Have you ever touched on Absolute Maximum Ratings?

[Jon Chandler]

Hi Dave,

Just have to say I've enjoyed your blogs but I will admit I haven't watched all of them

Have you ever talked about Absolute Maximum Ratings?  I've seen more than one question like "How much can I exceed the absolute maximum ratings and get away with it?"  I can picture you screaming "Zero!" but some people just don't get the concept.

At Digital-DYI, I've been involved in a number of forum threads about the PIC18F25K20, which the Amicus Compiler (aka free version of Proton Basic) requires.  It clearly does not have 5v tolerant pins but there are a few people claiming otherwise.  In the pin descriptions in the data sheet some of the pins are listed as Schmidt Trigger and others as TTL, which some people claim means 5v tolerant.

When the absolute maximum ratings of Vdd + 0.3 v on any pin are are pointed out, the argument becomes "the maximum rating for Vdd is 4.5v and the pin is rated at Vdd + 0.3 v so that means the pin is rated at 4.8 volts.  And everybody knows that 5v regulators always put out less than 5 volts, so therefore....."  I'm certain your reaction is the same as mine to logic like that.

And of course when the errors of that argument are pointed out, the reply is "I've been in this business for 20 years and I do this all the time."  You may indeed....thanks for the warning - I'll stay away from anything you've designed.

This example is focused on my current head-banging exercise, but I've heard the same nonsense again and again when it comes to absolute maximum ratings.  A reinforcement of good design practice might be a good thing.



Jon


One of the forum threads

[EEVblog]

I've touched on this in terms of LEDs and overclocking (Rigol).
You can exceed absolute maximum ratings and other specs on some parts, but you really need to know what you are doing, have data to back it up, and usually have a very specific need to do so, and be aware of any potential downsides.
But generally it's almost always, a bad thing to do.

Dave.

[Simon]

as far as i know manufacturers make parts that will differ very slightly from one another but that are targeted at a certain operating range, be it voltage, frequency, current or otherwise. to make sure all parts can operate at the intended spec you might end up with a part that will go a little higher but this might not be valid for any part and the manufacturer will not guarantee the part after absolute ratings have been surpassed. I've put 7V into a pic on a number of occasions (accidentaly) and it has worked afterwards, but i'd certainly not take that pic and put in into a project that I expect to work for any amount of time

[Zero999]

The absolute maximum voltage rating for most 5V PICs is 6.5V so I'm not surprised it survived 7V.

The thing is, most parts aren't intended to be operated at their absolute maximum ratings but their recommended operating conditions. I know this applies to supply voltage, power dissipation and current rating but I don't know about CPU frequency all I know that parts from Intel are often tested and binned accordingly so they're probably run at 10% under their maximum usable frequency.

[saturation]

If you exceed the AMR, the manufacturer no longer guarantees the performance in the spec sheet.  YMMV after that.

I've frequently taken parts to AMR to purposefully fail them and see what happens.  Often, the AMR is conservatively rated by 50-100% so a 5V can easily reach 7.5V without damage, and potentially even 10V.  You dont' want to design this way, but it tells you how parts are made and rated; beyond AMR you are now outside the realm of 'lump matter discipline' and sometimes a component can have novel and new behaviors that are often exploited for different techniques.

[Simon]

The absolute maximum voltage rating for most 5V PICs is 6.5V so I'm not surprised it survived 7V.

The thing is, most parts aren't intended to be operated at their absolute maximum ratings but their recommended operating conditions. I know this applies to supply voltage, power dissipation and current rating but I don't know about CPU frequency all I know that parts from Intel are often tested and binned accordingly so they're probably run at 10% under their maximum usable frequency.

it's 5.5V for a pic,

I think with processors and speed the quality of silicone differs over the wafer, I believe that the centre is best so chips from the same wafer are speed tested and then given their set speed, presumably the fastest processor coming out of the centre is going to cost a whole lot more and most of us buy the mid-range and budget ones from the outer parts of the wafer that are more numerous and therefor cheaper.

passionate overclockers often research particular processors and want to know when it was made because when the processors are released and people start to overclock them they find that certain batches perform better and these parts will have a fairly high second hand value compared to another part sold second hand. I once got a AMD processor up to 115% it's rated speed (so the bench mark stopped calling it an XP2600+ and called it a XP2900+ which did not even exist as a processor  ) and that was with not really much over voltage but it was down to luck that I had a good part.

The thing that prevents extreme over clocking is heat, the faster you go the more voltage you need and the more voltage you use the hotter it gets and the hotter it gets the higher the voltage needed to keep stable and the higher the voltage the hotter it gets and so it runs away where there is a point you can't get anymore speed, this point varies depending on how mad you want to go on cooling

[Zero999]

it's 5.5V for a pic

Which PIC?

Have you actually read the datasheet?

5.5V is normally the maximum recommended operating voltage, the absolute maximum rating is higher: 6.5V for all the PICs I've seen.

If a current limiting resistor is used, there's also no maximum rating for the input voltage - it's possible to connect a PIC input to the mains via a 1M resistor. The maximum input voltage of Vs+0.3V and -0.3V is for constant voltage inputs only.

[Simon]

yes your right, my memory is getting bad already, I suspect most people will not bother to put any more than the standard 5 volts into a pic unless their trying to.... overclock it ?

It would be interesting to know how Rigol got 100 MHz out of 40 MHz ADC's in the DS1052 and DS1102 scopes, a 250% overclock is impressive to say the least let alone making it a mass produced thing

[squeezee]

yes your right, my memory is getting bad already, I suspect most people will not bother to put any more than the standard 5 volts into a pic unless their trying to.... overclock it ?

It would be interesting to know how Rigol got 100 MHz out of 40 MHz ADC's in the DS1052 and DS1102 scopes, a 250% overclock is impressive to say the least let alone making it a mass produced thing

They also produce identical 80MHz and 100MHz versions of that ADC, often with a mature product on a mature process you're going to have a majority of parts being binned as the top-end product. Of course you don't want to discount your higher spec part and kill the margin, but you also need to fulfill your higher volume orders for the cheaper product. So you end up marking and selling them with the lower model number instead.

Rigol just has to test them to make sure they meet the 100MHz spec, or perhaps they get them with the guarantee they're 100MHz parts.