Fullscreen HTML5Med Res (??MB)HTML4

Digital Sound Effects Module This inexpensive, compact module can play back up to eight different sound effects, lasting a total of 60 seconds or more. It’s powered from a lithium cell or from a lowvoltage AC or DC supply and can be used with model railway layouts or any other application requiring sound effects. T HIS LITTLE MODULE is quite simple but we’ve packed a host of features into it. You can upload a variety of sounds from a computer via its USB port and it will then play back the sounds when triggered. It’s small enough to be hidden away inside a vehicle, model or wherever and it can be triggered by a microswitch, reed switch, pushbutton, sound or light detector, etc. The most obvious use is to hide it inside a model car or train, to produce an engine sound and a horn or whistle effect. Or you could build it into a door to play back a sound each time it’s opened. You could even hook it up to a pet door so that it plays a sound to let you know when your pet enters or leaves the house. Alternatively, you could fit it with a pushbutton for sound effects while playing a game or have it triggered whenever equipment is used or the fridge door is opened. In fact, the possibilities are endless. In operation, the unit drives an 8-ohm speaker and if the speaker is properly baffled and efficient enough, the playback volume level can be quite loud (more so with an AC/DC supply than a button cell). The playback time can be up to 60 seconds or more, depending on the sound quality used. The module has two inputs to trigger different sets of sound effects and 70  Silicon Chip each trigger can be assigned to any set of the eight possible sound effect slots. When triggered, it can either randomly pick one sound from the selected set or you can have it cycle through them in sequence. To keep the unit small and the cost low, it uses virtually all SMDs (surface mount devices). We’ve chosen the easiest SMDs to solder so that just about anyone can build it, given some patience. The circuit is based around two ICs, a PIC microcontroller and an LM4819 low-power audio amplifier. Up to 108KB of the PIC’s internal flash memory can be used for sound storage but if that isn’t enough, it can be expanded to over 1MB (more on this later). PWM sound generation We initially considered using a PIC microcontroller with an inbuilt DAC (digital-to-analog converter) for sound playback. Unfortunately, few PICs contain an audio DAC and those that do require a regulated supply of 2.73.3V. This isn’t really suitable for use with a lithium cell as they can drop below 2.7V under load or if a bit flat. Rather than add the complexity of a boost regulator to maintain the voltage, we decided to use a standard PIC with two high-speed PWM outputs. These are used to drive low-pass filters, so that we effectively build our own By NICHOLAS VINEN simple DAC. In practice, this works quite well and gives performance comparable to a dedicated 10-bit or 12-bit DAC, with quite an acceptable level of distortion – typically less than 0.2%. Block diagram Fig.1 shows the general arrangement. IC1 produces two PWM waveforms, each with a duty cycle variable from 0-100% in 64 steps (26). The output from pin 7 (RP2/ PWM0) is determined by the six most significant bits of the 12-bit sampled waveform being played back, while pin 2 (RP0/PWM1) has a duty cycle based on the six least significant bits. This second output is used to provide smaller output voltage steps for better resolution. These two square waves each pass through low-pass RC filters, to remove most of the high-frequency harmonics and produce voltages which are proportional to the input duty cycles. The 34kHz -3dB roll-off point ensures that there is little attenuation of audible frequencies. After filtering, the signals are mixed with a ratio of 64:1, to reconstruct the 12-bit digitally-sampled voltage level. Refer to the panel later in this article (Using PWM To Reproduce PCM Audio) for a detailed explanation of how the two 6-bit PWM outputs are combined to give the equivalent of a 12-bit output. We chose six bits per output for siliconchip.com.au 187.5kHz 6-BIT PWM (6 MOST SIGNIFICANT BITS) MICRO CONTROLLER IC1 RP2/PWM0 7 LPF (34kHz) x1 MIXER LPF (34kHz) x1/64 RP0/PWM1 2 (LONGER TIMEBASE) 4 3 Vdd LPF (34kHz) 5 20k 20k 8 8 SPEAKER AUDIO AMPLIFIER (IC2) 187.5kHz 6-BIT PWM (6 LEAST SIGNIFICANT BITS) 7 DUAL 6-BIT PWM DAC (~11 BIT EQUIVALENT) Fig.1: block diagram of the Sound Effects Generator Module showing how the PIC micro reproduces the audio. IC1 generates two PWM square waves based on stored audio data and these signals fed through low-pass filters before being mixed with a 64:1 ratio. The output of the mixer is filtered further and then passed to IC2, a low-power audio amplifier which drives the 8Ω speaker in bridge mode. two reasons: (1) a total of 12 bits gives a good compromise between the memory required to store an audio file and the resulting playback quality; and (2) this allows us to have a PWM frequency well above the -3dB point of the required low-pass filters, so that the latter are reasonably effective. The output from the mixer passes through another low-pass RC filter to further remove switching noise and is then fed to the non-inverting input of audio amplifier IC2. As shown, this stage drives the speaker in bridge mode. This not only maximises the audio output power (important given the low supply voltage of ~3V) but also avoids the need for a large DC-blocking output capacitor. IC2 operates with a gain of +1 for the