Mini sine wave oscillator

This project is a variant of the 'Miniature Audio Oscillator' by the great Rod Elliott.

I have made very few changes to the core schematic. I removed the potentiometer used for adjusting the frequency and replaced it with a DIP switch, allowing for easy switching between three different frequencies: 20 Hz, 1 kHz, and 20 kHz. These options should be sufficient for quickly testing various audio equipment.

Additionally, I added a third dual opamp. One half of the opamp is used as an output buffer, while the other half serves as a voltage rail splitter and virtual ground. The schematic is designed for a single rail voltage: a 9V battery should be sufficient, but it can also work with higher voltages up to 30V (depending on the opamps used). If Li-ion batteries need to be used, then it is recommended to connect 3 or 4 batteries in series.

The output signal, as stated in the original article, has very low Total Harmonic Distortion (THD) at approximately 0.12%. While I do not have the necessary measurement equipment to gauge the distortion of my variant, I estimate it to be below 0.5%.

I created this project using the EasyEDA software, with the aim of making it relatively easy to replicate at home. The design employs a single-sided PCB and through-hole components.

This is the schematic:

And this is 3D view of the PCB:

For this project, I used a photosensitive dry film for the first time. It wasn't easy, but after a number of trials and errors, I finally managed to produce a fairly good PCB. Here's how it turned out:

The oscillator utilizes a 4-channel DIP switch to control the activation of different value capacitors, thereby altering the output frequency.

The formula for calculating the frequency is F = 1 / (2π × R × C),

where R = R6 = R7 = 10k and

C = C1_A = C2_A = 680 pF when all switches are OFF. These two capacitors can also be 820 pF. To achieve a frequency close to 20 kHz, I handpicked two 680 pF capacitors with higher actual values.

When the left two switches are ON, C = C1_A + C1_B = C2_A + C2_B = 680 pF + 15 nF ≈ 15.7 pF.

When all switches are ON, C = C1_A + C1_B + C1_C + C1_D = C2_A + C2_B + C2_C + C2_D = 680 pF + 15 nF + 680 nF + 100 nF ≈ 795 nF. Note that C1_D and C2_D are included for fine-tuning the lowest frequency of 20 Hz. While not strictly necessary, omitting these capacitors will result in a slightly higher lowest frequency, around 23-24 Hz.

The output voltage measures approximately 3.5 Vpp or 1.24 Vrms and can be adjusted with the potentiometer.

The schematic is compatible with various types of opamps. I tested it with TL072 and NE5532. The current consumption with the TL072 is approximately 10-11 mA, whereas with the NE5532, it is around 34 mA. For extended battery life, the TL062 is the optimal choice; however, I do not have any available at the moment. With the TL072, the output signal begins to visibly distort when the supply voltage drops below 8V.

Here some screenshots from the oscilloscope:

And here is a video of the oscillator at work:

Project files can be downloaded from here: MiniOsc. Use them on your own responsibility!

Shaker for PCB etching

This is an old project, that was sitting in the cabinet for a very long time and I just finished it a week ago. It uses an old computer DVD drive (no one ever uses them anymore). It is completely stripped from all electronics and mechanics except for the parts that move the tray in and out. The motor inside is a simple DC motor.

The shaker uses an 8-pin microcontroller PIC16F15313 and a motor driver module based on DVR8833 chip.

There are two NO reed switches that are mounted on the case and two magnets glued on the moving tray. When one of the switches is closed by the magnet the microcontroller reverse the direction of the tray. There is also an potentiometer connected to an analog pin which control the movement speed of the tray.

The shaker is supplied with 12Vdc from an power adapter. It will work with voltages from 6V to 12V. For the microcontroller there is a 78L05 voltage regulator.

At first there was an power resistor mounted on the bottom of the aluminium sheet and connected directly to the 12V rail. The idea was to heat up the container with the etch solution, but because the container is plastic there was not much heat transfer going on and I removed it. Now I heat up the solution in the microwave oven first, and then I put the PCB in and use the shaker.

The circuit is assembled on piece of perfboard.

 

When using this shaker make sure the speed is not too high, because there is a risk of spilling potentially dangerous liquid around.

The code for the microcontroller is created in MPLAB X v6.08 and can be downloaded from here: Shaker_Source_Code.

IR remote code reader (NEC protocol)

This is a quick project on breadboard for decoding and displaying the signals from a IR remote.

The technical details of the NEC IR transmission protocol can be read here.

The project uses PIC16F628A microcontroller and 4 digit 7-segment common cathode LED display for showing the codes. The first two digits represent the address for the receiving device, and the next two digits represent the command code. The format of the codes on the display is hexadecimal.

