Fullscreen HTML5Med Res (??MB)HTML4

PIC n’Mix Mike Hibbett’s column for PIC project enlightenment and related topics Audio Spectrum Analyser – bandwidth expansion I t’s a bumper issue this month as we complete our Audio Spectrum Analyser project. The circuit is based around a dsPIC33 DSP (digital signal processing) microcontroller and a low-cost 320 × 240 colour LCD display sourced from eBay. The software has been developed using Microchip’s free FFT (fast Fourier transform) library, and Adafruit’s free graphics library that was compatible with our display. Those libraries provided the complex software, leaving us with figuring out how to use and modify them to work with our processor of choice, and our schematic. Nowadays, for the many projects that make use of freely available pre-existing source code, that modification work is the hard part. Once done, adding new features like changing the operation of the system in response to a keypress or drawing cool logos is a relatively simple task, fun to do and highly rewarding. We complete the project this month by adding functionality to the buttons, testing out the new LDO voltage regulator, and building a 3D-printed enclosure. Display recycling The first step was to salvage the LCD display from the PCB assembly of the previous design; we had only purchased one LCD module and it would take weeks for a new one to arrive from eBay. Separating the boards was done by cutting each pin from the two headers binding the PCBs together. The next step was to remove the pins one by one from the LCD module with the aid of a soldering iron and then cleaning up the holes for fitting new a new header strip. When cutting off header pins, leave enough length on the pins in the LCD module so that you can easily pull on them with pliers. Once done and with the plastic spacer carefully cut away, each hole was cleaned up with solder wick to allow a new 0.1-inch strip of header pins to be inserted. The original PCB was discarded. We could have salvaged the processor, but at a cost of just a few pounds, we did not want to take the risk of building a circuit using an IC with heat or mechanical damage. True enough, with care, removal is possible, and you may find it an interesting desoldering challenge! We, however, were on a tight timescale, so the board was discarded. (We did salvage the electret microphone, a simple two-lead device – plus, it was the only one we had. Removal is a simple job of heating both leads simultaneously. This could be done quickly with a low risk of damaging the part.) New PCB Now we can start to populate the new PCB. First though, let’s look at the PCB under the microscope to see the manufacturing quality. As you can see in Fig.1, the alignment of the key manufacturing processes is excellent. The drill holes are very well centred in the pads, and the solder resist layer – which is very important to help prevent solder bridges between component pads – is both well aligned around the pads, and also hugs very close to each pad – maximising the amount of this protective layer that is available. This is a very good job by PCBWay. Choose your PSU Fig.1. New PCB, close up. The solder mask and drill alignment to the copper are excellent. 48 A key point to note in the schematic, shown in Fig.2, is that we now allow for two power supply choices – if you build this circuit, populate only one! The choice is based around IC1, a switching regulator that has a wide input voltage range, and IC3, a simple linear regulator. We discussed the pros and cons of each in the previous article; this month, we are populating the linear regulator version, as it offers the lowest noise into the rest of the circuit (the original PCB used the switching regulator, and we were disappointed by the additional noise, although it did not appear to limit performance.) The linear regulator circuit takes an input of 3.3V to 6V – ideal for four AA batteries – so we will power the circuit from this supply. We will also validate on a bench power supply what the current consumption is across that voltage range, so we can get an expectation of the battery life. The inclusion of an on-off switch in the enclosure design will ensure that we prolong battery life as long as possible. Returning to FFTs We start our design upgrade by considering the more complex requirement, which affects both hardware and software change – moving from a maximum input signal of 1kHz to a maximum of 20kHz, covering the full human hearing range. Let’s recap on our hardware design, as that limits some of our design choices. First, our spectrum analyser display is fixed at 256 data points, as we only have 256 horizontal pixels to play with. There is no point adding more data points; we cannot display them. Using extra data points would mean having to average those extra data points into one – which is no better that simply using a lower FFT size. So, a 256-point FFT remains our design specification. What