non-inverting output and a gain of -1 for the inverting output, giving an overall gain of 2. It’s able to deliver about 100mW to the speaker, which produces quite a reasonable volume if the speaker is efficient. In practice, the available power is limited by the lithium cell. Fig.2(a) shows a scope grab of the audio output when reproducing a sinewave. It’s zoomed in far enough to show the remnants of the highfrequency PWM signal but you can also see the curved sinewave shape. When we change the scope’s time base to “zoom out”, we see from Fig.2(b) siliconchip.com.au Table 1: Playback Time vs Sample Rate & Bit Depth Sampling Rate & Bit Depth No Flash Chips One Flash Chip Two Flash Chips 8kHz, 8-bit 14s 80s 125s 11.025kHz, 8-bit 10s 58s 105s 8kHz, 12-bit 9.5s 53s 97s 11.025kHz, 12-bit 7s 38s 70s 22.05kHz, 12-bit 3.5s 19s 35s 32kHz, 12-bit 2.5s 13s 24s 44.1kHz, 12-bit 1.5s 9.5s 17.5s 48kHz, 12-bit 1.5s 8.5s 16s that the waveform is quite smooth (ignoring supersonic frequencies). Interpolation While the PWM outputs operate at around 187.5kHz, the audio sampling rate is a lot lower. If we simply changed the PWM duty cycles at the sampling rate of the audio file being replayed (eg, 11,025Hz), the output would have visible steps as shown in Figs.2(c) & 2(d). This would result in extra harmonic content in the audio output which would sound quite bad, especially at lower sampling rates due to the larger effective step size. In fact, the audio produced using this technique sounds rather “crackly” – not good! The simplest solution is linear interpolation. This involves changing the PWM cycle a little for each pulse, for the same total change over time but in smaller increments. In fact, Figs.2(a) & 2(b) show the identical waveform to Figs.2(c) & 2(d) but the former have the linear interpolation enabled. As you can see, the resulting waveform is much smoother and it sounds a lot better too. This interpolation requires a lot more processing in the PIC. Each time a new sample value is loaded, it must calculate the required slope and given the low PWM resolution (six bits), this is often going to be a fractional value so we need to do some fractional maths to generate a smooth ramp. The PIC18F27J53 is (just) powerful enough to do this with some carefully written code. With a 187.5kHz PWM September 2012  71 Fig.2(a): a close up of the audio output from the module (output of IC2), showing the residual PWM signal that isn’t filtered out, plus the smoothly varying level of the sinewave which is being played back. Fig.2(b): the same sinewave signal as Fig.2(a) but with a longer timebase. The low-pass filtering of the scope’s input circuitry has rendered the switching residuals invisible, leaving just the smoothly varying output. Fig.2(c): the same sinewave (11.025kHz sampling rate) being played back without the linear interpolation code active. The resulting steps cause audible artefacts, especially with lower sampling rates. Fig.2(d): another view of the non-interpolated sinewave with a longer time base, clearly showing the steps which result from the limited time resolution available at low sampling rates. update rate and a maximum instruction clock rate of 12MHz, we have just 12M/187.5k = 64 instructions to perform these calculations. In the end, we were able to make the code fast enough, using an 8-bit fractional sample-position counter and a handoptimised 8 x 12-bit multiply/scale function to integrate the computed delta (ramp) value over time. Circuit description Now take a look at Fig.3, the complete circuit diagram. The three low-pass filters and mixer shown in Fig.1 are implemented using three resistors (two 10kΩ & one 620kΩ) and two 470pF capacitors. This is possible 72  Silicon Chip because the two first-stage low-pass filters and the mixer are combined. You can essentially think of it as two low-pass RC filters with a common capacitor. In addition, the different resistor values effectively form an attenuator between the two PWM outputs, to give the correct (approximate) mixing ratio. The relatively small capacitor value (470pF) was chosen to minimise distortion due to loading on the microcontroller outputs, which have limited current capability. The second low-pass filter is similar to the first and is connected between the mixing node (ie, the junction of the 10kΩ and 620kΩ resistors) and the non-inverting input (pin 3) of amplifier IC2. In this configuration, IC2 only needs two additional components to operate: a 1µF supply bypass capacitor and a 10nF capacitor to filter its internal half-supply voltage generator. This latter capacitor also determines how long it takes to go into and out of sleep mode, which is used to minimise power consumption when no sound is being played. We want to play back sounds immediately when triggered, so the 10nF capacitor gives a turn-on time of just 10ms. Audio amplifier IC2 drives the speaker in bridge mode via CON4. The circuit is DC-coupled so IC1 is programmed to deliver an average siliconchip.com.au CON3 10 + D1 BAT54C 47 F 25V A2 POWER IN 5 – 24V 1 F Vdd OUT IN K – Q1 DMP2215 REG1 LM2936MP-3.3 A1 100 F 16V GND PWDET 20 2 15 3 16 PWDET CON2 10k CS1 CS2 SCK 1k* SDO ICSP/TRIGGER 1 GND TRIG2 TRIG1 2 3 Vdd 4 5 SDI 1k* CON1 14 6 1 F D–/RC4 D+/RC5 1 F 620k 6 1 10k 3 4 VUSB 2 Vss1 8 14 1 8 4 1 SDO SDI SCK CS1 VO1 BYPASS 8 SPEAKER 5 CON4 GND 7 Vss2 19 Vdd 5 2 6 1 8 Vcc SDI SDO SCK CS WP IC3 AT25DF 041A GND 4 3 5 SDO 2 SDI HOLD 6 SCK 7 1 CS2 DIGITAL SOUND EFFECTS MODULE SDI WP IC4 AT25DF 041A SDO SCK CS GND 4 HOLD 3 7 100nF (OPTIONAL) A2 K A LM2936MP DMP2215L