The schematic is extremely simple:

The program is written in C using MPLAB X IDE 6.0 and compiled with the xc8 compiler.

The source code can be downloaded from here.

Driving NeoPixel type ARGB LEDs with PIC

 I have ordered 10 pcs of addressable LED from Aliexpress some time ago and tested them with my Arduino boards and the library from Adafruit and they work as expected. But of course I am more a PIC guy, so I start to thinker with some PIC microcontrollers trying to drive these LEDs. And it is not easy! So the major difference between 8-bit PIC micro and a ATMEGA 328P is that the ATMEGA command rate of ATMEGA is the same as clock rate, so it has 16 MHz clock frequency and 16 MHz command rate but the PICs have command rate 4 times slower than clock frequency. So a PIC clocked at 16 MHz will have 4 MHz command rate and 32 MHz PIC will have 8 MHz command rate. In order to achieve similar performance as ATMEGA, the PIC must be clocked at 64 MHz. 

The difficulty comes from the very high frequency and the format of the output signal. The frequency of NeoPixel signal is ≈800 kHz and the "1" have 800 ns high followed by 400 ns low. The "0" is 400 ns high followed by 800 ns low. And 1 instruction of 32 MHz PIC microcontroller take 125 ns to execute. It is impossible to write a C code that can create such signal in 9-10 instructions. This is achievable only with carefully written Assembler. I had success with PIC16F1847 clocked at 32 MHz.

It was pain in the a**, because there are very little info and tutorials about writing mixed code (C and assembler) for xc8. For example I couldn't find a way to declare a variable in BANK0 in assembly code because the variables in C code evidently take precedence and occupy all the free space in BANK0 first. And it is important the variables to in the same memory bank as PORTB, because changing banks takes one additional instruction. So I has to declare the variables in the C code specifying the exact address... 

PROCESSOR 16F1847 #include <xc.inc> #define SER_OUT PORTB, 4 EXTRN _colorByte EXTRN _index GLOBAL _sendByteASM, _sendByteASM2 psect ASMcode, class=CODE, local, delta=2 _sendByteASM: MOVWF _colorByte ; copy parameter to the local variable _colorByte MOVLW 8 MOVWF _index START: BSF SER_OUT BTFSC _colorByte, 7 GOTO ONE BCF SER_OUT LSLF _colorByte, 1 NOP NOP DECFSZ _index, 1 GOTO START GOTO ENDLABEL ONE: LSLF _colorByte, 1 NOP NOP BCF SER_OUT DECFSZ _index, 1 GOTO START ENDLABEL: return _sendByteASM2: MOVWF _colorByte ; copy parameter to the local variable _colorByte BIT0: BSF SER_OUT BTFSC _colorByte, 7 GOTO ONE0 BCF SER_OUT NOP NOP GOTO BIT1 ONE0: NOP NOP NOP BCF SER_OUT BIT1: NOP NOP BSF SER_OUT BTFSC _colorByte, 6 GOTO ONE1 BCF SER_OUT NOP NOP GOTO BIT2 ONE1: NOP NOP NOP BCF SER_OUT BIT2: NOP NOP BSF SER_OUT BTFSC _colorByte, 5 GOTO ONE2 BCF SER_OUT NOP NOP GOTO BIT3 ONE2: NOP NOP NOP BCF SER_OUT BIT3: NOP NOP BSF SER_OUT BTFSC _colorByte, 4 GOTO ONE3 BCF SER_OUT NOP NOP GOTO BIT4 ONE3: NOP NOP NOP BCF SER_OUT BIT4: NOP NOP BSF SER_OUT BTFSC _colorByte, 3 GOTO ONE4 BCF SER_OUT NOP NOP GOTO BIT5 ONE4: NOP NOP NOP BCF SER_OUT BIT5: NOP NOP BSF SER_OUT BTFSC _colorByte, 2 GOTO ONE5 BCF SER_OUT NOP NOP GOTO BIT6 ONE5: NOP NOP NOP BCF SER_OUT BIT6: NOP NOP BSF SER_OUT BTFSC _colorByte, 1 GOTO ONE6 BCF SER_OUT NOP NOP GOTO BIT7 ONE6: NOP NOP NOP BCF SER_OUT BIT7: NOP NOP BSF SER_OUT BTFSC _colorByte, 0 GOTO ONE7 BCF SER_OUT NOP NOP GOTO ENDLABEL2 ONE7: NOP NOP NOP BCF SER_OUT ENDLABEL2: return

The above code has two subroutines _sendByteASM and _sendByteASM2. The first one use a cycle to check and send the bits to the serial output pin (in this case RB4). The best timing I was able to achieve this way was "0" - 375ns/875ns, "1" - 875ns/500ns. And it worked.