we are changing, however, is the sample frequency, the rate at which we perform an analogue-to-digital conversion on the input signal. Previously, we used a sample rate of 2kHz, giving us the ability to display signals up to 1kHz. We would like to improve the design to be able to display signals up to 20kHz (now way beyond your author’s hearing range,) which will require us to increase the rate at which we perform analogue-todigital conversions from 2kHz to 40kHz. That is well within the capabilities of the dsPIC33’s ADC peripheral and should pose no problem to our application, as it does nothing else while collecting ADC samples. Practical Electronics | May | 2020 PIC n’ Mix files I C3 MCP 17 00T -3302E/ T T 3 C14 10µ F X 5R 10V 3. 3V – 6 V inp ut regulated P S U VIN 1 VOUT 2 The programming files discussed in this article are available for download from the PE website. C15 10µ F X 5R 10V GN D A nalogue p ower sup p ly f ilter 1. 5V – 5. 5V inp ut regulated P S U B attery B attery+ 2 B attery– 1 JP I C1 T P S 6 109 7 A -33DB VT 3. 3V 5 A _ 3. 3V L1 S RN 3010T A -100M 1 5 C1 100nF 3 C2 10µ F X 5R 10V 2 L VOUT R6 0Ω VIN EN C4 100nF GN D C3 10µ F X 5R 10V k C12 0F (op en circuit) S D med ia R5 10kΩ C13 0F (op en circuit) C5 10µ F X 5R 10V 4 Mic 1 3 2 R4 2.2 Ω 1 JP 2 N o connections used f or mech anical sup p ort only P I Cki t 3 H ead er 6 5 4 3 2 8 R1 . kΩ 1 MCLR 2 3. 3V 3 GN D 4 P GD 5 P GC 6 7 1 JP PJ 1 JP 1 R3 . kΩ 8 3 9 10 11 12 13 14 C8 100nF R2 . kΩ I C2 d sP I C33EP 256 GP 502-I / S O MCLR A VDD A N 0/ O A 2O UT / RA 0 A VSS A N 1/ C2I N 1+ / RA 1 RP I 4 7 / T 5CK / RB 15 P GED3/ VREF– / A N 2/ C2I N 1-/ S S 1/ RP I 32/ CT ED2/ RB 0 RP I 4 6 / T 3CK / RB 14 P GEC3/ VREF+/ A N 3/ O A 1O UT / RP I 33/ CT ED1/ RB 1 RP I 4 5/ CT P LS / RB 13 P GEC1/ A N 4 / C1I N 1+ / RP I 34 / RB 2 RP I 4 4 / RB 12 P GED1/ A N 5/ C1I N 1-/ RP 35/ RB 3 T DI / RP 4 3/ RB 11 VSS T DO / RP 4 2/ RB 10 O S C1/ CLK I / RA 2 VCAP O S C2/ CLK O / RA 3 VSS MS / A S DA 1/ S DI 1/ RP 4 1/ RB 9 RP 36 / RB 4 CVREF2O/ RP 20/ T 1CK / RA 4 VDD T CK / CVREF1O/ A S CL1/ S DO 1/ RP 4 0/ T 4 CK / RB 8 S CK 1/ RP 39 / I N T 0/ RB 7 P GED2/ A S DA 2/ RP 37 / RB 5 P GEC2/ A S CL2/ RP 38 / RB 6 MI C+ 2 C10 3. 6 p F 28 VCC LED C11 10µ F X 5R 10V C7 100nF MI C– 4 27 26 4 25 3 24 9 23 5 22 6 21 7 20 C6 10µ F X 5R 10V 19 18 CS S DO D/ C S DI 10 12 13 17 11 16 14 S CK T -CLK T _ DI N T _ DO T _ CS T _ I RQ 15 2 C9 100nF RES ET GN D LCD p anel S 2 S 1 This is a very simple change in the software. Just two lines of code change, as shown below: #define SAMPLE_RATE 2000 Changes to: #define SAMPLE_RATE 40000 And for the ADC peripheral: PR1 = 32200; Changes to: PR1 = 1610; Signal input Now let’s look at the hardware changes required to support this higher input frequency. The input circuit from the microphone to the ADC of the processor includes an amplifier, which is buried within the microcontroller itself, and so Practical Electronics | May | 2020 the detail is not obvious from the schematic in Fig.2. The key point is that the amplifier includes two external components, a resistor and a capacitor (R4 and C10 on the schematic) which will provide some low-pass filtering as well as setting the gain of the amplifier. Filtering out higher frequency signals is desirable as these can appear as phantom signals at lower frequencies, through a process called aliasing. It’s complicated, but the key point is to filter out unwanted higher frequency signals, otherwise they will interfere. We will build the circuit up slowly so we can first test the microphone sensitivity, and to get a good understanding of the amount of gain we will need to set in the amplifier. We are going to do this empirically (ie, by guessing component values and measuring the results) rather Fig.2. Updated circuit schematic. than through simulation, as we have no information on the microphone to model its characteristics. We start by assembling the voltage regulator circuit based around IC3, R6 (using a zero-ohm link or very low value resistor), C14 and C15. Then we add the microphone circuit consisting of R5, set to a value of 10kΩ. C5 was not connected, as we want to test the microphone circuit and op amp circuits independently. To help with these measurements we purchased an AFG (arbitrary function generator), the ISO-TECH AFG-21012 on eBay. At £150 it’s a serious investment, but very useful for projects like these where you would like to see the true performance of a circuit when presented with AC signals. An arbitrary function generator has many other uses too, so it’s a welcome addition to the lab. You can see our setup in Fig.3. 