BZX84B5V1 K A1 8 Vcc 100nF BAT54C 2012 IC2 LM4819 –IN (OPTIONAL) SC  VO2 +IN 10nF Vdd IC2, IC3, IC4 Vdd SHUT DOWN 470pF 470pF VddCore 10 F IC1 ZD1 BZX84B5V1 A 1 F Vdd 17 3 RC6 RA1 24 2 RB3 RP0/PWM1 23 7 RB2 RP2/PWM0 25 SCK1/RB4 IC1 1 MCLR PIC18F27J53 10k 26 SDI1/RB5 18 SDO1/RC7 28 PGD/RB7 27 PGC/RB6 * SHORT OUT FOR PROGRAMMING 28 620k Vdd A2 4 K 10k K 1 CR2032 BATTERY G D2 BAT54C A1 USB TYPE B D S D G S TAB (GND) IN GND OUT Fig.3: complete circuit of the Sound Effects Generator. IC1 generates the PWM waveforms which are filtered and then passed to audio amplifier IC2. IC3 and IC4 are optional flash memory chips for more storage space and these are controlled using a 5-wire serial bus. REG1 provides a regulated 3.3V rail when the unit is plugged into a USB port or is running from an external supply; the rest of the time, it runs off a CR2032 lithium cell. Sounds are triggered by pulling pins 4 or 5 of CON1 low and CON1 can also be used to program IC1 with an in-circuit serial programmer. modulated output of 50% to prevent a large DC voltage from appearing across the speaker. More memory The firmware occupies 20KB of IC1’s 128KB internal flash memory, leaving 108KB available for sound storage. This will be sufficient for some applications but if you want multiple sound effects or longer sounds, you will need more space than this. siliconchip.com.au In practice, the total flash memory can be expanded to either 620KB or 1132KB by adding one or two low-cost serial flash chips – IC3 and IC4. These each store 4Mbit (512KB) of data. IC1 automatically detects whether either or both chips are installed at power-up. Table 1 shows the total playback time available with various combinations of IC3 and IC4 installed. IC1 communicates with the flash chips using a 3-wire SPI (serial peripheral interface) bus plus two chip-select lines – CS1 and CS2. The specified flash chips (AT25DF041A-SSHF) were chosen for their wide operating voltage range (2.33.6V) and low power consumption. IC1’s minimum operating voltage is 2.15V but in practice, we expect that all the ICs will run down to about 2V. The supply voltage for IC3 and IC4 is critical during erase and write operations, during which time they September 2012  73 CON2 Q1 620k 10k D2 ZD109109121 1 F D1 SFX POWER 1 (MINI SPEAKER) 10nF IC2 TRIG1 TRIG2 GND CON1 ICSP 620k 470pF POWER IC3 470pF 1 1 F 100 F 100nF 1 IC4 ICSP 10k 10k 5 10 10 F 1 10k BAT1 47 F 1k 100nF 1 (BUTTON CELL HOLDER) REG1 1 1k 1 F IC1 PIC18F27J53 1 F SPKR TOP OF PC BOARD SPKR UNDERSIDE OF PC BOARD Fig.4: the SMD parts all mount on the top side of the PCB while the through-hole parts, including the cell holder, are mounted on the bottom. CON1 is a friction-fit for programming but can be soldered in to connect the trigger inputs if you don’t want to solder wires direct to the PCB. Note that there is room for a small speaker to be taped to the bottom of the PCB but an off-board baffled speaker will give better results. run from a regulated 3.3V rail derived from an external PC’s USB port, via D1 and REG1. Sleep mode When the module is not plugged into a USB port and not playing any sounds, IC1 goes into sleep mode to save power and the whole circuit typically draws less than 10µA from the CR2032 cell. If IC3 or IC4 are installed, they are placed in “Deep Power-down” mode which, according to the data sheet, gives them a typical current consumption of 15µA each. You would expect then that installing IC3 and/or IC4 would reduce the standby cell life substantially. However, we measured the actual sleep current for IC3 and IC4 at about 2µA each. This likely reflects manufacturing process improvements since the AT25DF041A data sheet was written and we expect most constructors will find that installing these chips has little effect on cell life. During playback, IC3 and IC4’s operating current is negligible compared to that of IC1 and IC2, due to the low data rate (72kbits/s maximum). USB interface The PIC’s USB interface is used to transfer sound data for later playback. It’s also used to configure the various trigger options. The only external component required for the PIC to communicate via USB is the mini-B type connector (CON2). The necessary USB impedance-matching and pull74  Silicon Chip up resistors for the D+ and D- communication lines (pins 16 & 15) are inside IC1. In operation, the PIC monitors the USB VCC line, to determine when the unit is plugged in. This is necessary so that the internal USB module can be turned off at other times to save power. The method used will be explained shortly. Power supply When a CR2032 3V lithium cell is installed, it powers all the ICs via Mosfet Q1, which provides reverse polarity protection. Q1 is a P-channel type with its gate tied to ground via a 620kΩ resistor, so that it is switched on by default. However, if the cell is somehow inserted backwards, its gate will instead go positive compared to its source. In that case, Q1 switches off and its internal body diode is reverse biased, so no current can flow. Conversely, when it’s on, Q1 has a very low on-resistance (<0.2Ω). As a result, there is very little voltage drop across it, given the low current drain from the battery (<50mA). As stated, the circuit can also be powered via the USB port or from an external DC or AC supply. In these cases, the 3.3V supply for IC1-IC4 comes from REG1, an LM2936 low-dropout linear regulator. This is especially important for USB communications, as IC1 requires a supply rail that’s close to 3.3V for proper USB operation. When an external supply is present and the