Fullscreen HTML5Med Res (??MB)HTML4

Add bling to your hifi amplifier! Pt.1: By Nicholas Vinen 100dB Stereo LED Audio Level/VU Meter Give your hifi system WOW factor with this spectacular stereo VU meter. It uses 80 high-brightness SMD LEDs to give any stereo amplifier/mixer a highly colourful dual-bargraph display which simultaneously shows the average audio signal level plus peak levels. And it can display signal ranges up to a whopping 100dB. Y EARS AGO, some big and expensive stereo power amplifiers sported large dual VU meters to indicate the power levels in both channels. But while they looked quite impressive, they were a bit of a gimmick since their analog meter movements could only display signal averages. They certainly weren’t fast enough to display the peak signal levels which would have been a big advantage. And of course, professional audio mixing desks also typically have VU metering but these days it is usually based on LED arrays which show average and peak signal levels, just like this new SILICON CHIP design! The SILICON CHIP Stereo LED VU Meter uses no less than 80 high-brightness SMD LEDs to give a dual bargraph dis32  Silicon Chip play of the average audio signal level, with dots indicating the peak levels. It can be configured to display a dynamic range between 40dB and 100dB, depending on your application. It’s suitable for monitoring line level signals or power amplifier outputs. Whether or not you actually need to add this meter to a piece of equipment, once you see it in action, you’ll want to fit it in anyway! We’ve put a video of it operating at the following URL so you can see for yourself: www.siliconchip. com.au/Videos/StereoVUMeter If you are a keen hifi enthusiast, this VU meter can show you how much headroom you have from your audio amplifier, ie, it can show how many more decibels it is capable of delivering before clipping and this is indi- cated with the peak dot display. It can also be used when recording or mixing, to ensure that the incoming audio signal(s) are consistent with each other and none of them are going to overload and cause excessive distortion or loss of dynamic range. One important feature of an audio level meter is that it should update relatively fast, so you can see the dynamic nature of the signal, but not so fast that your eye can’t track it. Traditional “VU Meters” were designed with a response time of 300ms (to 99%), to give a reasonable impression of signal loudness but also because the needle could only move so fast. Since this meter uses digital technology, it can show the peak and average level simultaneously and the peak siliconchip.com.au +3.3V LEFT INPUT CON1 +1.65V x 23 AN9 x 23 RC7 AN2 22k RC8 RC6 AN0 RC5 RC4 AN11 +3.3V RC3 RC2 RIGHT INPUT CON2 VR1 BRIGHTNESS +1.65V RC1 RC0 (HPF) (HPF) 22k LPF x 23 AN1 VDD (8 MODE LEDs TOTAL) +3.3V +3.3V A LED81 A λ LED82 40dB A λ LED84 80dB K +3.3V A λ LED83 60dB K +3.3V λ 100dB K S1 RB3 RA7 RB4 RA10 RB5 RB12 RB6 RA9 RB10 RB7 RB11 K RB2 RB8 RB9 λ λ λ λ LED3 λ λ λ λ LED2 λ λ λ λ λ λ λ λ A7 A6 A5 A4 A3 A2 A1 A (80 LEDs TOTAL IN MATRIX) AN3 22k LED4 A8 IC1 PIC32 AN10 x 23 LED34 LPF A10 A9 LED33 22k RC9 LED32 (HPF) LED31 10-BIT ANALOGTO-DIGITAL CONVERTER (HPF) LED1 A K K D Q1 K1 K2 G D S Q2 K3 K4 K5 K6 K7 G D S G Q3 D S K8 G S S2 Fig.1: a simplified circuit showing how the Stereo VU Meter works. The audio signals (left & right channels) first pass through a high-pass filter/attenuator, followed by an active low-pass filter and two gain stages. The signals before and after the gain stages are then fed to microcontroller IC1 which does the peak and average calculations, then drives a multiplexed LED display using outputs RB2-RB9 to control eight cathode-driving Mosfets (Q1-Q8) and outputs RC0RC9 to drive the LED anodes directly. level can have a fast rise time and a slow fall time. The fast rise-time allows the circuit to “catch” those very fast and short peaks, while the slow fall time allows you to better see them when they occur. For purists, we’ve implemented a VU-style meter mode so you can stick with the traditional 300ms rise/fall time averaging if that is what you want. The two bargraphs, one for each stereo channel, consist of 40 individual LEDs. With a dynamic range of 40dB, that means that each LED lights for a 1dB increase in signal level. If you select the 60dB range, that gives 1.5dB/ LED; the 80dB range gives 2dB/LED; and the 100dB range, 2.5dB/LED. These are SMD LEDs with rectangular lenses roughly 2mm square – and they are incredibly bright pin-points. An on-board pot can be set reduce the brightness for a darker room! We’ve used a green/yellow/amber/ red colour scheme. The colour shifts simply give a warning that you are approaching the clipping level. A pushbutton can select one of four siliconchip.com.au full-scale signal levels: -10dBV, 0dBV, +4dBu or +7dBV. Different equipment will have different line levels and one of these will suit most devices: -10dBV (316mV RMS) for some battery-powered consumer equipment such as iPods and mobile phones, 0dBV (1V RMS) for other consumer equipment, +4dBu (1.228V RMS) for some professional gear and +7dBV (2.24V RMS) for CD, DVD and Blu-ray players and some other equipment. You can make fine or coarse adjustments to these