49 Fig.3. AFG – our brand new Arbitrary Function Generator. We connected the AFG to a small loudspeaker and generated a 1kHz tone, placing the loudspeaker next to the microphone. Powering up the circuit – remember, the microphone requires power – we discovered that the microphone has an output that varies in amplitude from 5mV to 150mVpk-pk (peak to peak) when the sound varies from quiet to loud. We now know the amplification that we want to achieve; from 5mV – 150mVpk-pk up to 3Vpk-pk, a gain of between 20 and 600. It’s time to wire up the op amp and experiment with setting the gain. To do that, we must wire up the processor sub-system and load the original application, as the op amp must be configured by software. So, we assemble the remaining circuit. The microphone amplifier circuit based around R4, C10, and the op amp built into the microcontroller can be viewed as the circuit shown in Fig.4. This is (loosely) an inverting amplifier, gain controlled by R4, with a low-pass filter determined by R4 and C10. The gain is the ratio of R4/Rin; we didn’t include a resistor in the input circuit, relying on the impedance of the microphone. This drove a bit of trial and error to find values for R4, and we settled on 2.2MΩ, a gain of approximately 220. That’s quite a large value resistor, but it made the device nicely sensitive to ambient level sounds. You can go higher, but the device quickly becomes too sensitive and can be easily overloaded, clipping the audio signal and generating misleading harmonics on the display. 2.2MΩ works really well. We implemented some low-pass filtering by adding C10 with a value of 3.6pF, giving a cut-off frequency at approximately 20kHz. This will not be a very effective low-pass filter, but it will remove some of the higher frequency signals (such as oscillator and SPI bus noise from the board itself) being aliased into the displayed frequency range. Improving this circuit with a switchable op amp filter would have improved the quality of the displayed signal, but that can be considered a future improvement. Why is attenuation for higher frequencies important? Having higher frequency signals present at the input to the ADC will result in aliasing – those higher frequency signals appearing to be lower in frequency and will end up being displayed on the LCD. You can see this effect in Fig.5. Filtering these out or minimising them greatly before they reach the ADC will stop them generating ‘fake’ signals on your display. It was interesting playing with the sound generated by the AFG, especially as the frequency is raised out of hearing range, but with the spectrum analyser still faithfully displaying the signal. For the author, a quiet sound level stopped being audible at around 8kHz, but the signal was of course still visible on the display. Raising the frequency above 12kHz resulted in the displayed signal dropping to zero, probably as a result of the loudspeaker and microphone limitations, as these are not high-fidelity devices. Increasing the signal level output from the AFG saw the signal reappear, right up to 20kHz. We’ve released a YouTube video demonstrating this effect, which can be viewed here: https://youtu.be/mGIGxUU9w1k Software First, we replaced the old EPE logo with the new PE one. We wrote a simple Python script to convert the magazine’s logo to a data file suitable for including in the program. The source code for this is provided in the project download file, should you be interested in changing any of the graphics. Next, we moved onto the more challenging software task – adding a user interface to the device, enabling the buttons to change the behaviour of the device. Unused I nsid e th e d sP I C p rocessor Fig.4. Microphone pre-amplifier circuit. T h is lower f req uency signal would be d isp layed in error Fig.5. Aliasing – we need to filter out high frequencies that can give erroneous data. S p ectrum d isp lay 200 O utp ut + 50 Logo 24 0 – T itle 8 5 R4 RI N 26 0 H igh er f req uencies can generate d ata p oints th at mimic lower f req uencies C10 I np ut In the previous article we gave the following desired functionality for the buttons: Button A – On each press of the button, cycle between peak frequency display, peak trace hold, none. When active, the peak frequency will be displayed in the format nn.n Hz directly above the peak signal. Button B – On pressing once, the display will show the frequency span for three seconds. On pressing the button again while the span is being displayed, the device will cycle through different span settings. Those are good and clear requirements, but now we come to implement these requirements, looking at the display layout (as shown in Fig.6) we realised that there is a lot of available screen space unused in the lower left of the display. The width of this spare screen space is 60 pixels, enough to display ten characters in a 5×7 font. There is space for eighteen rows of characters, so there is plenty of room to display useful information continuously, irrespective of button presses. Doing so will also simplify our software design, so we chose to take this route. The original Adafruit LCD driver software provided a simple 5×7 pixel character font, so we made use of that. We imported two functions from the original