronics H0378] 1 12V 500mA+ plugpack (tip positive) 1 3.5in SPI TFT LCD touch screen (LCD1) [Silicon Chip SC5062] 4 PCB-mount RCA sockets (CON1, CON2, CON4 & CON5) [Altronics P0208C/P0145A, Multicomp Pro PSG01547, Keystone 97x series, Cliff FC68371] 1 PCB-mount coaxial power socket (CON3; size to suit plugpack) [Altronics P0620, Jaycar PS0519] 1 rotary encoder (S1) [Jaycar SR1230 (D shaft) or Silicon Chip SC5601] 1 ‘scrubber’ knob [Adafruit ADA-2055 (D shaft), Multicomp Pro MP716XX (splined or D shaft)] 2 white momentary PCB-mount pushbutton switches (S2 & S3) [Altronics S1099, Jaycar SP0723, C&K D6R (no LED) series] 2 red momentary PCB-mount pushbutton switches (S4 & S5) [Altronics S1095, Jaycar SP0720, C&K D6R (no LED) series] 2 20-pin headers, 2.54mm pitch (for mounting the Pico W) 2 20-pin header sockets, 2.54mm pitch (for mounting the Pico W) 1 9-pin header, 2.54mm pitch (for mounting the PCM5102 module) 1 9-pin header socket, 2.54mm pitch (for mounting the PCM5102 module; eg, cut from a longer strip) 1 6-pin header, 2.54mm pitch (for mounting the PCM5102 module) 1 6-pin header socket, 2.54mm pitch (for mounting the PCM5102 module) 3 8-pin DIL IC sockets (optional; for IC1, IC2 & REG3) 2 A5 laminating pouches for top and rear decals, 120 microns preferred Hardware 4 10mm M3-tapped spacers 4 6mm M3-tapped spacers 2 M3 × 8mm panhead machine screws 4 M3 × 16-25mm countersunk head machine screws 4 M3 × 12-16mm countersunk head machine screws 10 M3 hex nuts 4 small self-adhesive rubber furniture bumpers [Bunnings 0262216] 1 small tube of thermal paste OR 2 TO-220 thermal washers Semiconductors 1 NE5532 dual low-noise op amp, DIP-8 (IC1) 1 24C256 256kb serial CMOS EEPROM, DIP-8 (IC2) [Jaycar ZZ8485] 1 7809 9V 1A linear regulator, TO-220 (REG1) 1 7805 5V 1A linear regulator, TO-220 (REG2) 1 MAX1044 switched capacitor voltage converter, DIP-8 (REG3) [element14, DigiKey, Mouser] 1 BAT54S 25V 200mA dual series schottky diode, SOT-23 (D1) 1 1N5819 40V 1A schottky diode, DO-41 (D2) 1 red 3mm LED (LED1) 1 blue 3mm LED (LED2) 2 white 3mm LEDs (LED3, LED4) Capacitors 3 220µF 16V electrolytic 2 10µF 16V electrolytic 11 100nF 50V X7R ceramic 4 220pF ±5% polystyrene, MKP, MKT or NP0/C0G ceramic 1 100pF 50V ceramic Resistors (all 1/4W 1% axial leaded) 4 10kW 2 5.6kW 4 4.7kW 10 2.2kW 1 1kW 4 10W Parts for optional stand 4 3mm acrylic laser-cut pieces [Silicon Chip SC6932] 4 small self-adhesive rubber furniture bumpers [Bunnings 0262216] 1 small tube of superglue WiFi DDS Function Generator Short-Form Kit (SC6942, $95): includes everything except the case, USB cable, power supply, labels and optional stand. The Pico W is supplied unprogrammed. linear regulator REG2 to supply the Pico, LCD screen and DAC module. The DAC module has an onboard 3.3V regulator to supply its needs, providing adequate filtering of any noise fed back into the 5V supply from the digital circuitry. The 3.3V rail that provides clamping protection for the trigger input and the rotary encoder debouncing is taken from the Pico’s 3.3V output pin. To reduce the 5V regulator’s heat dissipation, its input is supplied by REG1’s output (9V) rather than directly from the 12V supply. The unit draws around 210mA when driving two 10V peak-to-peak sinewaves into 600W loads, so REG1 typically dissipates no more than 630mW ([12V – 9V] x 210mA). The -9V rail for the analog electronics (primarily the op amps) is generated by a MAX1044 switched capacitor voltage inverter chip (REG3) running at its boosted frequency to avoid in-band noise. The two 10µF electrolytic capacitors set the MAX1044’s oscillator frequency to around 100kHz, well above the maximum sinewave frequency. Schottky diode D2 ensures the output voltage never rises above 0V, which could otherwise occur when operating the unit without the DAC module in place, as the +9V rail is established before the MAX1044 starts oscillating. Without D2, the rise in AGND’s voltage relative to GND is enough to exceed the limit of allowable voltage on the MAX1044’s output pin, causing it to fail. The optional laser-cut acrylic stand. The lettering on