Oscillating Options

Every microcontroller needs a clock source. The CPU, the memory bus, the peripherals—clock signals are everywhere inside a microcontroller. They govern the speed at which the processor executes instructions, the baud rate of serial-communication signals, the amount of time needed to perform an analog-to-digital conversion, and so much more.

All of this clocking action goes back to the source of the clock signal, namely the oscillator. Therefore, you need to make sure that your oscillator can support whatever performance you expect from your microcontroller. At the same time, though, some oscillator options are more complex or expensive than others, so your choice of oscillator should also reflect the importance of reducing cost and complexity whenever possible.

There are quite a few ways to generate a clock signal for a microcontroller. The datasheet for your particular device should provide a good deal of information about what types of oscillators you can use and how to implement them in a way that is compatible with the device’s hardware. This article will focus on the advantages and disadvantages of various clock sources so that you can better choose among the oscillator options discussed in your microcontroller’s datasheet.

So let’s start with a list, followed by a discussion of each option:

• Internal
  • Usually (as far as I know, always) a resistor–capacitor circuit
  • Phase-locked loop for frequency multiplication
• External
  • CMOS clock
  • Crystal
  • Ceramic resonator
  • Resistor–capacitor
  • Capacitor only

Internal Oscillators: The KIS Option

I’m an advocate of the Keep It Simple principle; consequently, I have a special appreciation for internal oscillators, and I encourage you to avail yourself of the internal oscillator whenever possible. No external components are required: You can safely assume that the frequency is well chosen since the oscillator was designed by the same people who designed the rest of the microcontroller. Also, the salient performance specs—e.g., initial accuracy, duty cycle, temperature dependency—are (hopefully) right there in the datasheet.

The dominant disadvantage with internal oscillators is the lack of precision and frequency stability. The baseline frequency depends on the values of the passive components that make up the oscillator circuit, and the tolerances for the values of these passive components are not particularly tight. Furthermore, capacitance and resistance are influenced by ambient temperature, so internal RC oscillators experience temperature “drift”—i.e., changes in temperature lead to changes in frequency.

In my experience, many applications can tolerate the shortcomings of an internal oscillator, especially when the frequency has been calibrated at the factory. With older microcontrollers, the internal oscillator might have tolerance as bad as ±20%. However, a newer device can give you ±1.5% (or better), which is accurate enough for RS-232 communication and even (in conjunction with clock-recovery circuitry) for USB.

Another way to expand the capabilities of an internal oscillator is manual “trimming”—if your microcontroller includes a trimming/calibration register, you can adjust the frequency by modifying the value in this register. This is a perfectly practical technique for low-quantity designs: Simply measure the clock frequency with an oscilloscope or frequency counter and then trim the oscillator accordingly.

A variation on the internal-oscillator theme is the phase-locked loop (PLL). A PLL allows a low-quality, high-speed internal oscillator to benefit from the stability and precision of an external oscillator. In general, a PLL doesn’t help you to avoid external components because it requires a reference clock that is usually derived from a crystal. An exception, though, is when you have a high-quality clock somewhere on the PCB but don’t want to use it for the microcontroller because it’s too slow—you could use a PLL to multiply this clock up to an acceptable frequency.

CMOS Clock

Another straightforward clocking option is the so-called “CMOS clock,” which falls into the for-lack-of-a-better-term category. “CMOS clock” is a vague (though convenient) way of referring to any clock signal driven by some other component on the board. The CMOS clock is a great option if your design already includes a clock signal with 1) a workable frequency and 2) electrical characteristics that are compatible with the microcontroller’s CMOS-clock-input circuitry. Often, though, this is not the case, so let’s look at two options for generating a CMOS clock.

First is the “crystal oscillator.” This is a good time to point out that a quartz crystal is not an oscillator; rather, a quartz crystal is the central component in a quartz-crystal oscillator circuit.

Crystal oscillators are handy devices that consist of a quartz crystal and the additional circuitry needed to generate a standard digital clock signal. Thus, you get the stability and precision of a crystal without worrying about load capacitance and the careful PCB layout needed to ensure robust operation with a standalone crystal.

The second option is a “silicon oscillator”. This term refers to oscillator ICs that are not based on quartz crystals or ceramic resonators. These devices are versatile and easy to use, and they can be quite accurate. For example, the LTC6930 series from Linear Tech requires only one bypass capacitor, and the vast majority of the parts are within .05% of the nominal frequency: Silicon oscillators are more reliable than crystals and ceramic resonators, especially in harsh environments subject to shock or vibration. But they’re also more expensive.