levels or set your own levels based on reference signals fed into the unit. The unit runs from a 12-15V DC supply so it can be powered from a small plugpack, 12V lead-acid battery or a low-voltage internal rail in an amplifier or similar equipment. It only draws about 50-150mA when operating, depending on the LED brightness setting. The PCB includes RCA input sockets and a DC socket for power, so it can be used in its own case with a clear lid or incorporated into another piece of equipment and hard-wired in place. The metering circuit’s signal-tonoise ratio is good enough to allow you to use the 100dB range with a +7dBV reference level and use the whole range of the device. The 100dB range can also be used with a lower reference level but you will need to use the noise nulling feature to get a blank display with no signal. This feature is especially useful to subtract source noise (eg, from the driving equipment) from the display when operating in the higher dynamic range settings. Other features include: software adjustments to cancel out variation in brightness between different colour LEDs; peak+average, peak-only and average-only modes; non-volatile mode and calibration settings; and mode indicator LEDs (see specifications panel for more details). Principle of operation Fig.1 is a simplified circuit diagram which gives an overview of the operation of the unit. It’s based on 32-bit microcontroller IC1 which operates at 40MHz and incorporates a 10-bit June 2016  33 Features & Specifications • • Display: two rows of 40 SMD 3216 (1206 imperial) LEDs Input signal: up to 2.33V RMS (+7.36dBV) or higher with changed resistors (eg, to suit power amplifier outputs) • • • Frequency response: 5Hz-20kHz, -3dB (see Fig.2) • Meter range: selectable 40dB, 60dB, 80dB or 100dB (1dB/LED, 1.5dB/LED, 2dB/LED or 2.5dB/LED) • Reference level: selectable -10dBV (316mV RMS), 0dBV (1V RMS), 4dBu (1.228V RMS), 7dBV (2.24V RMS) or custom levels (per-channel) • • Power supply: 12-15V DC, ~50-150mA • • Inputs: RCA sockets for signals, 2.1mm or 2.5mm ID DC socket for power Input impedance: approximately 37kΩ Modes: peak+average (dot/bar), average only (bar), peak only (bar), VU-style (peak+average or average only) Brightness adjustment: 10-100% via potentiometer (onboard trimpot or chassismounted) Other features: LED brightness matching for different colour LEDs in display, external/internal noise nulling, peak/average calculation period adjustment, supply reverse polarity protection analog-to-digital converter (ADC) with an input multiplexer. Seven of the analog inputs are used. The two 40-LED bar displays (80 LEDs in total) are multiplexed in eight groups, each group of 10 sharing a common cathode which is driven by one of eight N-channel Mosfets Q1-Q8, which are in turn controlled by the microcontroller’s outputs RB2-RB9. The LED anodes are driven directly by outputs RC0-RC9. Only a 4x4 portion of the 8x10 LED matrix is shown but you can see the general arrangement. The eight indicator LEDs (LEDs8188, four shown in Fig.1) are driven directly at their cathodes from eight microcontroller outputs. The anodes connect to the 3.3V supply via currentlimiting resistors. Actually, there are no current-limiting resistors for the matrixed LEDs, as such. Instead, their current is limited by the internal impedance of the transistors which drive the micro’s output pins, in combination with a softwarelimited on-time/duty cycle. This results in a current drive of around 1.2mA/LED at full brightness, for a total LED bar current of around 100mA. Since this is split between 10 anode drive pins, that means around 10mA per pin. (IC1 has an “Absolute Maximum” rating of 15mA/pin and an overall limit of 200mA). The analog signals are fed into CON1 & CON2 at left. The circuit is 34  Silicon Chip designed for line level signals with a maximum level ranging from around 316mV RMS (-10dBV) up to 2.33V RMS (+7.35dBV), to suit most consumer equipment and also some professional audio gear. The input divider resistors can be changed to allow much higher signal amplitudes, eg, to suit the outputs of a power amplifier. Further adjustments in full-scale level can be made using the pushbutton interface. The signals are AC-coupled and attenuated to no more than 1.16V RMS (3.3V peak-to-peak) using a resistive divider. This AC coupling serves as a high-pass filter (HPF) to remove DC and very low frequency signals (<5Hz). Note that there is another HPF at the ADC inputs of IC1. There is also an active low-pass 3-pole filter inserted immediately after the resistive divider, built around a dual op amp (for both channels) with a -3dB point of 20kHz. It provides a fast roll-off, with around 20dB of attenuation by 40kHz and also incorporates RF filtering. The overall effect of the low-pass and high-pass filtering on the frequency response is shown in Fig.2. Analog-to-digital conversion The signals from the output of the low-pass filters are fed directly to a pair of analog inputs on IC1. These signals are also amplified by 23 times and the amplified signal is fed to another pair of analog inputs. This signal is further amplified by another 23 times (ie, 529 times total) and fed to a third pair of analog inputs. The seventh analog input is used