LM2936 is powering the ICs, its output voltage will typically be above the cell’s voltage (nominally 3V). As a result, we need to prevent it from charging the cell, which could damage it. This function is also performed by Q1. The external supply voltage pulls Q1’s gate high via dual Schottky diode D2 and a 10kΩ resistor. One half of this diode conducts if an external USB supply is connected, while the other half conducts if an external supply is fed in via CON3. As a result, Q1 is switched off and no current can flow into the cell (since Q1’s internal body diode is also reverse biased). Note that dual-diode D2 is necessary so that you can’t accidentally feed power from CON3 into the computer’s USB port (if connected). Zener diode ZD1 protects both Q1 and pin 17 of IC1 from damage should the external supply be above 5.5V. Pin 17 of IC1 is used to detect when external power is applied, to enable the USB transceiver (this pin is 5.5V-tolerant and so can be used for this task). The software sets this pin as an interrupt source, so it can wake the micro when the USB interface is connected. DC/AC supply The external supply can be either 5-24V DC or 5-24V p-p (peak-to-peak) AC and is fed in via CON3. This suits many applications, including a model railway system with DCC, which uses a 15-22V AC square wave. For AC, one half of dual-Schottky diode D1 rectifies the supply voltage while for DC, siliconchip.com.au this diode provides reverse polarity protection. A 47µF 25V electrolytic capacitor filters the resulting supply rail while a 10Ω series resistor limits the in-rush current when power is first applied. This prevents D1 from burning out when the unit is first powered up. As with USB power, REG1 then provides the 3.3V supply for the ICs. REG1 can pass up to 50mA, which gives an instantaneous dissipation of around 1W with a 24V input. That would be too high if it were sustained but in practice, power is drawn in bursts by the audio amplifier. This lowers the average dissipation to an acceptable level. Trigger inputs CON1 serves both as an ICSP (incircuit serial programming) header for IC1 and as the trigger input connector. For programming, the two 1kΩ series resistors must be shorted out. These resistors normally protect the IC inputs from accidentally applied voltages above 3.3V during operation (eg, you can use a 0-5V trigger signal if necessary). Normally, to trigger a sound, either TRIG1 or TRIG2 is pulled to ground although the unit can be re-configured to invert the trigger logic. Software IC1’s software must perform a number of tasks. As explained earlier, it goes into and out of sleep mode as necessary, powering up the USB interface and the serial flash chips only when needed. Pin-change interrupts on pins 17, 27 & 28 are used as wakeup signals. When the USB interface is enabled, the module appears as a virtual serial port. The XMODEM protocol is used to upload audio files (8-bit or 16-bit mono WAVs). Configuration commands are sent as text over the serial port and the module responds to indicate that they have taken effect. You can also query some information from the module, such as how much memory is free. When you upload a WAV sound file, it checks that the format is valid and that there is enough free memory, then stores it. If a 16-bit file is uploaded, it is converted to 12-bit format on-thefly, to save memory and speed up the playback code. There are a number of configuration options such as whether the sounds are siliconchip.com.au Features & Specifications Module size: 59 x 28 x 13mm Trigger inputs: 2 Number of sound effects: 1-8, triggered round-robin or random Audio sampling rate: 8-48kHz Audio resolution: 8-bit or 12-bit Sound memory: 108KB, 620KB or 1.12MB Total playback time: 1.5-125 seconds depending on sampling rate & data memory (see Table 1) Output power: Approx. 100mW into 8Ω Supply options: CR2032 lithium cell, USB 5V, DC 5-24V, AC 5-24V peak-to-peak Cell operating voltage: 2.15-3.3V (2.3-3.3V with memory >108KB) Standby current: typically 9-14µA, depending on installed memory Standby cell life: >1 year Playback cell life: 4-24 hours, depending on sound volume, etc Configuration interface: USB (mini type B socket) USB protocol: virtual serial port (CDC), file transfer via XMODEM Computer operating system: Windows XP, Vista, Windows 7* * In theory, the module will work with Linux and Mac OSX using the CDC driver but we haven’t tested it. The driver will need to recognise our Vendor ID and Product ID (04D8, FD52). looped, whether the sound continues playing to the end of the file once the trigger input is released, which input has priority, how to deal with multiple sounds and so on. These are set using text commands over the USB serial interface and stored in IC1’s flash memory to be used when the unit is triggered (more on this later). Construction The Digital Sound Effects Module is built on a double-sided PCB coded 09109121 and measuring 28 x 59mm. Fig.4 shows the parts layout. The first job is to fit the surface-mount devices to the top side of the PCB. Start by laying the board flat on your workbench and fitting the USB connector (CON2). This has two plastic locating posts on the underside which go into matching holes on the PCB. Ensure that the socket end is at the edge of the PCB and that the connector is sitting flat, then solder one of the mounting feet. That done, check that the five pins are properly aligned on their pads, then solder the other three feet plus the five pins. You will have to angle the soldering iron when soldering the pins, as