Quartz and Ceramic

When you need seriously high precision and stability without the additional cost of a crystal-based oscillator IC, opt for the standalone-crystal approach. Parts with tolerance below 20 parts per million (i.e., 0.002%) are readily available. The oscillator circuit shown above is partially integrated into microcontrollers that support the standalone-crystal configuration; you will need to provide the correct load capacitors. The total load capacitance (CLTOTAL) is specified in the crystal’s datasheet, and the load capacitors are chosen as follows:

CLTOTAL=CL1×CL2CL1+CL2+CPCLTOTAL=CL1×CL2CL1+CL2+CP

where CP represents whatever parasitic capacitance is present. This calculation is pretty simple in practice: Choose a reasonable value for CP (say, 5 pF), subtract this from CLTOTAL, then multiply by two. So if the datasheet gives a load capacitance of 18 pF, we have

CL1=CL2=(18 pF−5 pF)×2=26 pFCL1=CL2=(18 pF−5 pF)×2=26 pF

Ceramic resonators are less accurate than crystals; common tolerances are 1000 to 5000 parts per million. They may save you a few cents if you don’t need the accuracy of quartz, but in my mind, the main advantage is that you can get ceramic resonators with integrated load capacitors.

Last, and Least . . .

There aren’t many situations that call for an external resistor–capacitor or capacitor-only oscillator. If for some reason you are opposed to the external-oscillator options discussed thus far, choose a microcontroller with an internal oscillator and use that. If, however, you are determined to dig out a passive or two from your box of spare parts, refer to the microcontroller’s datasheet for guidance on how to connect and design the oscillator circuit. Here is an example of how to connect the components, taken from the datasheet for the C8051F12x–13x (PDF) microcontrollers from Silicon Labs:

And you can refer to page 190 (PDF) of this same datasheet for an example of information on choosing component values.

I hope you now know enough to make an informed, confident decision next time you need to choose an oscillator for your microcontroller. Here are my recommendations in a nutshell:

• Internal oscillator whenever possible
• Silicon oscillator if the accuracy is adequate and the cost is acceptable—otherwise, quartz crystal

If you don't understand the reasons for having a clock circuit, you're probably not qualified to be designing mircrocontroller PCBs.

The clock circuit determines the speed with which the microcontroller operates. How many instructions per second it will execute is dependent on this.

Oscillating Options

Every microcontroller needs a clock source. The CPU, the memory bus, the peripherals—clock signals are everywhere inside a microcontroller. They govern the speed at which the processor executes instructions, the baud rate of serial-communication signals, the amount of time needed to perform an analog-to-digital conversion, and so much more.

All of this clocking action goes back to the source of the clock signal, namely the oscillator. Therefore, you need to make sure that your oscillator can support whatever performance you expect from your microcontroller. At the same time, though, some oscillator options are more complex or expensive than others, so your choice of oscillator should also reflect the importance of reducing cost and complexity whenever possible.

There are quite a few ways to generate a clock signal for a microcontroller. The datasheet for your particular device should provide a good deal of information about what types of oscillators you can use and how to implement them in a way that is compatible with the device’s hardware. This article will focus on the advantages and disadvantages of various clock sources so that you can better choose among the oscillator options discussed in your microcontroller’s datasheet.

So let’s start with a list, followed by a discussion of each option:

• Internal
  • Usually (as far as I know, always) a resistor–capacitor circuit
  • Phase-locked loop for frequency multiplication
• External
  • CMOS clock
  • Crystal
  • Ceramic resonator
  • Resistor–capacitor
  • Capacitor only

Internal Oscillators: The KIS Option

I’m an advocate of the Keep It Simple principle; consequently, I have a special appreciation for internal oscillators, and I encourage you to avail yourself of the internal oscillator whenever possible. No external components are required: You can safely assume that the frequency is well chosen since the oscillator was designed by the same people who designed the rest of the microcontroller. Also, the salient performance specs—e.g., initial accuracy, duty cycle, temperature dependency—are (hopefully) right there in the datasheet.

The dominant disadvantage with internal oscillators is the lack of precision and frequency stability. The baseline frequency depends on the values of the passive components that make up the oscillator circuit, and the tolerances for the values of these passive components are not particularly tight. Furthermore, capacitance and resistance are influenced by ambient temperature, so internal RC oscillators experience temperature “drift”—i.e., changes in temperature lead to changes in frequency.