to sense the position of the brightness pot. The micro samples all seven inputs, with the six signal inputs sampled continuously at around 40kHz and the brightness pot sampled every millisecond or so. For each channel, the software uses whichever signal gives the most accurate reading, ie, the unamplified input for higher level signals and one of the amplified inputs for lower level signals. This greatly improves its dynamic range, given that it only has a 10-bit ADC. A perfectly noiseless 10-bit ADC would give a dynamic range of around 20log10(210) = 60dB. CD-quality audio uses 16 bits and has a dynamic range of 20log10(216) = 96dB, so it is desirable for our VU meter to have a similar dynamic range. We could have used an external ADC however these mostly come in fine-pitch SMD packages and can be a little expensive. Instead, we have used the two gain stages described above, in combination with software input switching depending on signal level, to effectively provide the extra bits needed to achieve a 100dB dynamic range while only using the existing 10-bit ADC. Two momentary pushbuttons, S1 and S2, are used to configured the unit. These can be used to change a variety of settings such as the meter range, reference level, peak/averaging mode, level calibration, brightness calibration and so on. When pressed, S1 pulls input RB10 low while S2 pulls RB11 low. These are held high by internal pull-up resistors which are enabled by the firmware. In essence, the circuit concept is relatively simple but a lot of the work, including the peak and RMS calculations, averaging, display multiplexing and so on are done by IC1’s firmware. Full circuit description The full circuit diagram for the Stereo VU Meter is shown in Fig.3. The left and right channel input circuitry is identical so we will describe the left channel only, with part numbers for the right channel in brackets. From CON1 (CON2), the signal is AC-coupled via a 2.2µF ceramic capacitor and DC biased to +5.6V via a 22kΩ resistor. The 5.6V rail is half the siliconchip.com.au 11.2V op amp supply, allowing the signal to swing symmetrically between the op amp supply rails (ie, between 0V and 11.2V). This 22kΩ bias resistor forms a divider in combination with the input series resistor, attenuating the input signal by half while keeping the input impedance relatively high at around 37kΩ. As explained later, these resistor values can be changed to allow the unit to handle higher signal levels. As stated earlier, the 2.2µF capacitor and 22kΩ series resistor form a highpass filter for the input signal. Further high-pass filtering is also performed by the software. Dual SMD Schottky diode D2 (D3) clips the input signal, should it go below ground or above the 11.2V op amp supply rail. The BAT54S has a low forward voltage of around 0.3V, so it will conduct before the junctions in the op amp's inputs. Low-pass filter The following active 3-pole lowpass filter is built around an NE5532D low-noise op amp IC2a (IC2b). The first section of this filter, consisting of a 3.9kΩ series resistor and 680pF capacitor, also serves to attenuate much higher frequency signals (eg, AM radio) which may have been picked up by the signal leads. This filter is a special case of the Sallen-Key active low-pass filter (see www.beis.de/Elektronik/Filter/Act3PoleLP.html) which gives a -18dB/ octave roll-off using a single active device (in this case, an op amp). The values were chosen carefully, to give a near-flat bandpass response (ie, Butterworth characteristic) using E24-series resistors and E12-series capacitors (see Fig.2). Note that the 3.9kΩ resistor was chosen with the 11kΩ impedance of the preceding divider being taken into account. In other words, the resistance between the input and the 680pF capacitor should be 14.9kΩ; the two 22kΩ resistors are effectively in parallel when considering the impedance feeding the RC filter, hence we subtracted 11kΩ from 14.9kΩ to get 3.9kΩ. The filtered output from pin 1 of IC2a (pin 7 of IC2b) is fed to analog input AN0 (AN1) of microcontroller IC1 via a 2.2µF capacitor and 1kΩ series resistor. This resistor, in combination with dual Schottky diode D4 (D7), prevents the voltage at the micro’s input siliconchip.com.au Fig.2: frequency response for the analog portion of the circuitry. The steep (-18dB/ octave) roll-off of the low-pass filter can be seen, with virtually no attenuation below 15kHz, -3dB at 20kHz, -10dB at 30kHz and around -20dB at 40kHz (not visible). The bass roll-off is from two passive high-pass RC filters and gives a -3dB point of 5Hz and -10dB at around 1.6Hz. from going below -0.3V or above +3.6V. In addition, a 22kΩ resistor provides a DC bias of 1.65V, ie, the halfway point of the micro’s 3.3V supply. This keeps the signal within the ADC’s input range of 0-3.3V for input signals of up to 2.33V RMS (keeping in mind the 2:1 input attenuation). While the micro has internal clamp diodes to protect its inputs and it might seem like D4 & D7 are overkill, we discovered an interesting “feature” of the PIC32 series – if you drive an analog input pin beyond its supply rails, even within its input current rating, the chip may reset! Hence these clamp diodes are mandatory and as before, Schottky diodes are used