Advice on "absolute maximum ratings"

[stevenhoneyman]

I'm looking to use an existing part (which I only have two of - can't 'just try it'!) in a 3.3v circuit. The datasheet on page 3 here: http://media.digikey.com/pdf/Data%20Sheets/Austriamicrosystems%20PDFs/TSL2562,63.pdf says that the recommended supply is 2.38 to 3 volts max, but the absolute max is 3.8, and the i2c range is 1.25 to 3.6...

My question is really how maximum is maximum/absolute maximum usually? To me that looks like 3.3v should be fine as it's nowhere near absolute max, but then again if it was OK, why does the recommended supply range stop at 3? It also looks like the bus voltage can be higher than the supply voltage, which isn't usually the case. Maybe they just missed the ".6" off the end of the recommended supply?!

[seriouscoinage]

I believe that this means that the specs in the datasheet aren't guaranteed for voltages above 3V, but because the absolute maximum is 3.8V, the chip will not be damaged unless the supply is higher than 3.8V. I think you could safely try 3.3V without the chip being damaged, but there's a chance it might not work how you expect.

[AndyC_772]

Normally the "max" value in the recommended operating conditions represents the maximum voltage the chip will work at while still remaining within specification. Go above that limit and there's a risk that the chip won't work quite as well and/or for as long as it should. Above the "absolute maximum" figure, expect the chip to die quite quickly.

Personally I'd drop the 3.3V supply to somewhere nearer 2.5V, either with a small LDO, or even just with a silicon diode (and perhaps a small dummy load resistor to ensure you get the full, roughly constant 0.7V voltage drop). The I2C pins can, of course, be connected directly to your 3.3V I2C bus.

[T3sl4co1l]

You're also not guaranteed to get specified operation above 3.0V.  I should guess it's intended to be a 2.5V supply part.

In practice, "3.8V abs max" probably means the device avalanches in the 5-8V range.  If carrying much current in the process, destruction is sure to follow.  It may survive under current-limited conditions, but need not function properly while doing so (often, the charge injection that occurs causes disruption of analog and digital functions -- transistors go leaky due to charge diffusion -- and operation grinds to a halt).  You're also not guaranteed to get any consistent avalanche voltage from part to part; for all you know, one might start hogging current at 3.9V, or it might go double.

If you need a 3.3V supply (and don't want to bother with a 2.5V rail for just this thing), can you locate a substitute?

Tim

[stevenhoneyman]

Thanks for the replies. I like the diode idea... I'll toy around with that and fall back on adding a 2.5V regulator.

You're also not guaranteed to get specified operation above 3.0V.  I should guess it's intended to be a 2.5V supply part.

It looks that way, although it's being run at 2.8V by the device manufacturer.

If you need a 3.3V supply (and don't want to bother with a 2.5V rail for just this thing), can you locate a substitute?

Unfortunately that's not practical - I'm working with an existing FPC which this is a part of (in the BGA type package) so there's no room to bodge in a different part. Unless they all use the same pinout... hmmm... let me check a few.

[LukeW]

It's good practice to try and power it in the middle of the nominal voltage range whereever possible.

The current consumption of that device is so small, so adding something like a zener to provide 2.5V will be easy and won't waste much power.

If you don't have easy access to a 2.4-2.8V or so zener diode, using a standard rectifier to drop 0.6V or so is probably a good choice... but you want to bias it with enough current draw so it's actually fully forward biased, using a resistor, as another poster mentioned above. Otherwise it's unlikely to give you that full voltage drop.

[stevenhoneyman]

Hmm so there is one with a better supply voltage... just a single digit difference in the part number! http://ams.com/eng/content/download/250096/975518/file/TSL2561_Datasheet_EN_v1.pdf

I wonder how easily that chipscale package is to solder back down again.