In my experience, many applications can tolerate the shortcomings of an internal oscillator, especially when the frequency has been calibrated at the factory. With older microcontrollers, the internal oscillator might have tolerance as bad as ±20%. However, a newer device can give you ±1.5% (or better), which is accurate enough for RS-232 communication and even (in conjunction with clock-recovery circuitry) for USB.

Another way to expand the capabilities of an internal oscillator is manual “trimming”—if your microcontroller includes a trimming/calibration register, you can adjust the frequency by modifying the value in this register. This is a perfectly practical technique for low-quantity designs: Simply measure the clock frequency with an oscilloscope or frequency counter and then trim the oscillator accordingly.

A variation on the internal-oscillator theme is the phase-locked loop (PLL). A PLL allows a low-quality, high-speed internal oscillator to benefit from the stability and precision of an external oscillator. In general, a PLL doesn’t help you to avoid external components because it requires a reference clock that is usually derived from a crystal. An exception, though, is when you have a high-quality clock somewhere on the PCB but don’t want to use it for the microcontroller because it’s too slow—you could use a PLL to multiply this clock up to an acceptable frequency.

CMOS Clock

Another straightforward clocking option is the so-called “CMOS clock,” which falls into the for-lack-of-a-better-term category. “CMOS clock” is a vague (though convenient) way of referring to any clock signal driven by some other component on the board. The CMOS clock is a great option if your design already includes a clock signal with 1) a workable frequency and 2) electrical characteristics that are compatible with the microcontroller’s CMOS-clock-input circuitry. Often, though, this is not the case, so let’s look at two options for generating a CMOS clock.

First is the “crystal oscillator.” This is a good time to point out that a quartz crystal is not an oscillator; rather, a quartz crystal is the central component in a quartz-crystal oscillator circuit.

Crystal oscillators are handy devices that consist of a quartz crystal and the additional circuitry needed to generate a standard digital clock signal. Thus, you get the stability and precision of a crystal without worrying about load capacitance and the careful PCB layout needed to ensure robust operation with a standalone crystal.

The second option is a “silicon oscillator”. This term refers to oscillator ICs that are not based on quartz crystals or ceramic resonators. These devices are versatile and easy to use, and they can be quite accurate. For example, the LTC6930 series from Linear Tech requires only one bypass capacitor, and the vast majority of the parts are within .05% of the nominal frequency: Silicon oscillators are more reliable than crystals and ceramic resonators, especially in harsh environments subject to shock or vibration. But they’re also more expensive.

Quartz and Ceramic

When you need seriously high precision and stability without the additional cost of a crystal-based oscillator IC, opt for the standalone-crystal approach. Parts with tolerance below 20 parts per million (i.e., 0.002%) are readily available. The oscillator circuit shown above is partially integrated into microcontrollers that support the standalone-crystal configuration; you will need to provide the correct load capacitors. The total load capacitance (CLTOTAL) is specified in the crystal’s datasheet, and the load capacitors are chosen as follows:

CLTOTAL=CL1×CL2CL1+CL2+CPCLTOTAL=CL1×CL2CL1+CL2+CP

where CP represents whatever parasitic capacitance is present. This calculation is pretty simple in practice: Choose a reasonable value for CP (say, 5 pF), subtract this from CLTOTAL, then multiply by two. So if the datasheet gives a load capacitance of 18 pF, we have

CL1=CL2=(18 pF−5 pF)×2=26 pFCL1=CL2=(18 pF−5 pF)×2=26 pF

Ceramic resonators are less accurate than crystals; common tolerances are 1000 to 5000 parts per million. They may save you a few cents if you don’t need the accuracy of quartz, but in my mind, the main advantage is that you can get ceramic resonators with integrated load capacitors.

Last, and Least . . .

There aren’t many situations that call for an external resistor–capacitor or capacitor-only oscillator. If for some reason you are opposed to the external-oscillator options discussed thus far, choose a microcontroller with an internal oscillator and use that. If, however, you are determined to dig out a passive or two from your box of spare parts, refer to the microcontroller’s datasheet for guidance on how to connect and design the oscillator circuit. Here is an example of how to connect the components, taken from the datasheet for the C8051F12x–13x (PDF) microcontrollers from Silicon Labs:

And you can refer to page 190 (PDF) of this same datasheet for an example of information on choosing component values.