because they will conduct before the IC’s internal semiconductor junctions. Gain stages IC3a (IC3b) and IC4a (IC4b) are both configured as non-inverting amplifiers with a gain of 23, set by the ratio of the 22kΩ and 1kΩ feedback resistors. A 100pF capacitor across the 22kΩ resistor rolls off its frequency response, reducing high-frequency noise gain and making the whole circuit quieter. To achieve the 100dB dynamic range relative to 7dBV (2.24V RMS), we need less than 22.4µV RMS input-referred noise throughout the entire system. That is why we’re using NE5532 lownoise op amps. The total gain of the system from the input to analog input AN9 (AN10) is 0.5 x 23 x 23 = 264.5 times. A -100dB (2.24µV) input signal will thus be amplified to 6mV RMS. The 10-bit ADC can sense voltage steps of 3.3V ÷ 1024 = 3.2mV. So this level of signal can be (just) measured by IC1. Driving the LEDs Driving the 80 LEDs in the VU meter takes up a total of 18 output pins. Each bar of 40 LEDs is broken up into four groups of 10 and the cathodes of the LEDs in each group are tied together. These cathodes are driven by one of logic-level Mosfets Q1-Q8 which are in turn driven from microcontroller outputs RB2-RB9 (pins 23, 24, 33, 4144 and 1 of IC1 respectively). Because these Mosfets have such a small gate charge, no extra circuitry is required. Each Mosfet is switched on in turn for around 1.6ms. When Q1 is switched on (ie, pin 23 of IC1 [RB2] is high), LED1-LED10 can be lit when their respective anodes are driven high. The anodes are driven by outputs RC0-RC9 of micro IC1 (pins 25-27, 36-38 & 2-4 respectively). June 2016  35 +11.2V +11.2V 100nF X7R D2 BAT54S LEFT INPUT CON1 100nF X7R +3.3V X7R 2 22k 3 16V X7R 1nF 680pF 50V C0G 50V C0G 22k 8 3 1 IC2a 2 20k 22k 3.9k 8 3 1 2.2µF 2 4 22k 22k 100pF 100pF 50V C0G 50V C0G 1k D4 BAT54S 1k 3 IC2, IC3, IC4: NE5532D 5 1 22k 1k +3.3V 2 1nF 50V C0G 5 5 7 IC2b 6 20k 22k 50V C0G 22k +5.6V 2 1 +1.65V 2 3 680pF D5 BAT54S +5.6V D3 BAT54S 3.9k 1k 2.2µF 3 3 +11.2V 16V X7R 1k 22k 2.2µF 22k 2 1 4 2.2µF 50V C0G 1 2.2µF 2 D6 BAT54S 1 IC4a 100pF +5.6V 22k 8 3 1 IC3a 4 +1.65V RIGHT INPUT CON2 100nF 7 IC3b 6 6 D9 BAT54S 7 IC4b 2.2µF 100pF 50V C0G 22k 100pF 100pF 50V C0G 50V C0G 1k 22k 2.2µF 22k +1.65V 1k D7 BAT54S 1k +5.6V 22k 1k 2.2µF 3 D8 BAT54S +5.6V 2 1 +1.65V 22k 1k +5.6V 1 2 1 3 3 2 +3.3V +3.3V VR1 10k 4x1k A A +11.2V 8 2.2µF 5 16V X7R IN OUT REG2 MIC5201YM ADJ EN GND NC NC NC 3 4 6 7 1 1.5k 2 10Ω A 1k 2.2µF 16V X7R 12k A +5.6V K LED81 λ 60dB λ 80dB + D1 SS14 TPV+ 6.3V X5R K 4x1k A A K 33Ω 1W q CON3 TPG1 2.2µF 16V X7R TP3.3V REG1 MCP1703-3302E/DB 1 IN OUT GND GND 2 4 A A +3.3V 3 LED85 λ -10dBV 3.9k LED86 LED87 λ 0dBV λ 7dBV K K K +1.65V 6.3V X5R 47µF 6.3V X5R BZX84-C5V6 SS14 SC λ 4dBu LED88 K 10µF 3.9k 2016 K K A 12VDC POWER λ 100dB K 47µF ZD1 BZX84 A C5V6 λ 40dB LED82 LED83 LED84 DIGITAL STEREO AUDIO LEVEL/VU METER K A K A (NC) Fig.3: full circuit for the Stereo VU Meter. This shows the full 80-LED matrix at right, along with the power supply and details of the analog circuitry. Dual Schottky diodes D2 & D3, in combination with the 22kΩ series resistors, protect IC2 from excessive input signal levels while D4-D9 (with 1kΩ series resistors) prevent IC1’s internal input clamp diodes 36  Silicon Chip siliconchip.com.au RIGHT CHANNEL LEFT CHANNEL RPC6/PMA1/RC6 RPC5/PMA3/RC5 RPC4/PMA4/RC4 RPC3/RC3 AN8/RPC2/RC2 AN7/RPC1/RC1 AN6/RPC0/RC0 RS529 14 RS23 22 RS1 20 5 A10 4 A9 3 A8 2 A7 38 A6 37 A5 36 A4 27 A3 26 A2 25 A1 AN10/RB14/RPB14 LED7 λ λ λ λ λ λ λ λ λ λ λ λ LED4 λ λ λ λ LED3 λ λ λ λ LED2 λ λ λ λ λ λ λ 2 AN1/VREFq/RA1 8 5 CON4 A S1 AN11/RB13/RPB13 LED1 S2 A 23 K1 24 K2 33 K3 PGED3/RPB5/PMD7/RB5 41 K4 42 K5 43 K6 44 K7 1 K8 AN4/RPB2/RB2 AN5/RPB3/RB3 35 SOSCI/RPB4/RB4 RA7/PMA7/TCK RA10/PMA10/TMS/PGED4 RB12/PMD0/AN12 PGEC3/RPB6/PMD6/RB6 RPB7/PMD5/RB7 RA9/TDI/RPA9/PMA9 RPB8/PMD4/RB8 RPB9/SDA1/PMD3/RB9 34 32 31 30 λ LED78 λ λ λ λ LED77 λ λ λ λ LED76 λ λ λ λ LED75 λ λ λ λ LED74 λ λ λ λ LED73 λ λ λ λ LED72 λ λ K λ λ LED79 K G A A λ λ K LED71 K Q1 D S G Q2 D S G Q3 D S G Q4 D S G RA4/SOSCO/RPA4 Q5 D S Q6 RA8/TDO/RPA8/PMA8 G RA3/OSC2/CLKO/RPA3 VCAP RA2/OSC1/CLKI/RPA2 AVSS 16 D2qD9: BAT54S 3 1 λ K D 10 λ A λ 4 9 TPG1 12 λ LED40 3 IC1 PIC32MX170 PIC3 2MX170 qF256D 13 λ LED80 1 RB11/RPB11/PMD1/PGEC2 11 λ ICSP RB10/RPB10/PMD2/PGED2 BR λ λ λ 18 AN3/RPB1/RB1/PGEC1 TPBR λ LED50 RPC7/PMA0/RC7 λ LED39 LED49 RPC8/PMA5/RC8 K LED8 λ LED38 LED48 RPC9/PMA6/RC9 AN0/VREF+/RA0 λ λ LED37 LED47 19 MCLR AN9/RB15/RPB15 AN2/RPB0/RB0/PGED1 λ λ K LED31 LED41 LS1 X7R 40 VDD 28 VDD λ LED6 21 17 AVDD LED9 LED5 15 X7R A 10k 100nF K A λ LED36 LED46 100nF X7R LS23 λ λ K 100nF LS529 λ LED35 LED45 10Ω λ LED34 LED44 LED10 A LED33 LED43 A LED32 LED42 +3.3V 2 VSS 6 VSS 29 VSS 39 REG2, IC2qIC4 8 7 6.3V X5R 4 1 G 10µF CATHODE BAND A D S Q7 D S G Q1-Q8: 2N7002P LEDS K D G S Q8 S MCP1703T-3302E/DB 4 (GND) (IN) 1 (GND) 2 3 (OUT) from conducting, depending on the signal applied. Also shown in more detail on this diagram are the three different signal bias levels of 0V (GND), 1.65V (half supply for 3.3V) and 5.6V (half supply for 11.2V). The amplified and filtered input signals are fed to microcontroller IC1 which drives the LED bargraphs via Mosfets Q1-Q8. siliconchip.com.au June 2016  37 Simulating VU Meter Response The adjacent