Posts Tagged ‘cnc’

Using TMP100 temperature sensor with your PIC – PR5

January 31st, 2012 No comments

In this post I will describe how you can connect the TMP100 temperature sensor from TI to your PIC as well as receive temperature data. The sensor I am using is a generous sample from TI. Compared to the DS18S20 that I used before, this sensor is much faster, it takes 40ms to take a measurement with 0.5 degrees accuracy. The communication protocol is I2C. It could be a bit difficult connect it to your circuit since it is using a SOT-23-6 package.

For this project, I will be using my PIC prototyping board which carries a 18F4520. But you can pretty much use the C18 code in this post, with any PIC having a hardware I2C module.

Package problems!

Because SOT-23-6 is so small the first thing I needed was to create an adapter to make TMP100 compatible with my breadboard. This was a perfect excuse for improving my CNC skills :). So I created an adapter in EAGLE, cut it out using the CNC and then solder the sensor and 6 pins. The sensor is now breadboard ready!

The TMP100 adapter


The connection is really simple. Add two pull-up resistors to the SDA and SDL lines. I used 10kOhm. The datasheet also suggests an optional capacitor on the supply.

TMP100 pinout


TMP100 Schematic

As you can see from the schematic there are two pins called ADD1 and ADD0. Those pins determine the I2C address of the sensor. You can find more details in the datasheet. I connected those two lines to GND therefore the address of the sensor I am using is 0x90. That’s all on the hardware side.

TMP100 – Point Register

I2c is the language this sensor talks :). So we need to use your PICs I2C hardware capabilities. Initialize your communications using


This sensor has a Point Register (PR)which (guess what!) points to the address of the register you want  to read or write 🙂 Before doing anything else we should point PR to the temperature register. To do that we have to initiate an I2C communication, send the address of the sensor and tell that sensor whether we like to write or read from it. This is done using a single byte. You see, the address is 7-bits starting from the MSB and bit 0 indicates whether is read or write operation (0=Write 1=Read). Let’s do an example. Our sensor’s address is 0x90. We wish to read something from the sensor so the byte should be 0x91. If we wish to write something to the sensors it should be 0x90. That’s it.

Therefore to write to point PR to the temperature register we should issue the following

WriteI2C(0x90); // Call the sensor with WRITE
WriteI2C(0x00); // Write Temp reg to point register;

Notice the 0x00? That is the address of the temperature register. The sensor is now ready to give us some temperatures.

TMP100 – Read that temperature

To read the temperature off the sensor, we need to call it using its address and READ (what we said above, remember?) and the sensor will transmit two bytes of data back. We read that using the appropriate I2C commands shown below. We have to acknowledge the reception of the bytes to let TMP100 know that we are ready for the next byte. Temperature is read as follows

first_byte = ReadI2C();
second_byte = ReadI2C();

The temperature is made out of two bytes. If you don’t care about getting any decimals you can just use the first byte. The 4 most significant bits of the second byte contain information on the decimal value of the temperature.

TMP100 – C18 Library

I made a really simple library for tmp100 in C18. Just include the tmp100.h file to your project and make sure you initiate the I2C (the OpenI2C() command shown in this post) before calling any functions.


OK that is pretty much. You should be able to get some measurements out of that little sensor. If you need further help take a look at the example below

#pragma config OSC = HSPLL
#pragma config WDT = OFF
#pragma config PWRT = ON
#pragma config LVP = OFF
#pragma config PBADEN = OFF
#include <p18cxxx.h>
#include <i2c.h>
#include <usart.h>
#include <delays.h>
#include <p18f4520.h>
#include <stdio.h>
#include "tmp100.h"
void main(void) {
    unsigned int raw_tmpr;
    char str_tmpr[8];
              USART_RX_INT_OFF &
              USART_ASYNCH_MODE &
              USART_EIGHT_BIT &
              USART_CONT_RX &
    OpenI2C(MASTER,SLEW_OFF);   // Initialize I2C module
    tm_setconf(0x60);   // Set sensor to full sensitivity
        raw_tmpr = tm_gettemp();    // Get 2 bytes of temperature
        tm_tostr(raw_tmpr,str_tmpr);    // Convert the temperature to text
        printf("Temperature is %s\r",str_tmpr); // And output through USART