they are under the main body of the connector. Don’t worry about solder bridges at this stage; just make sure they are soldered correctly. It’s then simply a matter of using solder wick to clean up the bridges (note: adding a bit of flux paste makes this much easier). Finally, check that the bridges are gone using a magnifier; if not, add more flux and fix them. The four SOT-23 (small-outline transistor package) devices can be installed next (ie, D1, D2, Q1 & ZD1). Be sure to remove them from their packaging one at a time so you can’t get them mixed up (they look virtually identical). In each case, it’s just a matter of placing a small amount of solder on one of the pads, then reheating the solder while you slide the device into place. If it isn’t aligned properly, simply reheat the solder and nudge it until it is correctly aligned. The other pins can then the soldered. Follow with the ICs, taking care to get the orientation correct. In each case, pin 1 is indicated with a dot on the PCB. IC1 and IC2 should have a divot near pin 1 while the other two ICs (if fitted) have a bevelled edge on the same side as pin 1. As before, it’s just a matter of applying some solder to one of the end pads, then reheating this solder as the device September 2012  75 Serial Commands For The Sound Effects Module Commands are sent to the Digital Sound Effects Module by typing them into the serial terminal. There are three basic types of command: those which give you information, those which are used to upload sound files and those which are used to change the module’s configuration. Most commands have an immediate effect and respond with information after you press the enter key. If there is an error (eg, you mistyped the name of the command), it will respond with information about what has gone wrong. Having prepared the sound files, the next step is to use the Send command to upload them. If you have a speaker wired up at this stage, you can then check that everything is working using the Play command. Here is the list of available commands with some information about how to use them. Send Ready for file via XMODEM Abort Transfer aborted Command: “Info” Description: once you have uploaded a sound file, you can set some options which determine how it is played back. By default, when triggered, the sound will play once and won’t stop until the end of the file (unless interrupted, see below). If you want it to loop as long as the trigger input is held on, use the “loop” option (or “once” if you don’t; this is the default). If you want the sound to stop as soon as the trigger input is released, rather than wait for playback to complete, use the “partial” option (the default is “complete”). Example: Options 1 loop, partial Sound #1: 22050Hz, 12-bit, 12.8s, 415KB, loop, partial Description: displays the firmware version, amount of memory installed and free, what sound files are loaded and the configuration settings. Example: Info SILICON CHIP Sound Effects Module v1.0 Total memory: 1131.9KB Free memory: 721KB Sound #1: 22050Hz, 12-bit, 12.8s, 411KB, loop, stop immediately Trigger #1: NO, sound #1, priority, random Trigger #2: NO, no sounds, round robin Unsaved configuration changes Command: “Clear all” Description: deletes all sounds loaded into the Module, freeing up all memory for new sounds Example: Clear all Memory cleared, 1131.9KB free Command: “Clear last” Description: deletes the last sound loaded into the Module, freeing up the memory it occupied. Example: Clear last Sound #2 cleared, 721KB free Command: “Send” Description: initiates the upload of a sound file to the module. After a successful Send command, the sound is uploaded via XMODEM. The sound file is given the next available index, starting with #1. Example: Send Ready for file via XMODEM Saved to index #1 Command: “Abort” Description: cancels a pending Send command. Can be used if the transfer failed for some reason but the unit is still waiting for it to finish. You can also re-start a transfer by doing a Send command again. Example: 76  Silicon Chip Command: “Play <index>” Description: immediately plays back the sound loaded in the specified location. The USB interface does not respond during playback. A response will be sent once playback is complete and the serial port interface is then ready for more commands. Example: Play 1 Playing file #1 (12.8s)... Playback complete Command: “Options <sound index> <options>, <option> ...” Command: “Sounds <trigger index> <sound index>, <sound index> ...” Description: allocates one or more sounds to a trigger index (1 or 2). Sounds can be allocated to either or both trigger inputs. This determines which sounds are played back when the specified trigger input is activated (one at a time, see below for information on how they are chosen). Example: Sounds 1 1 Trigger #1: NO, sound #1, priority, random Command: “Trigger <trigger index> <option>, <option> ...” Description: sets the options for trigger 1 or 2. The available options are “NO” or “NC” to set the input mode to suit normally open or normally closed switches (or active low and active high signals, respectively), “priority” (which allows it to interrupt sounds which are triggered by the other input) or “nopriority”, “roundrobin” (with multiple sounds allocated, they are played in sequence) or “random” (with multiple sounds, one is randomly selected each time). Example: Trigger 1 priority, random Trigger #1: NO, sound #1, priority, random Command: “Save” Description: configuration commands (except for Send) are not permanently saved until this