The second subroutine check and send every bit separately and there I was able to achieve timing much closer to the required. "0" - 375ns/875ns, "1" - 875ns/375ns. 

Then I got a more modern PIC - PIC16F15344, which have 4 very interesting modules: Configurable Logic Cell (CLC) each of which can be set as 4-input AND, AND-OR, D-type flip flop, J-K flip flop and couple of other types. Also I saw a video from the great Ben Heck where he is using the SPI output from ESP32 with some external logical chips to form the output signal compatible with NeoPixel. 

My thought was to feed the color bytes to the SPI module (configured to work at 800 kHz) and the output (clock and data) to use somehow to form impulses with different length and then combine them with CLC. The following screenshots are the settings of different modules used in this project.

The CLC1 is configured as AND-OR and the signal from the SPI is directly routed to the output. This will be needed later. 

For creating the waveforms of "0" and "1" I used the Complementary Waveform Generator (CWG). This module is used to create a signal for driving half-bridge or full bridge circuits and among other setting there can be set a dead time. So I fed the signal from CLC1 (which is a copy of SCK signal and have 50% duty cycle or 600 ns high) to the CWG module and set the dead time of the rising edge to be about 400 ns and when inverted this will be the "0". The dead time of the falling edge is set to be around 200 ns increasing low time to 800 ns and when inverted form the "1" waveform.

CLC2 is configured as 4-input AND. There I combine the inverted output from CWG1A, CLC1 and SDO from SPI to create the "0" signal:

Finally, all is combined at CLC3 which is set as AND-OR cell:

Here the inverted signal from CWG1B is "AND"-ed with the SDO signal to produce the "1" signal. Then both "0" (from CLC2) and "1" are "OR"-ed to form the final output signal which is routed to one of the pins - in my case RC4/pin6. 

Here some scope screenshots:

The timing here is much better and the beauty of this solution is that there is no interrupts, no assembler code. All of the above is just setup of registers. I am using MPLAB Code Configurator to generate all the code and the actual work is done by the hardware modules and for sending a single byte to the NeoPixels are needed only 2 lines of code. Here is the function to send the 3 bytes for red, green and blue:

void sendPixel(NeoPixel p) { SSP1BUF = p.red; while (!SSP1STATbits.BF); SSP1BUF = p.green; while (!SSP1STATbits.BF); SSP1BUF = p.blue; while (!SSP1STATbits.BF); }

Bellow is a video demonstration how it work with the assembler code. I adapted some of Adafruit library functions for this demo: rainbow, their table for gama8 function and the function for HSV color. 

HV rescue shield for Microchip ATmega series 8-bit microcontrollers

These days I decided to play a little with ATmega328P microcontroller and it was a total disaster. I used a PICkit 4 programmer and MPLAB X + XC8 for the code writing and compiling. Then I came to the brilliant idea to use MCC (MPLAB Code Configurator) which supposedly is easy way to configure and use the different modules inside the chip. And after setting the wrong fuses for the clock source the chip was effectively bricked. It worked, but only if I put a watch crystal at 32.768 kHz as a clock source. And the programmer cannot communicate at that low frequency.

Why these chips are designed this way? The ICSP interface in PIC microcontrollers have a dedicated clock line, so the programming of the chip is independent of the his internal configuration. I start digging for solutions in the web and tried all sorts of things. Most recommended solution was to attach external clock source to XTAL1 pin. I have signal generator and tried this with different frequencies but without success.

Finally I found this excellent page ARDUINO-BASED AVR HIGH VOLTAGE PROGRAMMER and following the latest schematic I put the ATmega328Pchip on a breadboard and connected a million jumper wires to the Arduino. It worked and the chip was rescued. Of course I connected it again to the PICkit 4 and successfully bricked it again!

So in order to avoid dealing with jumper wires on the breadboard I get the original schematic, removed all other interfaces except the ATmega and replaced a dual PNP+NPN transistor array with discrete transistors. Also I removed the DC-DC voltage convertor because I can connect external voltage source for the 12V line. Here is the modified schematic:

And here is the finished shield for Arduino:

It worked perfectly, the chip was saved and I am ready for another bricking :)

The board was edited with EasyEDA. You can download the project files from HERE. Inside are included Gerber files. Use these on your own responsibility. The Arduino sketch for this project you can download from the original project page HERE.

Driving 4 digit 7-segments TM1637 display module with PIC microcontroller

TM1637 display modules are cheap chinese modules that are offered in different colors, with digital dots or with colon. Usually these are 4 digit, but there are 6 digit modules also. I bought mine for 1.84 USD delivered.