I connected the USART TX/RX line using my CP2102 and used the cutecom program on Ubuntu to read the output on my PC. The screenshot below shows the terminal

The terminal on Ubuntu

Incoming search terms:

  • tmp100 (45)

Path Optimizer for pcb-gcode

November 13th, 2011 8 comments

The last couple of days I used the CNC machine to create a PCB for a project I am working on. As the CNC was etching away the copper, I noticed that between paths (i.e. while the spindle was not cutting) there was a lot of unnecessary travel. This happened quite a few times during my 35 minute job. Because I am using slow feedrate, about 10in/min, that unnecessary travel must have added 2-3 minutes. So if you are here looking how to optimize gcode you are at the right place 🙂

So to solve this, we need some optimization. What a better thing to spend your afternoon than firing up python, calling a friend and colleague over your PC, and solve this problem. So me and Mike, decided to solve this the easiest and quickest way possible, just by using brute force method to combine the different paths together. The goal is of course to minimize the unnecessary travel of the spindle.

Since this is a python script, it is multi platform. I personally tested it on my Ubuntu and Windows XP.

Some results

Let me just show you a couple of screen shots of a job, run on a simulator. This is the PCB of my last project. First let’s see the screenshot before optimizing the code

Spindle path before optimization

The white line represents the travelling between two paths. As you can see there are a lot of those white lines! 🙂 Now let’s see the second screenshot which is the same PCB job, but this time with optimized gcode

The path after optimization

As you can see the white lines are reduced. Also notice the clock (below the STOP sign)  of the simulator. We just saved a few minutes!

The program

First of all we assume that you already installed python on your system. Download the optimizer script. There is no GUI yet (feel free to expand the code, use git) so we need to run it from command line. Just write

python your_file_here

The script will run and give you some stats on the performance of the optimization.

Screenshot from my terminal

As shown, in my PCB gcode, the optimized code reduced the travel time from 33 inches to 14. That is a decrease of about 60%. The optimized file will have the same filename with an additional .optimized at the end.

Please note that this was a one-afternoon project just for the fun of it. All comments and feedback are welcome. I hope you find it useful.

I would like to thank Mike as we developed this together.

You can find the project on GitHub

Incoming search terms:

  • cncopt py (78)
  • g code optimizer (34)
  • gcode optimizer (33)
Categories: CNC Tags:

How to flatten the CNC table

November 3rd, 2011 No comments

Hey guys, recently we got a medium-low end CNC machine in the lab I am working. We got it mainly for prototyping PCBs of our projects. I am still in the process of fine-tuning the machine. The learning curve was (is!) pretty steep and makes it even worst if you have little time to spend working the machine.

Anyway, when I printed my first PCB using 15 mils for copper width, I noticed some inconsistencies on the final result. One factor of this problem is the fact that the cutting table is not flat in relation to the XY axis of the machine. In other words, the distance between the table and the spindle is not constant along the plane.

This could happen for many reasons, your table may not be flat, the machine may not be exactly horizontally fixed to place etc. After asking around the web, the good fellas at the forum, suggested to flatten the table surface by milling into it. So what that means is that you pass the cutting tool up and down, cutting out a square area on your table. Since the cutting tool is always at the same depth, the square are should be completely flat with relation to your cutting tool. Genius isn’t?

A nice and flat square on the wooden table!

In order to do that, you need the appropriate g-code of course. I search around a bit and found a software for that, but for some reason (bugs?) it wasn’t working well for me. For that reason and also because I wanted to learn Java Applet programming in a day :), I created a simple tool to create that square area for you.

Head over to CNC Tools page to find my online tool for for a flattening the CNC table. The tool will load into your browser and after you input your data it will give you the g-code and the estimated time to finish the job. You can test the generated g-code on a CNC simulator.

Let me know if you have any questions, suggestions or bug to report.

Categories: CNC Tags: