Path: Home => AVR overview => Applications => ATtiny crystals => Crystal divider   Diese Seite in Deutsch: Flag DE Logo
Tiny XTAL oscillator Applications of
AVR single chip controllers AT90S, ATtiny, ATmega and ATxmega
A crystal divider with ATtiny25

Crystal divider with ATtiny25

This describes a crystal-driven ATtiny25 that divides a crystal frequency by an adjustable constant. The divider can either be The drawings can be downloaded as LibreOffice-Draw file from here, the calculations as LibreOffice-Calc file from here.

Crystal divider with a mouse piano

The crystal divider's schematic with the mouse piano This is all you need:
  1. The four switches of the mouse piano select one of sixteen voltages, that are measured with the ADC1 input pin of the ATtiny25.
  2. The crystal X drives the ATtiny25, if its fuse is programmed for the external crystal.
  3. The frequency of the crystal is divided by the selected prescaler plus a compare value and a software divider and determines the output frequency.
  4. The output signal drives the PB0 pin, its inverted signal is on PB1.
  5. The controller can be programmed via the ISP6 pin connector.

Two-Dividers with the mouse piano

As an example, a 2.097152 MHz crystal is divided by factors between 1,048,576 and down to 32. This yields the output frequencies of 1 Hz, 2 Hz and, by a factor of 2N, up to 32,768 Hz. For this scheme see the sheet "TwoDivider" in the LibreOffice-Calc file here.

The possible dividers These are all possible dividers, by which the crystal frequency can be divided. The minimum possible divider is two, which would result in a frequency of 1,048,576 Hz with the given crystal. This is not exact, because the interrupt service routine would not function correct with a prescaler value of one, because each interrupt requires 22 clock cycles at a prescaler value of 1, which limits the minimum TC0 compare value to roughly 25 clock cycles. That yields a minimum divider of 50, and the maximum frequency is 41.9 kHz. If you need more than that: use a crystal with a higher frequency.

The maximum divider is 1024 * 256 * 256 * 2 = 134,217,728, which results in a frequency of 0.015625 Hz or 64 seconds per pulse.

These 16 different combinations of prescaler values, CTC compare values and software divider values with divider values of between 25 and 220 generate a table with the following content. Note that the dummy byte has been added to get any .db line to an even count, to fit it to the flash memory's word-wise structure.

CodeTable: ; 1=CS0x, 2=comp, 3=soft divider, 4=dummyDIP-Swf(eff) Hz
.db (1<<CS02)|(1<<CS00),255,4,001
.db (1<<CS02)|(1<<CS00),255,2,012
.db (1<<CS02)|(1<<CS00),255,1,024
.db (1<<CS02)|(1<<CS00),127,1,038
.db (1<<CS02)|(1<<CS00),63,1,0416
.db 1<<CS02,127,1,0532
.db 1<<CS02,63,1,0664
.db (1<<CS01)|(1<<CS00),127,1,07128
.db (1<<CS01)|(1<<CS00),63,1,08256
.db 1<<CS01,255,1,09512
.db 1<<CS01,127,1,0101,024
.db 1<<CS01,63,1,0112,048
.db 1<<CS00,255,1,0124,096
.db 1<<CS00,127,1,0138,192
.db 1<<CS00,63,1,01416,384
.db 1<<CS00,31,1,01532,768

Of course, you can equip the CodeTable with any other data to generate different frequencies. When changing, be aware that with a CS0x of 1 a minimum of 25 for the CTC compare should be respected, otherwise the device will not function correct. For CS0x=2 (prescaler value = 8) that minimum should be 3, other CS0x combinations are unlimited.

Getting the mouse piano's switch state

The calculation sheet "MousePiano" in the LibreOffice-Calc file here allows to play with the resistors of the mouse piano. Note that all ADC measurements are 10 bit and repeated 64 times to be added up to a 16-bit result.

As can be seen from the table, the upper resistor tolerance voltages are all below the lower tolerance voltages of the next combination of switches. If you change resistors and if tolerance areas overlap, the respective overlapping values in the plus-column turn red.

The conversion of ADC results to code table entries goes like this:
  1. Z points to the beginning of the ADC table entry below.
  2. X points to the first code table entry.
  3. Compare the ADC sum with the LSB and MSB of the ADC table. If carry is set (ADC sum smaller than ADC table entry), skip the whole conversion.
  4. If not, compare the ADC sum with the LSB and MSB of the next ADC table entry. If the carry is set (ADC sum is below the second entry), the correct entry has been found. Then read the three bytes that X points to (the TCCR0B, the OCR0A and the software divider) to their registers.
  5. If the carry is not set, add 4 to X and repeat from step 3 on.
  6. In any case, clear the ADC sum, set the ADC counter to 64 and start the next conversion.
The ADC table is as follows:

