Any one try overclock a PIC microcontroller? 2

[KammutierSpule]

Hi all!

I'm thinking trying overclocking a PIC microcontroller,that is running it at a high clock rate...

What do you think? Have you tryed it too?
What problems can I have?
How much can I overclock it?

thanks all.

 

[melone]

Why would you want to overclock the IC? You will be operating outside the guarenteed limits of the device. Could it work? Sure. Is it a good idea? No.

Overclocking will result in one of 3 scenarios:

1) Functionality appears unaffected, but you will overheat the die of the device (thus decreasing the livespan).

2) No longer appears to function. You will try to read / write faster than the die / peripherals can handle. Data will be corrupted, and potentially damage / destroy devices, i.e. if you corrupt a OTP device due to improper timing configuration.

3) Functionality is sporadic. You might be on the very edge of some timing requirements that allow it to work some times, while not at others. This is the worst case scenario, since you will not be able to repeatably recreate the failure mechanism, you will not be able to "solve" the problem.

 

[IRstuff]

An addendum to Melone's list:

4) Overclocking as mentioned by Melone is operation outside of the part's specifications, which, MAY work at some limited temperature or environmental range, but the part will almost certainly NOT work over its intended environmental and temperature range.

TTFN

 

[designchain]

And still another addendum:

5) if you design this in to a product and ship it you will reduce yield (increasing rework during mfg) and affect product robustness and reliability in the field.

6) if this is a one-shot project for your own use you can consider using a chip cooler on top of the controller to manage the die temperature. CMOS performance is directly affected by die temp: cooler=faster. Watch your setup/hold times because they'll change too - causing some of the problems alluded to above.

Mike

--
Mike Kirschner
Design Chain Associates, LLC

http://www.designchainassociates.com

 

[KammutierSpule]

Hi thanks all for the replys.
Seams you are all right...

If I overclock it, it will be for a single time project.. or just some experiences... not for comercial stuff... its for my own use...

and.. when I tell overclock I mean... if the max clock of the PIC is 20Mhz for example.. so.. I will stry run it at 25Mhz no 30 ou 40..
I know it will afect too...
but some of the things you are tell like temperature well.. I dont think that for just more 5Mhz it will burn...

My initial question is more if anyone have some experiences in that..

But I will make some too..
I really dont know If I will need more speed..
well..
let me work..

thanks all again.

 

[melone]

Why do you "feel" that 5MHz is alright, and 10MHz is not? As an engineer, I would highly recommend not exceeding specifications!

 

[KammutierSpule]

hehe yee.. maybe its only a obcetion

 

[raivo]

I once tryed to overclock 8051 from 12 to 16 MHz and another from 16 to 20. I used theese on room temperatures and had no problems. IMHO you may try overclocking if your application is for your own use and you make some strong reliability tests before you decide that it is ok. Also i know that AD and DA converters will not tolerate any overclocking (errors will creatly increase)

 

[OperaHouse]

Not exactly related, but I saw a Microchip design that ran a resistor always connected to the supply over the micro to heat it up. Evidently this was their effort to deal with the lower temp spec of the standard product.

 

[peebee]

I worked a summer job at a wafer fab plant. After the IC packages were completed, they all went into sorting machines. The machine did some tests on them, and separated them into GOOD and BAD piles.

But the GOOD piles were further separated, kind of into GOOD, BETTER, and BEST. I'm not sure exactly what they were testing for, but I suspect speed was a key determinant.

Now I'm getting into the realm of speculation, but I suspect the main difference between, say, an 800MHz chip and a 1.2GHz chip is how it tested out prior to shipping, NOT any inherent difference in the design. They would have sold your chip rated for a higher speed if they could have.

You might be able to overdrive the chip, and I suspect the probability of physical damage, ie, smoke, is low. But the probability of goofy operation is probably pretty high. And those errors are likely to appear rather sporadicly, and perhaps only on certain operations. If you want the chip to perform correctly, I would strongly advise against overdriving the clock speed.

 

[IRstuff]

yuo're quite right, up to a point.

As mentioned above, YIELD is what drives the separation. Any given wafer will have a distribution of performance. In general, particularly for MOS devices, room temperature performance will predict at temperature performance, so devices are binned by how much speed margin there is at room temperature. This is the primary reason that overclocking works at all, the extra margin at room temperature that's gobbled up at high temp.

However, the range is generally not much more than about 20%. Much higher performance than that requires either a die shrink, process change or design change.

TTFN

 

[VolkerR]

One more thing:
if you just want to run a pic at 40MHz, there is no need for overclocking. Most 18Fxxx PICs are rated for 40MHz, and there are now many 18F types available with similar functionalities, partly even pincompatible to 16Fxxx devices. The 18F devices are software-kompatible, too (run the old PIC16 instruction set plus new instructions)
Take a look at the Microchip website.

Volker