command is executed. If you don’t save configuration changes, they will be lost when the unit loses power. Example: Save Configuration saved siliconchip.com.au Digital Sound Effects Module: Parts List 1 PCB, code 09109121, 28 x 59mm 1 PCB-mount button cell holder (Jaycar PH9238) 1 CR2032 lithium cell 1 5-pin header, 2.54mm pitch (CON1) 1 SMD USB connector, mini-B type (CON2) (eg, Altronics P1308) 1 8Ω mini-speaker (eg, 27mm or 40mm diameter) 1 100mm length 2-wire ribbon cable 1 USB cable, type-A plug to miniB plug 4 M3 x 9mm tapped Nylon spacers 4 M3 x 6mm machine screws Semiconductors 1 PIC18F27J53-I/SO microcontroller programmed with 0910912A.hex (IC1) 1 LM4819 audio amplifier [SOIC8] (IC2) (Digi-Key LM4819MXCT-ND) OR 1 LM4889MA 1W audio amplifier [SOIC-8] (IC2) (Element14 1286916) 1 LM2936MP-3.3 50mA 3.3V LDO regulator [SOT-223] (REG1) (Element14 1469062) 1 DMP2215L P-channel Mosfet [SOT-23] (Q1) (Element14 1713864) 1 5.1V zener diode [SOT-23] (ZD1) (Element14 1431236) 2 BAT54C dual Schottky diodes [SOT-23] (D1, D2) (Element14 1467518) Capacitors (SMD 3216/1206 unless specified) 1 100µF PCB-mount low-profile electrolytic (eg, Element14 9452567) 1 47µF 25V PCB-mount low-profile electrolytic (eg, Element14 1165523) 1 10µF 4 1µF 1 10nF 2 470pF BitScope Digital + Analog w Ne del o M Pocket A nalyzer Everything in one tiny 2.5" package ! 100 MHz Digital Oscilloscope Dual Channel Digital Storage Oscilloscope with up to 12 bit analog sample resolution and high speed real-time waveform display. 40 MSPS x 8 Channel Logic Analyzer Captures eight logic/timing signals together with sophisticated cross-triggers for precise multi-channel mixed signal measurements. Serial Logic and Protocol Analyzer Resistors (SMD 3216/1206, 0.25W 1%) 2 620kΩ 2 1kΩ 4 10kΩ 1 10Ω Optional parts for longer playback time 2 AT25DF041A-SSHF-B 4Mbit flash memory ICs [SOIC-8] (IC3, IC4) (Element14 1636622) 2 100nF ceramic chip capacitors [3216/1206] Capture and analyze SPI, CAN, I2C, UART & logic timing concurrently with analog. Solve complex system control problems with ease. Real-Time Spectrum Analyzer Display analog waveforms and their spectra simultaneously in real-time. Baseband or RF signals with variable bandwidth control. Waveform and Logic Generators Generate an arbitrary waveform and capture analog & digital signals concurently or create programmable logic and/or protocol patterns. Multi-Channel Chart Recorder Record to disk anything BitScope can capture. Allows off-line replay and waveform analysis. Export captured waveforms and logic signals. Protocol Analyzer Note: the PCB & the programmed PIC micro are available from SILICON CHIP PartShop. Digital Oscilloscope is slid into place. That done, solder the diagonally opposite pin, then solder the remaining pins, ignoring the inevitable solder bridges. Removing the solder bridges Once the device is in place, apply a thin layer of flux paste along both rows of pins, then clean up the solder bridges with solder wick. That’s done by first placing the solder wick alongside (but not on top of) the pads. The soldering iron is then placed on top of the solder wick and the wick gently slid towards the solder on the pads. As the wick heats, it will start to melt the flux and the excess solder, creating visible smoke. At that point, you can slide it right up against the pins and most of the excess solder should then be sucked into the braid. siliconchip.com.au Repeat this procedure until all the solder bridges are gone. It’s not strictly necessary to clean off the flux residue provided you are using no-clean (noncorrosive) flux. However, if you do want to clean it off, this can be done using pure alcohol (eg, isopropanol). For a more detailed description on soldering in SMD ICs, refer to pages 80-82 of the June 2012 issue. The passive SMDs are next on the list. These include nine 3.2 x 1.6mm (3216) resistors and 8-10 similarly sized ceramic capacitors. The resistors have their value code printed on top but the capacitors will be unlabelled. As before, the best tactic is to remove them one at a time from their packaging, so you don’t get them mixed up. Regulator REG1 is the last SMD components to be installed. It’s mounted Spectrum Analyzer Compatible with major operating systems including Windows, Linux & Mac OS X, Pocket Analyzer is your ideal test and measurement companion. bitscope.com/sc September 2012  77 30 Using PWM To Reproduce PCM Audio 20 10 0 2 0 10 20 30 0 2 2 1 2 2 3 24 Fig.5(a): 32-sample sinewave with 30 voltage steps (5-bit resolution) 8 7 6 5 Uncompressed digital audio is stored in Pulse-Code Modulation or PCM format. This consists of a series of numbers which represent a proportional voltage at a point in time. The voltage is sampled at a fixed rate (the sampling rate) and stored. The resulting numbers form a representation of the audio waveform. Refer to Fig.5(a); this shows a sinewave converted to 5-bit PCM with 32 samples. With five bits, we have about 30 voltage steps and as you can see, some of the sample points (blue) don’t quite line up with the original waveform (red). In reality, we use more than five bits but this is just an example. At the bottom of Fig.5(a) is a plot of the five data bits. Consider the first sample (left-most blue dot), which has a value of 15. This is encoded as 20 + 21 + 22 + 23 (1 + 2 + 4 + 8) and hence the bits corresponding to these numbers are high (one) whereas the top bit, 24, is initially low (zero). Some bits then flip as the sampled value changes and a new sample is binary encoded. Now look at Fig.5(b). We’ve