photo shows an original VU meter. According to Wikipedia: “The original VU meter is a passive electromechanical device, namely a 200µA DC d’Arsonval movement ammeter fed from a full-wave copper-oxide rectifier mounted within the meter case. The mass of the needle causes a relatively slow response, which in effect integrates the signal, with a rise time of 300ms. 0 VU is equal to +4dBu, or 1.228V RMS across a 600-ohm load . . .” The SILICON CHIP Stereo LED VU Meter does not attempt to exactly replicate the operation of a VU meter. For example, the copper-oxide rectifier causes a VU meter to be inaccurate for low-level signals and thus makes a reading below -20dB difficult. We’re still using proper RMS calculations and showing readings on a range of at least 40dB (depending on the range setting of the unit). However, we have been able to implement a simulation of the ballistic properties of the VU meter needle in our software. When a 1kHz sinewave is applied to a standard VU meter which had previously had no signal applied, it should reach 99% of the final reading in 300ms and have an ultimate overshoot of between 1% and 1.5% of the reading. We’ve achieved this with a simple needle inertia simulation that tracks the needle position and velocity along with a target position (based on the current RMS reading), an acceleration coefficient and a damping coefficient. These two coefficients were tuned to achieve the required response, as stated above. The cathodes of LEDs81-88 are driven by microcontroller outputs RA9, RB12, RA10, RA7, RA8, RA3, RA2 and RA4 respectively (pins 35, 10, 12, 13, 32, 31, 30 and 34). The 1kΩ currentlimiting resistors result in a drive of around 0.5mA each. Controls Tactile pushbutton switch S1 is used to cycle through the available meter ranges (and performs other functions when pressed and held or pressed in combination with S2). Similarly, switch S2 cycles through the four different reference levels. Their associated input pins, RB10 and RB11 (pins 8 & 9), are also used for programming and debugging IC1. There is no conflict as long as neither switch is pressed when the in-circuit serial programming (ICSP) tool plugged into CON5 is being used. Brightness pot VR1 forms a voltage divider across the 3.3V supply and the voltage at its wiper is sensed via analog input AN11 (pin 11) of IC1. The brightness is controlled by adjusting the LED matrix anode drive period. Power supply The power supply is relatively simple and uses only linear regulators but its design is critical to achieve the stat38  Silicon Chip ed performance. 12-15V DC is applied to socket CON3, or else wired directly between TPV+ and TPG1. Schottky diode D1 provides reverse polarity protection and drops less than 0.4V. The 11.6-14.6V DC at its cathode is fed to two resistors. One is a 33Ω 1W SMD type which connects to the input of 3.3V low drop-out (LDO) regulator REG1. This is in a relatively large SOT223 4-pin package that is soldered to a considerable copper area which includes thermal vias to conduct heat to the back of the board. With a 14.5V input, 3.3V output and maximum load of 125mA, these components dissipate (14.5V – 3.3V) x 0.125A = 1.4W. Around 0.5W will be dissipated in the resistor with the remaining 0.9W in the regulator, hence the copper plane heatsinking on both sides along with thermal vias under the regulator package. SMD multi-layer ceramic capacitors are used bypass the input and filter the output of REG1. As well as supplying the microcontroller and LEDs, the 3.3V output flows through a pair of seriesconnected 3.9kΩ resistors to generate the 1.65V rail, which is bypassed with a 47µF capacitor. This is important since the varying current demand on the 3.3V rail as LEDs are switched causes ripple and we don’t want this to couple into the input signal path. The 11.6-14.6V at the cathode of D1 also flows to adjustable low dropout regulator REG2 (MIC5201). The 10Ω series resistor forms a low-pass filter in combination with its 2.2µF input bypass capacitor to reduce the amount of ripple reaching REG2. It also causes a voltage drop of up to 0.25V. REG2’s output is set at 11.15V (nominal) by the 12kΩ and 1.5kΩ resistors. REG2’s minimum input voltage is around 11.6V – 0.25V = 11.35V. With a load current of only 25mA, REG2’s dropout voltage is less than 150mV, so it should stay in regulation. This is important since REG2 exists to eliminate ripple on the op amp supplies; even with a 100dB CMRR (common mode rejection ratio), several hundred millivolts of ripple can have a significant impact on the overall signal-to-noise ratio of the system. The 5.6V half-supply reference for the op amps is derived from the 11.2V supply with a 1kΩ resistor and 5.6V zener diode ZD1, which is bypassed with another 47µF capacitor. This further attenuates any ripple which may make it through the regulator. Microcontroller IC1 has three 100nF ceramic supply bypass capacitors, with its AVCC supply filtered by one of these in combination with a 10Ω series resistor. A 10µF ceramic capacitor between pin 7 (VCAP) and ground provides an output filter for its internal 2.5V core regulator, while a 10kΩ pullup between the 3.3V rail and pin 18 (MCLR-bar) prevents spurious resets. PCB layout The PCB design is