board running at 16MHz with 3.3V ?


I've looked at the many arduino boards, but i always see boards running at 3.3V with clock at 8Mhz or boards running at 5V with clock at 16Mhz

Do you know if it's possible to have a board with 3.3V / clock 16MHz ?

The fact is i have devices that use 3.3V, so this voltage perfectly fits my need, but I also use NewSoftSerial which requires clock at 16Mhz, and the developer of this lib says it's hardly impossible to make it work with a 8 MHz clock...

Any idea about such a board 3.3V / 16 MHz ?


The Seeeduino board runs @ 16MHz with 3.3V

Cool !

And does it exist a mini seeduino equivalent to the arduino Pro mini ?

Not yet, but I am sure Seeedstudio is reading your request right now... Or you could take the Arduino Pro mini 5V - 16MHz and supply 3.3V, like Seeeduino does (bypassing the integrated voltage regulator)

Ohh if Seeedstudio wants to know my need, here it is :)

And i think my need is the need of many persons:

Aim is to have the smallest piece of arduino compatible (like the Pro mini) with clock at 16MHz to have NewSoftSerial, with the smallest power needed (3.3V) and the smallest piece of xbee connected to it easily

Or even better, to have the smallest piece of arduino capable of 2 uarts, 1 isn't really enough

Some kind of a sanguino mini pro in fact (644p is quite interesting compared to a 328p)

And of course with having the lowest energy consumption in order, in the long run, to have a wireless small piece that can send small amount of data from time to time for years

By the way I've seen in your shop, it's surely what i'm looking for... But this shield needs 3.3V or 5V ?

Because i've bought Xbee module from libellium but this shield is sooo big, I really don't understand why... (it connects on 5V provided by ICSP)


Operating the chip at 16 MHz with 3.3 volts is beyond the specified maximum frequency “Safe Operating Area” for the mega168 chip.

Please take a look at the charts in section 26.3, figure 26-2 “Maximum Frequency vs. Vcc, ATmega48/88/168”, which appears on page 304 of the latest version of the datasheet:

There is a linear line between 10 MHz at 2.7 volts, and 20 MHz at 4.5 volts, which is a slope of 5.56 MHz/volt. So at 3.3 volts, the highest frequency within the “Safe Operating Area” would be 13.34 MHz. At that slope, 3.78 volts it the point where 16 MHz is within the “Safe Operating Area”.

At 3.3 volts, 16 MHz is essentially overclocking the chip by 20%. Many chips probably do run just fine that way, at room temperature, and if the 3.3 volts is accurately regulated (likewise, many PCs run fine when substantially overclocked). But it is still running the chip beyond the maximum “Safe Operating Area” specified in the datasheet. A good engineering practice would be to consider the tolerance of the voltage regulator and use a clock frequency within the safe operating area at the lowest possible voltage the regulator might output.

Perhaps Seeeduino rigorously tests the overclocking ability of every board? Somehow I doubt that.

Thanks Paul for those technical information !

In fact, I was wrong in my summary, I've told never NewSoftSerial would support 8Mhz, which is completely false.

This feature is possible and is in the roadmap of the team

I've been mistaken with another thing the lib shouldn't be able to do, that is support for asynchonous multiple objetcs More info here :

So as a conclusion, be patient and use NewSoftSerial with arduino pro mini 3.3V / 8MHz

Now looking for the smallest xbee shield working with 3.3V...


@enzo, if you need a small footprint for XBee + Arduino, then take a look at the Funnel IO

It runs at 8MHz, but maybe form factor is more important than speed? Are you sure you need 16MHz?

Ah yes, very interesting this board !

I need now 16MHz in order to be able to use NewSoftSerial
But as I said previously, this lib should work at 8MHz in a near future

I’m just waiting for such a good feature

Thanks for the link NKC

Read this quote from this thread (2nd page, 5th post):

FWIW, I've been running the ATmega168 at 3.3V with a 16 MHz resonator on about a dozen units with no ill effects whatsoever. Same reason as crashingdutchman, i.e. to connect a radio module which doesn't tolerate 5V.

So far, all my boards work just fine, at room temperature anyway... One of the units has been running continuously for over two months now.

Lots of people overclock the PCs too, without any problems. Many have done the mods in the winter and had them work flawlessly until the summer.

Running for 2 months at 3.3 volts regulated and room temperature wouldn't be nearly as good a test as running at 3.0 volts and 50 deg C for even 15 minutes. If you can do that, then you know your chip has some margin. If it fails the moment you go under 3.2 volts or raise the temperature a bit, then you know you're running right near the limit.

