How to Double Clock Frequency Using Only Digital Logic

I recently rotated hobbies back to microcontrollers and electrical engineering, and one of the projects I’ve been working on required a doubling of clock speed from an Arduino Nano, which has an Atmel 328p running at 16 MHz on-board (here’s a hint: my target was 25.175 MHz).

The challenge?  Well, I wanted to do it only with parts on hand, and preferably with digital logic – I do have plenty of analog components in stock at DQYDJ headquarters, but a hack is a hack.  And, no, overclocking the Arduino or changing the clock in general wasn’t a path I was willing to go down.

Plus – where’s the fun in that?

If you want a real answer for a product that you’ll be shipping (or you want to do something non-sloppy where fellow EEs won’t look down on you), use a PLL.  If you want the glory of using 7400 series gates in ways they weren’t built to be used?  You’re reading the right article!

How Did You Double the Frequency?

First, not that it matters for the project (but in the circuit section you certainly will care about how I tuned it for 32 MHz), but I first set the fuses on the 328p to produce a buffered clock on pin 8 (labeled D8 on the Uno/Nano – search ‘CKOUT’ in the 328P data sheet for more).  Now, I mentioned above that I didn’t want to use analog components – but, we exist in reality.  “Analog” effects dictate everything in engineering, and digital logic is no different, however much it wants to pretend life is all 1s and 0s.  My goal was, therefore, to produce a second clock 90 degrees out of phase with the original using the delay of digital components… then XOR the two clocks together.

For the uninitiated, XOR, or exclusive or, is a function which will return ‘true’ when and odd number of inputs are ‘true’.  For a two input XOR function, that means only one will be high.  Don’t worry, I drew it in my engineer pencil-scratch (give me a break, I type for a living):

xor creating clocks

So, the clock will be high when only one of the two out of phase clocks is high… which happens to be twice as often as the original clock.  And that’s the theory, which I’ve conveniently drawn for you here:

Yeah, It's a Diode Symbol.  I Couldn't Find the Delay.

Yeah, It’s a Diode Symbol. I Couldn’t Find the Delay.

How Did it End Up in Reality?ugly breadboard mess

Theory is great, and if you just wanted the refresher turn back now.  The actual implementation details were particularly hilarious because while I did have a not gate – the perfect gate to add delay since you can force it to make low to high and high to low transitions in the same clock – I didn’t have an XOR!

Well, you can always get around things like that – (A AND !B) OR (B AND !A) did the trick.  I also needed a very fast gate to handle the 32 MHz signal – which is approaching the limits for the 74LS parts I mostly had handy.  So, here’s my actual parts list:

  • HD74LS04P – Not Gate.  Delay
  • HD74LS08P – And Gate.  XOR
  • HD74LS32P – Or Gate.  XOR
  • SN74F10N – 3 Input Nand Gates.  Clean up the ugly clock from the hacked XOR

Here’s the schematic, drawn to the best of my ability.  Note: I can’t take responsibility if you break something by using this information, somehow.  Besides, your mileage will almost certainly vary, as actual time delays will vary from part to part.  This is the combination which worked best to get a usable clock out of the 74LS32 for me, which I then cleaned up with the very fast (in comparison) 74F10N.  Note that I’m using the 74F10N as a not gate – merely tying unused inputs to high.

The Clock Pins look like inputs - they aren't.

The Clock Pins look like inputs – they aren’t.  Click to zoom.

And, because I know you’re interested in seeing that rats nest wired up, you should find that over to the right somewhere.

And Your Results?  Did It Produce a Beautiful 32 MHz Square Wave?

No, quite ugly in fact.  Of course, it was good enough for our purposes.  The duty cycle moves between 50% and 60% high, but the project it went into appears stable even with the sloppy clock.  Here’s a picture of my oscilloscope for internet points and some hard proof that this scheme worked:

Bravo; good show.

Bravo; good show.  (Click to zoom)

 Further Reading

If you just need a usable clock, this will work.  Don’t use it in any sort of mission critical application – I’ll be replacing this with a real circuit later.  For proof of concept, it worked well.

For budding EEs, read about what others have said about the digital logic frequency doubler:

So, don’t use it at work, you’ll be laughed at.  Don’t use it in a product, you’ll likely face reliability issues – and will pull your hair out when you need to substitute parts.

But, hey, when you’re too lazy to use the right part/build the right circuit… it works in a pinch.  But we’ll get to that in the next article on the results of all of this ridiculousness.

Looking for other DQYDJ electronics posts?  There’s only one – way back from 2012, FFT on an Arduino.



  1. […] top out at 20 MHz. [PK] wanted to generate VGA signals without overclocking, He did this by doubling the clock frequency with digital logic. The ATMega generates a clock, an inverter delays that clock so it is 90 degrees out of phase, and […]