taken the three most significant bits (MSBs) and re-numbered them to start from zero, without changing the data. The resulting sample values are plotted above. The resulting waveform has the same general shape as the original but it lacks the fine details since the least significant bits (LSBs) are gone. This is a type of audio decimation; dropping some of the LSBs is an easy way to reduce the amount of data required to store a PCM stream but it also reduces the audio quality. It’s how we convert 16-bit audio to 12-bit for storage in the Digital Sound Effects Module (as mentioned early in the main article). In Fig.5(c) we plot the two LSBs missing from Fig.5(b). This 4 30 3 2 1 0 2 20 0 10 20 30 0 10 21 2 SUM 2 Fig.5(b): three most significant bits with corresponding waveform 0 3 0 2 0 10 20 30 0 0 10 20 30 Fig.5(d): summing waveforms with appropriate scaling reproduces original wave shape; deviations are due to limitations of the 5-bit resolution 21 Fig.5(c): two least significant bits with corresponding waveform in a similar manner to the ICs and SOT-23 devices. Through-hole parts Now flip the PCB over and mount the cell holder. The two electrolytic capacitors can then be fitted, with their longer leads going through the holes marked “+” on the PCB. 78  Silicon Chip Finally, complete the assembly by fitting four M3 x 9mm tapped Nylon spacers to the corner mounting holes. These are secured using M3 x 6mm machine screws. Programming If you don’t have a pre-programmed PIC (available from the SILICON CHIP PartShop), program it now. To do this, temporarily short out the two 1kΩ resistors (eg, using a lead off-cut) and then plug (or solder in) a 5-way pin header in for CON1 (friction will hold it in place). That done, connect an ICSP (in-circuit serial programmer) with its pin 1 orientated as shown. If you are using a siliconchip.com.au shows the “residual” or the difference between the original and decimated waveforms. If we add this waveform to the decimated version with the correct ratio, we reconstruct the original 5-bit audio data, as shown above. This summing can occur either digitally or in analog. The 12-bit audio used in the Digital Sound Effects Module sounds almost as good as 16-bit audio but only requires 75% as much memory. Using the technique shown here, we split each 12-bit sample into the six MSBs and the six LSBs. These sample values are each then converted to an analog voltage using a pair of 6-bit PWM (pulse width modulation) outputs on the microcontroller. A sample value of 0 give us a 0% duty cycle and a value of 63 (26 - 1) gives us a near-100% duty cycle. The PWM square waves pass through low-pass RC filters which remove most of the switching noise and give us smoothly varying voltages which are proportional to the duty cycles and hence sample values. All that’s left is to then sum the two resulting analog waveforms in the correct ratio (64:1). This is achieved using a resistive voltage divider/mixer and we then have an analog signal which closely approximates the original audio waveform, to within about 0.025% (1/4095). PICkit programmer, you can power IC1 at 3.3V from its inbuilt power supply. Alternatively, you can connect a USB cable from a PC for power (although the device won’t be detected yet). The software, 0910912A.hex, is available from the September 2012 downloads section on the SILICON CHIP website (www.siliconchip.com.au). Unplug the CON1 pin header when you have finished programming IC1. and tells you how to determine which COM port number has been assigned. That done, download and install the free serial terminal program TeraTerm Pro (available from www.ayera.com/ teraterm/). Launch it and open the port assigned to the USB driver. Now type “Info” and press enter and you should get a response confirming that the module is working and showing its firmware version and the amount of free memory. Installing the driver Assuming that the PIC micro has been programmed, the next step is to plug the unit into a USB port and check that it is detected. A message should pop up adjacent to the System Tray saying “Installing device driver software” (Windows 7) or “Found New Hardware” (XP) – see Fig.6. You will then be prompted to locate a driver. You need to use the SILICON CHIP USB serial port driver, which is also available for free download from the SILICON CHIP website. This is the same driver as used for the Maximite and several other SILICON CHIP projects. The accompanying panel details the driver installation siliconchip.com.au Oversampling Using 6-bit PWMs gives a maximum switching frequency of 188kHz (the 12MHz CPU clock divided by 26 or 64). We want a switching frequency well above 20kHz so that we can filter out most of the switching harmonics without affecting the audible frequencies (20Hz-20kHz). A typical audio sampling rate is between 8kHz and 48kHz, giving between 23.4 and 3.9 PWM pulses per audio sample. To give a smooth output waveform without lowering the switching frequency (for the reasons explained above), we generate intermediate steps for the samples so that we can update the PWM duty cycle after each pulse. You can think of this as a joining-thedots exercise (see Fig.5(b)). The original samples are shown as blue dots but we could also put dots anywhere along the red line segments joining them to get an estimated intermediate sample value that we can then use to update the PWM duty cycles, making the output less “steppy” and more smooth. This is shown in the scope grabs in the main article. Preparing the sound files You now need to prepare