crucially important to achieve the desired performance level. The relatively compact PCB means that the switched LED supply lines inevitably run somewhat near the front end and the tiny amount of signal that couples in is picked up and amplified by the op amps. This is mostly ignored by the software (as explained later) but it does make the noise level in the left channel slightly worse than the right (but still below -100dB) due to its greater proximity to the LEDs. The most critical part of the layout, as is typical, is the ground track routing. While all points connected to ground in the circuit must be joined together, there are a large number of ways this could be achieved, many of which would cause digitally switched siliconchip.com.au currents to cause voltage shifts across different parts of the analog front-end, ruining the performance. The grounds for input sockets CON1 & CON2, regulators REG1 & REG2 (and their output filter capacitors but NOT input bypass capacitors) and the bottom end of the two half-rail bias generators are connected together, then brought back to pin 16 of IC1 (AGND). This pin is also used as the negative reference voltage for the ADC. This routing means that the analog grounds and half-supply rails have very little AC voltage between them, as little current flows through these tracks and that which does is linearly proportional to the input signal. AGND is connected to GND within IC1, and also by tracks running under it, so ultimately these pins all connect to GND. The op amp negative supply pins and their bypass capacitors are connected back to the ground pin of REG2 and the more-or-less constant supply current flows from there back to AGND and on to the main ground network. Firmware operation While the software is conceptually simple, it has quite a few modes and features. Its three main tasks are to continually sample the analog inputs, perform the average and peak calculations based on the results and drive the multiplexed LED display. The first and last are “real time” tasks so they are interrupt driven, with priority given to the LED multiplexing, as any delays could cause the display to flicker. The LED multiplexing is achieved using three of the five 16-bit timers internal to IC1: Timer1, Timer2 and Timer3. IC1 runs at 40MHz, in order to perform the required calculations fast enough to provide a rapid display update (around 40 updates per second). Timer1 is set up to trigger an interrupt every 65,536 (216) clock cycles, ie, every 65,536 ÷ 40MHz = 1.6384ms. Since there are eight cathodes to cycle through, this gives a display refresh rate of 76Hz. When this interrupt is triggered, the code branches to the Timer1 interrupt handler, which switches the LED anode drive outputs (RC0-RC11) low, then cycles the next Mosfet drive output (RB2-RB9) high. It then calculates which anodes to drive to light the appropriate LEDs and brings some combination of RC0-RC11 high. siliconchip.com.au Parts List 1 double-sided PCB, code 01104161, 177 x 75.5mm 1 white switched PCB-mount RCA socket (CON1) 1 red switched PCB-mount RCA socket (CON2) 1 PCB-mount switched DC socket, 2.1 or 2.5mm inner diameter (CON3) 1 5-pin header, 2.54mm pitch (optional, to program IC1) (CON4) 2 PCB-mount mini tactile switches (S1,S2) (Jaycar SP0611) 1 10kΩ SMD trimpot, TC33X type (VR1) (element14 1689863) 5 PC stakes (optional; see text) 1 12-15V DC 150mA+ power supply Semiconductors 1 PIC32MX150F128D-I/PT or PIC32MX170F256D-I/PT 32-bit microcontroller programmed with 0110416A.hex, TQFP-44 (IC1) 3 NE5532D dual low-noise op amps, SOIC-8 (IC2-IC4) 1 MCP1703-3302E 16V in, 3.3V out, 250mA low-dropout regulator, SOT-223 (REG1) 1 MIC5201YM 20V, 200mA adjustable low-dropout regulator, SOIC-8 (REG2) 8 2N7002P logic-level N-channel Mosfets, SOT-23 (Q1-Q8) 1 BZX84C5V6 5.6V 1/4W zener diode, SOT-23 (ZD1) 1 30V 1A Schottky diode, SMA/ DO-214AC or Mini2 SMD package (D1) 8 BAT54S dual series Schottky diodes, SOT-23 (D2-D9) 60 high-brightness green SMD 3216/1206 LEDs (LED1-30, LED41-70) 8 high-brightness yellow SMD 3216/1206 LEDs (LED31-34, LED71-74) 8 high-brightness amber SMD 3216/1206 LEDs (LED35-38, LED75-78) 4 high-brightness red SMD 3216/1206 LEDs (LED3940,LED79-80) If the brightness trimpot is set at a level below 100%, before the interrupt handler completes, it enables Timer3 and sets it for a value proportional to 8 high-brightness blue SMD 3216/1206 LEDs (LED81-88) Capacitors (SMD 2012/0805 unless specified) 2 47µF 6.3V X5R 2012/0805 or 3216/1206 2 10µF 6.3V X5R 2012/0805 or 3216/1206 11 2.2µF 16V X5R 6 100nF 50V X7R 2 1nF 50V C0G 2 680pF 50V C0G 6 100pF 50V C0G Resistors (SMD 2012/0805, 1% 0.125W unless specified) 2 22kΩ or 22.1kΩ 1% 0.5W 3216/1206 14 22kΩ 4 3.9kΩ 2 20kΩ 1 1.5kΩ 1 12kΩ 19 1kΩ 1 10kΩ 1 33Ω 5% 1W 6331/2512 2 10Ω Optional parts Tapped spacers and M3 machine screws for mounting, shielded cable and twin-lead for hard wiring, case with clear lid Optional laser-cut case 1 set 3mm clear acrylic pieces 1 small tube acrylic glue (solventbased) 4 M3 x 10mm machine screws 2 M3 x 12mm tapped Nylon spacers 4 M3 shakeproof washers Where to buy parts The PCB, programmed microcon­ troller, case pieces & red/white RCA sockets are available separately from the SILICON CHIP Online Shop. We’re