display driver, ‘draw pixel’ and ‘draw character’. It took an hour to adjust the Adafruit code to suit the Microchip development environment, at which point we were successful at displaying text – but on its side. The Adafruit driver software expects the display to be driven in portrait mode, not landscape as we are using. No problem, back to hacking code. This driver is going to become somewhat specialised to our application. The next step was to construct a ‘draw string’ function, to display the ‘Peak’ and ‘Span’ text, and the values calculated on each display update. The draw string routine was easy to implement, and then converting the calculated values into displayable strings was done using a standard ‘C’ function, sprintf(). No work required there. Placing this text on the display revealed an issue – for the author at least – the text 320 Fig.6. Mapping out our LCD real estate. Practical Electronics | May | 2020 Handling button presses is often a tricky part of a project’s P O W ER T I MER1 A DC software design; there is no uniO N IN T IN T versal ‘one solution fits all’, and custom designs invariably involve interrupts and timers, Conf igure S tart an A DC S tore A DC LCD conve rsion samp le which can be difficult to set up and debug. Let’s start our design by Conf igure B utton remembering what our requireA DC and buttons ch ange IN T ments are, and understanding what capabilities we have availEnable able on the processor. When interrup ts H O LD S P A N W h ich a button’s state changes from button? ‘not pressed’ to ‘pressed’, deS tart S tart bounce the press, then perform W ait f or 512 T imer2 T imer3 A DC samp les the action. Here we are looking for interT I MER2 YES IN T rupt capability on the pins that H old ? we are using as button input, PORTB pin 5, and PORT B pin T oggle Comp uter 6. Fortunately, the processor H O LD FFT provides a ‘change notification’ interrupt for all GPIO pins, so we have the ability to detect Red raw T I MER3 LCD IN T button press and button release. Once a button press is detected we need to wait for a Disp lay p eak Cycle th rough f req uency short time, typically 30-50ms, S P A N settings and re-check the input to see if the button is still pressed. If it is, perform whatever action is Fig.7. The final application flowchart. The parts in the required. We will use a timer dash line boxes are the additions for handling the to provide the delay, to avoid button presses. tying the processor up during those 30-50ms. When the timer expires, was too small to read without glasses. it will jump to an interrupt routine that Fortunately, the Adafruit graphics library will perform the desired action, if it finds for printing characters has a ‘double size’ that the button is still pressed. To simplify option, scaling the font to 10×14 pixels. the software, we will use two timers, one This limited us to five characters across for each button. The processor has five the display, but the text is very clear. We timers internally, so it’s not a waste of can accommodate a width of five characlimited resources. You can see in the apters in the design by cunning selection of plication flowchart in Fig.7 that the flow display formatting. of execution is very simple. The code for Next, how to handle the buttons. As this is actually quite simple; the complex the program is a simple continuous loop, part was finding out how to configure shown in Fig.7, we could have detected the the timer, and what names are used by button press within the loop. Two issues the compiler for the interrupt routines. cause us to avoid this: at the lower ‘span’ Thankfully, these details are very, very frequency settings, the processor will similar between the different families of spend a long time taking ADC samples so microcontrollers, so this task took just a we cannot rely on the amount of time the couple of hours. loop takes; button press responses will be We took the opportunity to add severdelayed. We also need to de-bounce the al different sample rates, with the SPAN button press actions, which will require button cycling between them. Providing a checking the button level 30 to 50ms after range of sample rates yields a significant it has been pressed – this will slow down benefit to the design. The FFT algorithm the display update. So, we take the more is fixed at 512 ‘bins’, providing 256 data complicated route, and use an interrupt points. At a 40kHz sample rate, which routine to detect when the buttons have equates to a maximum displayed frebeen pressed or released, and a timer to quency of 20kHz, the resolution of the trigger an action 30 to 50ms later to see if display is 78Hz. This means that two the button is still in the same state. This signals close in frequency to each other, way we can react very quickly to key presssay a 50Hz and a 60Hz signal, would be es and be more efficient with the use of lumped into the same vertical line. Prothe processor, allowing it to