the sound file(s) so that they can be uploaded to the module. They must be saved as mono 8-bit or 16-bit PCM WAV files with a sampling rate of between 8kHz and 48kHz. To check the format of your files or to convert them if necessary, you can use a free sound-file editing program called “Audacity” (http://audacity. sourceforge.net/download/). Load up Audacity and open a sound file. At the left side of the window, you can see whether it is stereo or mono. If stereo, use the “Stereo Track to Mono” option from the “Tracks” menu to mix them together. At the bottom of the screen, select the desired sampling rate (labelled “Project Rate (Hz)”). Keep in mind that higher sampling rates give better audio quality but also use more memory. There’s no point selecting a higher rate than that of the original file (which will be the default). Volume adjustment The volume adjustment can now be done. The module plays the file back with a fixed scale (the supply voltage will affect the volume somewhat). If your sound files are full-scale (ie, normalised), then the peaks may be slightly clipped due to the limited output power of the unit, especially if they have heavy bass. For best results, the audio file should be normalised to about 1dB below full-scale. This can be achieved by using the “Amplify” option from the “Effects” menu, then reducing the dB level shown by 1dB and clicking “OK”. If you find the sound is too soft or too loud, you can repeat these steps later but turn the dB level up or down as necessary. Keep in mind that as you increase the amplification, you may find the sound gets distorted due to clipping. In some cases though, this Fig.6: these are the messages that appear adjacent to the System Tray in Windows 7 (top) and Windows XP (bottom) when the Digital Sound Effects Module is initially plugged into a USB port (ie, before the USB driver is installed). September 2012  79 How To Install The USB Driver F OLLOW THESE steps to install the USB CDC driver in Windows 7 and XP (the procedure is similar for Vista): (1) Plug in the Module, open Device Manager and check that it has been detected as an “Unknown device” or “USB device” in the “Other devices” category. (2) If the Found New Hardware Wizard doesn’t automatically launch, right-click on the device and select “Update driver”. You may be asked whether you want to check Windows Update for a driver. If so, select “No, not at this time” or similar. (3) In Windows XP, choose to install the software (driver) automatically. In Windows 7, select the “Browse my computer for driver software” option. For Windows XP, select “Install from a list or specified location”. Then for both operating systems, browse for and select the folder containing the SILICON CHIP USB driver. (5) You will now likely get a warning that the driver is not signed. Choose to continue and install the driver anyway. (6) Verify that the driver installation is successful. The device should now show up under the “Ports” category in Device Manager. Make a note of the COM port number assigned. may be acceptable – it depends on the characteristics of your sound. Saving the file(s) Now use the Export function from the File menu. If you want to upload the file to the module with a 12-bit resolution, select “WAV (Microsoft) signed 16-bit PCM” from the dropdown at the bottom of the file dialog (it will be converted from 16-bit to 12-bit by the module). For an 8-bit resolution, select “Other uncompressed files”, then click the Options button. For Header, choose “WAV (Microsoft)” and for Encoding, select “Unsigned 8 bit PCM”. Then click OK. You can now select a location, type in a name and press the Save button. The file is then ready to be uploaded. Repeat this process if you are going to upload multiple sound files. With the files prepared, plug the module into a spare USB port, load 80  Silicon Chip up TeraTerm Pro (or if you prefer, another XMODEM-capable terminal program) and connect to the virtual serial port, as described earlier. You can now upload the files using the “Send” command, as described in the “Serial Commands For The Sound Effects Module” panel. Using the module With the sounds uploaded and the configuration set (don’t forget to save it!), you are ready to hook up the power, speaker and trigger inputs. Connect the speaker across the two solder pads. Any 8Ω speaker will do and the more efficient it is, the better. It will also sound much better if it is baffled. The simplest way to do this is to mount the speaker in a timber box. A tuned box will give the best sound quality but in general, any baffle is better than none. You can connect the speaker to the terminals either way around, since the phase doesn’t matter. There is also space to glue or tape a small (~27mm) speaker to the back of the PCB, next to the cell holder. You can do this if you’re really pressed for space and your sound requirements are modest. However, it will limit the volume and give poor bass response. The trigger switches must be connected between the trigger input pads and the nearby ground pad, either via a pin header connector or by soldering thin wires (eg, ribbon cable) directly to the PCB. You can use microswitches, pushbuttons, reed switches or even the output of a microcontroller or digital logic IC to trigger the unit. Having done all that, it’s just a matter of inserting the lithium cell into its holder or wiring up the external power supply (AC or DC) to CON3. You can then activate one of the trigger inputs and check that the sound(s) play back SC as they should. siliconchip.com.au