; Resistors: R0=4k7, R1=15k, R2=6k8, R3=3k3, R4=1k5, all 1.0%
  .dw 0,447 ; 0
  .dw 15360,15936 ; 1
  .dw 26432,27008 ; 2
  .dw 32512,33216 ; 3
  .dw 38144,38848 ; 4
  .dw 41280,41920 ; 5
  .dw 44160,44800 ; 6
  .dw 46144,46720 ; 7
  .dw 49408,49920 ; 8
  .dw 50560,51072 ; 9
  .dw 51712,52224 ; 10
  .dw 52544,52992 ; 11
  .dw 53504,53952 ; 12
  .dw 54144,54592 ; 13
  .dw 54848,55232 ; 14
  .dw 55360,55744 ; 15
  .dw 65535,65535 ; End of table

The last entry in the table stops further execution of the loop, because any sum value larger than 55,743 leads to an error condition. The maximum sum can be 1,023 * 64 = 65,472, if the ADC is at +5V. And comparision with 65,535 then, in any case, yields a carry.

A crystal divider for a center frequency +/- a deviation

Crystal divider with middle potentiometer Sometimes you need a frequency and a small band up and down of that. For that case a mouse piano would be less convenient, because all 16 CodeTable entries vary only by a CTC value difference of one or two down or up. And it is simpler to operate the device with a potentiometer rather than with a mouse piano.

The schematic shows how that goes: just the mouse piano exchanged with a potentiometer. The 1k decouples ADC1 in case of programming via the ISP6 interface, when USCK is used as clock pin.

As an example: I need a crystal generator with 77.5 kHz, that can be varied up and down around that frequency. First I have to identify the best crystal for task. The spreadsheet "crystals" provides a table of all commercially available crystals and their casing (either HC18 or HC49, the O is for complete oscillators, which are not needed here). You'll find that spreadsheet in the LibreOffice Calc file here.

If you enter your desired frequency in cell C7 you'll get all dividers that are necessary for the crystals to generate this frequency in column C. If too many "-" appear, you might have to change the prescaler value in cell C5 with the drop-down field values of 8, 64, 256 or 1,024, depending from your frequency. When the frequency is extremely low, then you can also increase the register divider value in cell E5 to up to 256.

Column D then lists the deviation from your desired frequency in Hz. Positive values mean that the frequency is higher than desired, negative mean the opposite. Column E calculates the minimum deviating values and displays this in cell E7, which in my case is 19.38 Hz difference. If you go down the list, you'll find the one(s) with the smallest difference high-lighted in green in column E. In my case the 20.0 MHz crystal had the smallest difference.

Frequency vs potentiometer degrees If you select this crystal in the drop-down field in cell B3, you'll get the divider value of that crystal in cell I3. You can now vary around that value: with a step width of 1 in cell I5 you'll get the 7 values smaller than that in the cell range I16:I22, and, above the 135 degrees or a DIP of 8, the seven values smaller than that in the cell range I7:I14. If you set the step width in cell I5 to 0.5, every second step leads to an increase/decrease, with the step width on 2 in that cell each step is by two, with 4 each step increases and decreases by four. The resulting frequencies are seen in column J. In my case, the frequency range so is from 73 to 82 kHz with roughly 600 Hz per step, as can be seen in the diagram.

Note that those divider values are only valid for values between 1 and 256 (smaller values are set to 1, larger values to 256) and that the prescaler and register divider value are the same for all those entries.

If you work with a prescaler of 1, the dividers with less than 30 have a red background. Below this value the division routine does not work correct. With a prescaler of 8, this minimum divider is 4.

Also note that the CTC compare value for OC0A is the divider value minus one. This is reflected in the values in column M and in the export data provided in the fields I25:I29 for export to the assembler source code (select this range, Ctrl-C to copy this and Ctrl-V to insert it into the source code).

To enable this mode, set cMode in the source code file to 1, re-assemble and burn the hex code to the flash of the ATtiny25. The constant clock is automatically set to 20,000,000 for the higher crystal which should be attached.

Software for the tn25-crystal divider

The software is available in assembler format as xtaldiv_tn25_v1.asm here. Please consult the lines in the section "Adjustable Constants" of this source code. You can select either the mouse piano version (cMode = 0) or the potentiometer version (cMode = 1). By default, the For these settings use the LibreOffice-Calc file here to calculate your own combination and paste the received source code to the .asm file. You can also change crystal frequencies for both modes in this section in the constant "clock" by selecting another crystal.

If you'd like to use the software in an ATtiny45 or ATtiny85 instead of an ATtiny25, just change the include All necessary changes are automaticly done by the assembler.

The program uses the sleep mode. In any case the controller sleeps for more than 99% of the time. Only very small divider values cause higher active time shares.

Fuse setting for 2 MHz crystal Fuse setting for 20 MHz crystal When programming do not forget to change the fuses of the ATtiny25 to work with the external crystal instead of the internal RC oscillator (which is default). And remove the DIV8 fuse, too. If you forget that, all your frequencies will not be correct. The 2 MHz crystal needs the fuse settings on the left, the 20 Mhz crystal the ones on the right.

Mounting the device And: yes, the tn25 is specified for and works fine with a 20 MHz crystal, as this build-up demonstrates.

©2022 by