I hope you now know enough to make an informed, confident decision next time you need to choose an oscillator for your microcontroller. Here are my recommendations in a nutshell:

• Internal oscillator whenever possible
• Silicon oscillator if the accuracy is adequate and the cost is acceptable—otherwise, quartz crystal

Thank you, I appreciate it, Mr. Kakkar, I was wondering about the crystal circuit :)

Thank you Mr. Beckert for your advice and explanation.

Thank you very much Mr. Gwade! I started my curiosity after briefly looked over the old micro controller's specifications, and I started wondering about the usage of external clock circuits. Your explanations, and advice greatly helped me! :D Thank you again Mr. Gwade.

Before starting to design your own microcontroller board I suggest you to purchase an evaluation board to get familiar with the particular microcontroller hardware and software development tool. For virtually every microcontroller family there number of evaluation boards available, e.g. http://www.microchip.com/developmenttools/productdetails.aspx?partno=atxmegaa3bu-xpld

Regarding clock circuits, most of the time you can use microcontroller's internal oscillator, especially for the beginner's project.

- A microcontroller needs a clock for its internal operation. To make it universal and flexible, clock circuit (RC circuit) or Quartz Crystal (for high frequency stability) is configured outside the microcontroller. Clock circuit is essential for sequencing internal operations.

- A microcontroller is a single chip computer with internal memory, input/output and associated circuits. No external hardware is required for a small application.

- If you are new to this work, it is advised to purchase a low cost microcontroller evaluation board and experiment on it. You also need to acquire hardware knowledge to design your own board. Later you can plan your own microcontroller board.

I think it is very basic concept of the microprocessor. As the main unit of a microcontroller is a processor. Therefore, it requires clock signal to execute the instructions you provide. Many microcontrollers have their internal oscillator. You can use that oscillator. Due to area constraints the internal oscillators are mostly the RC circuit. Hence, to achieve better performance an external Crystal based oscillator is used. You can add two very small sized capacitors in parallel with it to avoid the external noise as well.

Please disable master clear reset before using microcontroller. I was using PIC microcontroller mostly and I used separate 10K resistor (in VDD) and 0.1 uF capacitor (GND) in the master clear reset pin to disable it.

To start designing a PCB first check your circuit in the Breadboard. Then make a rough layout in a paper which will help you to position your components. Then you can start designing your PCB. You can find many tutorial videos (in youtube also) to learn the PCB designing.

The clock is rather like a drumbeat that tells everything when to act. A microprocessor is an example of a finite state machine. Each time the clock signal transitions, it causes flip flops and other circuit elements to change, advancing the machine to the next state. The consequences take a small amount of time to propagate through the circuitry before everything settles down to the new state. This delay sets the maximum clock rate.

As noted above, the internal clock rate that actually controls the execution speed is often derived from another frequency source such as a crystal oscillator. PLLs are one way to multiply a relatively slow external frequency into a much faster internal one. This is simpler (and cheaper) than implementing high frequency external clocks.

One practical aspect of selecting clock frequency is how fast do you actually need the processor to run? Faster isnt necessarily better. I have seen cases where the processor was runing much faster than the frequency of update of external sources of inputs it used in its control logic. Speed translates to heat which can reduce reliability depending on available cooling etc. In one case thermal imaging showed the processor practically glowing, and we were able to reduce the clock speed to improve reliability/extend life. When what is controlled is an essential or safety related function this is an important consideration.

All microcontrollers require clocks. A clock doesn't mean the time of day, a clock means a heartbeat ( square wave) by which all operations in the processor are performed . Let me know which processor you want to use and perhaps I can advise on an appropriate clock circuit.

Hi Fa Shin,

I see you have had a lot of help about the microcontroller clock circuits. Indeed most microcontrollers have build in clocks, and you might only need to have an external one to achieve a specific frequency or accuracy.

With regards to PCB design it can be a bit of a steep learning curve. As someone mentioned earlier, you may want to give one of the Arduino boards a try first. You can quickly grow out of them, but you need to feel comfortable with getting them to work first, as, when it comes to your own PCB, you need to KNOW that your "firmware" should work, you can program it ok and that it's the PCB that is at fault!

After that you need to have a play with some PCB design software, I have used Eagle myself for many years, however the licensing has changed now and many users are now switching to KiCad.