the sides has been removed from the final version. Further attention has been paid to limiting the coupling of digital signal noise to the linear elements of the project. The high-speed digital signals supplied to the DAC module have 22W series resistors to reduce noise induction due to spikes and ringing generated from any trace capacitance and impedance mismatch. The ±9V supplies pass through RC low-pass filters comprising 10W resistors and 220µF electrolytic capacitors in parallel with 100nF ceramic capacitors before being applied to the op amp. These components decouple the supplies from the digital ground plane while keeping the voltage drop Scope 5: the signal level change is minimised at sweep step boundaries to avoid generating transients. 48 Silicon Chip manageable. This allows full-scale output into 600W before any increase in distortion. Filtering DACs produce wideband high-­ frequency noise as output level changes are in discrete steps. In the unfiltered blue and orange trace waveforms seen in Scopes 3 & 4, the output is not a pure sinewave, and signal changes occur at closer intervals than the sampling rate would suggest. This is because the DAC is oversampling, smoothing out changes by producing intermediate values at 384kHz. Second harmonic sinewave dis- Scope 6: a 10kHz square wave at the DAC output (blue trace) and after filtering (green trace). Both show significant overshoot compared to the improved version (see Scope 2). Australia's electronics magazine siliconchip.com.au extremes in each half-cycle improves this (Scope 2). The waveform becomes quite rounded once the frequency rises past 16kHz (see Scope 1). The green trace in Scope 7 shows smoothing of the slope of a 5kHz triangle wave by the analog filter. Note that the filter doesn’t noticeably increase the rounding at the top of the triangle wave, which is an artefact of the DAC’s digital filter. Triggers Fig.7: the two-pole Sallen-Key low-pass filter is designed to attenuate the signal by less than 0.1dB at 70kHz but more than 20dB by 384kHz. tortion is significantly decreased by oversampling, particularly at higher frequencies. A digital filter in the DAC chip provides a 60dB reduction in artefacts above the Nyquist limit, which is half the sampling frequency. The digital filter also restricts the signal slew rate. Analog filtering is provided to further reduce unwanted high-frequency signal components. The two-pole Sallen-Key filter built around IC1 is designed to be maximally flat within the passband and just a fraction of a dB down at 70kHz (see Fig.7). The filter has a gain of two. The values of the 2.2kW resistors connecting from the DAC outputs to the input of the op amp based filters are lower than might be expected. That is because the DAC module has 470W series output resistors. The result of this filter can be seen in the substantially smoother green traces in Scope 3 & 4. Square waves are the most impacted by both the digital and analog filters. Scope 6 shows the DAC and filtered outputs of a 10kHz square wave. The DAC’s digital filter generates some overshoot at the top of the cycle and some ringing at the bottom. Setting the last sample halfway between the two Scope 7: the filtered output of a 5kHz triangle wave doesn’t have noticeably more rounding at the top than the unfiltered signal. siliconchip.com.au A short 3.3V trigger pulse is produced at the start of burst, step and sweep waveforms on channel A. In sweep mode, the trigger output is set at the start of the sweep and reset at the end of the first step. For step mode, the trigger signal alternates at the end of each up/down staircase. To enable more reliable oscilloscope triggering at very low frequencies, a trigger output pulse is also produced for the midpoints of channel A sine, square and triangle signals. The trigger pulse is synchronised to signal calculations rather than the DAC output. As Scope 8 shows, the trigger can be seen to lead the DAC output signal change by several milliseconds. This is because the calculated values are passed to the DAC using five 512 sample DMA buffers, which