also offering a set of parts containing all the SMDs except for the microcontroller & LEDs. Sets of 10 red, amber, yellow, green or blue high-brightness 3216-size SMD LEDs with diffused lenses are available separately, so that constructors can choose their own colours. the brightness. When the interrupt handler for Timer 3 is subsequently triggered, RC0-RC11 are brought low, cutting off the LED drive to reduce the June 2016  39 Handling Higher Amplitude Signals As presented, the circuit is designed to accept sinewave signals up to 2.33V RMS (6.6V peak-to-peak or 3.3V peak). Signals above this level will cause D2 & D3 to conduct and the meter will simply show a full -scale reading but no damage should occur. This is not sufficient to monitor the outputs of a power amplifier. Some equipment may also produce line-level signals above 2.33V RMS. In this case, it’s simply a matter of changing the input divider so that, with the maximum input signal level, the resulting voltage does not exceed 3.3V peak-to-peak. The parallel combination of the two divider resistors should be kept to 11kΩ. This means that the “lower” leg resistor will need to be reduced in value, probably to somewhere in the range of 11-15kΩ. Consider a power amplifier like our Ultra-LD Mk.4 (August-October 2015) which can deliver 135W into 8Ω or 200W into 4Ω. This requires an output voltage of √(135 x 8) = 33V RMS or 93V peak-to-peak (33V x 2.828). For 200W into 4Ω, the output will be √(200 x 4) = 28.3V RMS. So the meter would need to handle 33V RMS to monitor the outputs directly. This means a divider ratio of at least 93V ÷ 3.3V = 28.2 is required. We set the bottom leg to 11kΩ (ie, the resistor from the 5.6V rail) since the other resistor value will be much larger and have negligible contribution to the divider’s output impedance. The other resistor will then need to be at least 11kΩ x (28.2 – 1) = 300kΩ. This happens to be an E24 value; if it wasn’t, we would have chosen the next higher value. The parallel resistance is 1 ÷ [(1 ÷ 11kΩ) + (1 ÷ 300kΩ)] = 10.6kΩ, which is close enough to 11kΩ. This new divider reduces the input levels by 23dB compared to the original design, so the +7dBV setting is in reality now +30dBV, which would indicate full amplifier power into 8Ω. Note that the other reference levels will be effectively increased by the same amount. average display brightness. Both interrupt priorities are set to level four, the highest used. An error in the silicon? This multiplexing method works very well, giving a stable and bright display. At least, it did until we enabled the ADC. Even with ADC interrupts disabled, it caused the LED display to flicker. We couldn’t understand why, since the timer period should not be affected by the operation of the ADC and the flickering occurred even without enabling interrupts for the ADC unit at any location in the code. So why should the ADC interfere with the multiplexing? We got a clue by examining the Mosfet drive signals for Q1-Q8 using an oscilloscope. Periodically, one of the Mosfet gate drive pulses would be extremely short; rather than the expected 1.6384ms, it measured something like 50μs, appearing as just a spike on the screen. Following that occurrence, the next pulse for that same Mosfet would occur about 11.5ms later, or roughly seven timer periods, rather than the expected 13.1ms (eight timer periods). 40  Silicon Chip This pointed to the possibility that, having completed the Timer1 interrupt handler, the processor would sometimes immediately re-enter it and thus the code would then assume some time had passed and switch to the next multiplexed bank of LEDs prematurely. For some reason, this occurred only when the ADC was enabled and active. We can’t figure out how the software could have caused this, since when the ADC is in auto-sampling mode, with interrupts disabled, it operates autonomously. Simple solution In the end, we came up with a simple solution: we set Timer2 to the same time base and period as Timer1. When the Timer1 interrupt handler is about to exit, it resets Timer2. At the start of the Timer1 interrupt handler, we check the value of the Timer2 counter. If it’s less than half the expected value, indicating that the Timer1 interrupt should not have occurred yet, we exit the interrupt handler without doing anything. This allows us to ignore the occasional spurious event. Our guess as to the nature of this problem is that an implementation bug in this series of PIC32s causes some in- terrupt handlers to re-triggered upon exit if something else is going on in the chip simultaneously. A “gotcha” Another issue we ran into is that the display also started flickering when we added the code to update the states of LEDs81-88. Unlike the previous issue, this turned out to be our fault. LED86 is driven by output RB12 which is on the same port as the pins driving Mosfets Q1-Q8 (ie, RB2-RB9). The code to update the state of LED86 read: LATBbits.LATB12 = !