EDIT - a recent comment says they are running outside of the envelope, so I guess it is not uncommon (but I will have to try it next summer around Yuma AZ).

Sparkfun has a new "OpenLog", and according to them (in the comments on the product page), there has been some kind of stepping or revision that allows 3.3v and 16MHz. I can't find any confirmation of this. (Of course Atmel is sometimes slow - Digikey has TSSOP8 ATtiny45s but Atmel doesn't show anything on their site).

I've been running the ATmega168 at 3.3V with a 16 MHz resonator on about a dozen units with no ill effects whatsoever.

So the manufacturers who design this system deliberately say it will only run at a slower speed for what advantage?

Precisely none.

If a device is capable of reliably running faster then it would say so in the data sheet and the manufactures would have a better device. It can't, so they don't say it can. Any one who thinks he knows better that the manufacturer and designer of a device is fooling himself and possibly other gullible people. Get a grip.

I should get some liquid nitrogen and a 32Mhz resonator...

Chips tend to age and have problems at higher temperatures. A particular batch that is tested is only guaranteed to work inside the envelope - but there may be many "golden" units that can operate far outside.

If I remember right, the big CPUs are just rated - if it works at 2Ghz, it gets a 2Ghz label (the test conditions are more strenuous to guarantee it for the life of the chip). If it doesn't, they test at 1.8, then 1.6 and so on.

You want your car and its electronics like the radio to work when you start it in the hot sun. Most are rated to 85C or even higher (and down to -40 - note that around -35C many electrolytics freeze and lose their capacitance - another problem).

It depends what it is for. If it is not going to be outside of room temperatures it will probably work. But I wouldn't count on it for "general purpose" because that can be under much different conditions.

The Jeenodes are basically using the atmega328 3.3V at 16MHz. I have used the nodes in a sensor network for my home and I didn't see any loss in performance or data. I also have an outdoor temperature sensor and didn't give me any problem (yet). As the author says:

Using an ATmega168/328 at 16 MHz with 3.3V exceeds the specs by some 2.7 MHz - but it works just fine.

As the author says:

The author is a complete idiot.

What he is saying that the few he has tried work so far. I am so not impressed with this sort of shabby approach. But you might be.

The author is a complete idiot.

that's not very nice or scientific in fact it just your opinion

I have been doing overclocking of pc's for years and with a regulated environment you can exceed the specification that the manufactures provide as they are just a guide anyways

my final year dissertation was on "practical overclocking to increase processing performance" intel said chip runs @ 2.66ghz I proved it is still running @ 3.8ghz almost 2 years on, so am I an idiot too?

results from exceeding specification vary with silicon batch as every batch of silicone has some differences but all work within a set tolerance just like passive devices +- x% Like anything else in this world, crystals inherently possess imperfections, or what we often refer to as 'crystalline defects'. The presence of most of these crystalline defects is undesirable in silicon wafers, although certain types of 'defects' are essential in semiconductor manufacturing. Engineers in the semiconductor industry must be aware of, if not knowledgeable on, the various types of silicon crystal defects, since these defects can affect various aspects of semiconductor manufacturing - from production yields to product reliability.

atmel do sell a 3.3v part and why would they tell anyone to buy a part and under/over clock it when they can sell you one designed for xyz purpose, its cheaper for large production to make lots of one thing and just change the label thus more profit.

if this was an MOD product all part would have better high/low temp tolerances and better voltage range and cost that extra ££ more

If I was atmel I would do the same make 1 datasheet with very conservative figures to allow for manufacturing defects and as AMD have shown with the TLB bug one error can lead to end of a product, memory manufactures use a testing process called "speed binning" to determine which chip gets the faster or slower product code within a product family.

the only way to see if your batch of atmel chips will run at xy frequency is to test it and look for errors in the output of your function, no datasheet will be able to tell how that bit of silicone will react in your circuit as even the board traces can add resistance/capacitance/inductance to a running circuit.

I have a working pic rated as 8mhz running at 12mhz (same product family as a 20mhz chip), I also have a atmega8-8pu with a 16mhz crystal and a arduino 16mhz bootloader and that runs fine, maybe i should take my hot air gun and heat it until I get errors :D

I have also had the atmega32-16pu running error free @ 20mhz 5v

my temp range is about (12c-35c) for my projects if you needed very high or low temps or very accurate data that has no error handling then it would be best to stay within datasheet guidelines.

if like me you love to experiment then try it and report your findings to us :)

I have not done any testing of these chips at 3.3v lowest for me is 4.7v so I am intrested

What he is saying that the few he has tried work so far. I am so not impressed with this sort of shabby approach.

to push the boundaries of knowledge with experiments and testing is what science is about 8-)