allow accurate timing for the transfer of samples to the DAC by the Pico’s hardware. The time lag can vary between four and five buffer times (10-13ms), Scope 8: the trigger output pulse comes 10-13ms before the generated signal since the signal is buffered before transmission by the DAC. Australia's electronics magazine May 2024  49 depending on where the DAC and CPU are up to emptying and filling buffers. At frequencies of 10Hz and below, this jitter becomes a small fraction of the waveform and triggering is relatively stable. Channel A sweep, burst and step waveforms may be triggered using an external signal or an SCPI command. When enabled, a rising or falling edge will trigger the function. Two 4.7kW series resistors, a 100pF parallel capacitor and two schottky diodes clamping the signal within the 0V & 3.3V supply rails protect the Pico’s trigger input pin from ESD or accidental application of the wrong voltage. Schmitt triggers are configured on the Pico’s GPIO inputs, reducing false triggering from noisy sources, such as switches and relays. Component selection PCM5102A DAC modules are available from many online suppliers. The PCB accommodates either of the two most common module layouts (see the photos below). The one with L-shaped pinouts is preferred, as it has separate 3.3V regulators for the analog and digital supplies to the PCM5102A IC. I tested an OPA2134 op amp, but the NE5532 is preferred as it is significantly cheaper and has slightly lower 2nd harmonic distortion across the 5-30kHz range. Some components related to the NE5532 filters need close tolerances and linear dielectrics for good performance, so make sure the parts you These are the two PCM5102 modules that can be used in this design. We recommend the one on the left; we expect it to have lower output noise as it has separate 3.3V regulators for the analog and digital functions. use meet the specifications in the parts list. If suitable leaded capacitors are unavailable, SMD capacitors with an M3216/1206 or M2012/0805 footprint can be soldered across the pads on the PCB, assuming they meet the minimum specifications. 1% tolerance resistors are adequate throughout the design. The 3.5” LCD SPI touchscreen is commonly available and is the same as used in the Micromite LCD BackPack V3. It is available from the Silicon Chip Online Shop, along with the PCB for this project and some other parts. The footprints for PCB-­ mounting RCA sockets are unfortunately anything but standard. The two most common footprints can be accommodated on the PCB and are available from multiple suppliers, eg, Altronics has P0208C or P0144A, while other major parts suppliers stock at least one of Cliff FC68371, Keystone 970 series, Lumberg 1553-02 or Multicomp Pro PSG01547. Switched or unswitched varieties can be used. The centre height of the RCA connector barrel varies between types, which will affect the case drilling. We recommend using headers to mount the PCM5102A and Pico modules to facilitate commissioning and debugging. Sockets for the three 8-pin DIP ICs are optional. Rotary encoders come with either splined or D-shaped shafts; either can be used. I prefer ‘scrubber’ knobs with a small dimple on the top. The most commonly available is the Adafruit 35mm Scrubber Knob, which fits a flatted shaft. Whichever knob you select, make sure that the shaft fitting of the encoder matches the knob’s. Next month Part two of our series on building the WiFi DDS Function Generator next month will have all the construction details, programming instructions, testing procedures and usage instrucSC tions. U Cable Tester S B Test just about any USB cable! USB-A (2.0/3.2) USB-B (2.0/3.2) USB-C Mini-B Micro-B (2.0/3.2) Reports faults with individual cable ends, short circuits, open circuits, voltage drops and cable resistance etc November & December 2021 issue siliconchip.com.au/Series/374 DIY kit for $110 SC5966 – siliconchip.com.au/Shop/20/5966 Everything included except the case and batteries, see our website for postage rates 50 Silicon Chip Australia's electronics magazine siliconchip.com.au