(Sensitivity == 1); This looks like an atomic operation, setting the bit LATB12 which controls the state of output RB12. However, it is actually a read-modify-write operation, ie, the value of special register LATB is transferred to a general purpose register, the value of bit 12 is changed and the result is then transferred back to the special register. But if the Timer1 interrupt handler is triggered in the middle of this procedure, it may change the value of special register LATB. When the special register is written back later by the above code, that overrides the change made by the interrupt handler. One possible solution is to temporarily disable interrupts while updating LATB12 but that’s clumsy. The elegant solution is to take advantage of the special set/clear/toggle registers available for all the GPIO registers (and many others) on this chip, ie: if( Sensitivity == 1 ) LATBCLR = 1<<12; else LATBSET = 1<<12; These operations are atomic, ie, can’t be interrupted. If an interrupt handler is triggered simultaneously, it will be executed either before or after the clear/set operation, eliminating the flicker. Analog computations As explained earlier, the input signals (left & right channels) are applied to six different analog inputs: two with an overall gain of one half, two with a gain of 11.5 times and two with a gain of 264.5 times. However, rather than switch between sampling these different inputs depending on signal level, we simply sample all six, all the time. The code in the ADC conversion siliconchip.com.au We used green, yellow, amber and red LEDs in the bargraphs but you can change these to suit your requirements. Pt.2 next month has the assembly details. Note: prototype board shown. complete interrupt handler routine decides which to use, depending on the values returned. If the value from the input amplified by 264.5 times is very near either rail (ie, the conversion result is very close to either zero or 1023), the value from the 11.5 times amplified input is used instead, multiplied by 23 so that it has the same scale as readings from the other input. If that lower-gain input is also found to be very near either rail, the value from the third, attenuated input is kept instead, multiplied by 529 (23 x 23). The result is then stored in one of two 1024-entry stereo buffers for later analysis, to keep the interrupt handler routine quick so it won’t interfere with LED multiplexing. The micro’s peripheral clock is set to run at half the rate of the main clock, ie, 20MHz. The ADC clock input divider is set to a factor of two, giving 10MHz. Each conversion takes 12 cycles, plus eight cycles for the sample-and-hold buffer to stabilise, for a total of 20 cycles each (10MHz ÷ [20 x 6] = 83.3kHz). We average each pair of sample values to reduce system noise, resulting in a final sampling rate of 41.6kHz, sufficient for measuring signal frequencies up to 20.8kHz (the Nyquist limit). Once a sample buffer is full, the code’s main loop calculates the average of all values and subtracts it from the sample values to remove any DC offset or low-frequency signals which have not been rejected by the analog filter. It then computes the RMS and peak values for each channel. The resulting RMS values are then stored and averaged over a configurable number of 1024-sample intervals, so that it varies smoothly. Similarly, the highest peak values across multiple 1024-sample buffers are computed. These RMS and peak values are converted into decibels using a logarithmic calculation, taking into account the current range and reference level selection. The display is then updated to show the peak and/or average results (depending on the display mode). Other functions Button presses are sensed using a pin change interrupt and debounced using Timer4 (S1) or Timer5 (S2). When a press is detected, the interrupt handler sets a flag and the main loop updates the mode and then saves the new setting to flash memory. These saved settings are automatically loaded each time the unit is powered up. Long presses (>0.5s) have a different effect compared to short presses and this is determined by when the associated timer rolls over. As well as using these buttons to change the display range and reference level, Issues Getting Dog-Eared? LED Colours In our prototype, we used 30 green LEDs at the left end of each bargraph, followed by four yellow, four amber and two red. This is a purely aesthetic choice. You could make them all the same colour (eg, blue), or use a different combination than we did. While brightness matching between different colours is generally good enough, the software allows you to provide some drive compensation to reduce the difference in apparent brightness between different colour LEDs. We also decided to make LEDs8188 blue, in contrast to LEDs1-80 and the parts list reflects our choices. various combinations can be used to change settings, to be described next month in Pt.2. Noise nulling is implemented by storing the left and right channel RMS and peak values when S1 is held down, both in RAM and flash memory. These are subtracted from subsequent readings, with the RMS noise figure being subtracted in an RMS manner. That’s all we have space for this month. Next month we’ll get onto the PCB construction, testing, operation SC and fitting the unit into a case. Keep your copies safe with our handy binders Order now from www.siliconchip.com.au/Shop/4 or call (02) 9939 3295 and quote your credit card number. *See website for overseas prices. siliconchip.com.au June 2016  41