concentrate viding a range of spans down to 500Hz on reading ADC values, computing the means we can come down to a resolution FFT, and displaying it. Practical Electronics | May | 2020 of 2Hz – which means those two signals will be clearly visible across the display. The only impact of doing this (there is always a cost!) is what is referred to as temporal resolution – at 500Hz the signal acquisition takes a long time, 256ms. It would be easy to reduce the peak frequency even lower to get a more accurate reading on lower frequency signals, but at each halving of the peak frequency, the sample time doubles. For some applications this is perfectly acceptable, and it is a simple change in the software if you would like to test it. We also decided to display the word ‘HOLD’ when the HOLD button has been pressed, just for fun. It only added three lines of code. With these changes completed, the software took just 23% of available code space on the processor, leaving over 64KB for future additions. You can see the finished project in Fig.8. After testing it for a few minutes and then correcting two small software bugs, we removed the batteries and connected a bench power supply. We observed a continuous current draw of 120mA at 4V, and that the circuit continued to function down to 2.9V (at room temperature, at least.) Based on this, we decided to use just three AA batteries to power the project, which should provide 25 hours continuous use. 3D printed enclosure We’ve been using a 3D printer in the PIC n’ Mix lab for ten years now, since we first introduced the technology in February 2010. Then, 3D printers were an expensive novelty. Now, they have reduced in price from £1200 machines to £160 (the cost of the author’s current machine, a Creality Ender 3 – a printer we cannot praise enough.) By now, as hobbyists, we either own a 3D printer or know people who do and who are willing to print objects for us. From now on, expect to see 3D-printable enclosure designs for our projects, where appropriate. We will leave the detail of designing a 3D-printed enclosure to another series of Fig.8. The finished project, in its 3D-printed presentation case. 51 provide mounting points to the enclosure. We’ve seen both employed on complex products, but for this device the design is simple enough – four mounting posts, two buttons and a hole for the LCD – that we have decided to try the former approach. This reduces the number of 3D printed components required from three down to two, so it will be quicker and cheaper to build – on a hobbyist printer this will be very cheap, under a pound’s worth of plastic filament. For this project, we took a different approach, choosing to make a holder for the assembly, leaving the components exposed and clearly visible. We took this approach because we wanted to display the project at a Maker Faire, and it’s more interesting to be able to see what’s going on. You can see our design in Fig.9, created with the free design tool RS Design Spark Mechanical. Fig.9. Enclosure design (fft-case.stl file available for download from the May 2020 page of the PE website). articles, but the key point is that today, our recommendation is to use RS Components Design Spark Mechanical software for creating a design. The main reason is that this is an excellent application, easy to use and with many tutorials available online. There are better design applications available, both free and paid for (FreeCAD and Fusion 360 respectively) but for starting out in 3D CAD design with no previous experience, Design Spark Mechanical has, in our opinion, the best support for beginners. We’ve been using it for a year and are still happy with the capabilities. There are two common techniques employed when designing enclosures for electronic products; design a two-part enclosure that provides mounting points on one part for the electronics, or design a ‘chassis’ for the electronics, and have that chassis Final thoughts While this Audio Spectrum Analyser is not a particularly ‘useful’ design, it’s certainly been very educational, great fun to play with, and generated a lot of interest at our local Maker Faire, as the LCD danced to music and voices. With the inclusion of software libraries for graphics and text display, it offers building blocks for other projects, even unrelated to audio. We are already planning a bedside alarm clock for the version 2 hardware! The software, PCB and 3D printer enclosure files are all available on the magazine website. The PCB can be ordered from the PE PCB Service or directly from PCBWay, at this webpage: http://bit.ly/pe-may20-pnm This project has spanned two years, and we hope you have enjoyed the journey and the results. If you take this design and create something new, do tell us! Need a programmed PIC for your Practical Electronics project? Coming soon from the Practical Electronics online shop at: www.electronpublishing.com 52 Practical Electronics | May | 2020