Next, choose your controller. Personally im a great fan of the AVR8 and particularly the atmega 48 88 168 series of chips. They are used a lot on the arduino boards, they have plenty of memory, (built in 8MHz clock), and can be used with the free AVR studio software for writing and compiling the firmware C code, and there is a lot of support out there based on these chips. The data sheets are also pretty good, and have some sample code in them if you get really stuck. Though again it can be a bit of a challenge getting into them at first! The bonus with these chips is that they also have the option of switching the chip to "one wire debug mode" which means you can free up most of the input / output ports, otherwise used for programming and debugging.

For the PCB the things to make sure you have are; the power supply to the controller AND the programming pins (you will need to buy a programmer from the net usually something like a JTAG programmer, the higher end ones are "JTAG ICE" for example.) You could supply power from a 5V USB type adapter to keep things easy, or go for a simple linear voltage regulator (LM7805).

Have a look on the net at a few circuit diagrams for your microcontroller, and check you have all the basic essentials for programming and powering the chips. (Crucially for the ATMEGA you need: Reset, MISO, MOSI, SCK (-the ISP programming connections) plus the power connections - but thats it!)

Next you need a purpose! Think of a fun simple project, keep it simple, you dont have to put EVERYTHING on the board. A few LEDs are very good to have however, as they allow you a very simple debugging tool. You can switch LEDs off and on in your code, and know when that part of the code has worked ok. Maybe try a 7 segment display, or an LCD? N-Type mosfets are great for switching high power, higher voltage devices (you switch the 0v line of the device).

Once you have drawn up a schematic with your PCB software, you then arrange your board, and place track to connect the components, for a simple circuit there is nothing critical at this stage, and a 2 layer board is fine (and will cost the same as a 1 layer board!). You "sew" one track under another with "vias" where different signal tracks would other wise cross. Generally you keep tracks on the top layer traveling at right angles to tracks on the bottom layer, this keeps things simple, though the "auto route" function will help you do all this, though it does often make things look a mess! You can use the minimum hole size your PCB manufacturer says, eg. 0.3mm and track widths or say 10 mil (10/1000 of an inch!) which will cope with up to about 1amp of power.

You then have to export the PCB to a "gerber" format you can usually find someone else's basic file to run in your PCB software to do this. It basically converts your PCB into a set of files that allows a PCB manufacturer to make your boards - it contains masks for top and bottom, labeling, drilling info. etc. Then send off your PCB, PCBway.com is very cheap for this, and really is actually an argument for always making your own boards rather than buying the modules such as arduino!

Then solder away! With ICs getting rather tiny nowadays, it can be very hard to use the old stick type of solder, you can just about get away with it still with the ATMEGA chips and lots of flux, but you can get blobs of solder bridging pins. Some of the "solder wick" is good for removing this, especially when the area is whetted with some additional flux. Solder with some silver in it also helps with the surface mounting components, and my preference is to actually use "solder paste" and a dispensing syringe, as the solder paste tends to flow much better and avoids the trouble with bridging (most of the time!). For the ATMEGA you would place the chip, put a blob of paste on opposite corner pins, fix them with the iron, then lay a thin bead of solder past along the edge of the other pins, then stroke the iron slowly along the pins, and it should make a great job, and in this case you can actually use a reasonable large soldering iron tip.

After that, cross your fingers and power up, hope nothing goes bang, and get programming!

A bit of a crash course but I hope that was some help to you, and there is plenty of information out of the web to pad out the various aspects I mentioned! If you have any questions, feel free to message and I can try to help out!

best of luck!

Phil

a microcontroller is look like the human body and have inside it

a microprocessor which look like the human brain

and RAM & ROM which look like the human memory

and Input/output Interface circuit which look like the arms & legs of human

the clock is the Heart of the microcontroller which gives pulses for the brain and memory and arms & legs to work perfectly like the human heart gives beat push blood and make all human parts works

The microcontroller is a digital system composed of digital circuits. There are two types of digital circuits, the combinational circuits and the sequential circuits. The sequential circuits such as registers and memory needs clock for their operation.

In order to control the flow of data between the different building blocks of the microcontrller one needs a clock. So, the clock is needed to manage the operation of the microcontroller. The clock is needed also for the streaming the data across the communication ports.

The frequency of the clock must be highly stable to decrease its jitter for more precise timing. This is accomplished by using crystal controlled oscillator

Best wishes