Before I found that, I found XAPP 429, on interfacing 5V to
CoolRunner-II CPLDs. On page 3, it says "A simple series terminating
resistor is NOT an acceptable solution to interfacing 5V signals. In
this situation, the current applied to the pin is not the offending
factor. Since the cumulative damage to the gate oxide impedance is
caused by voltage, merely limiting the input current does not protect
the oxide."
Why is the Spartan-3 different than the CoolRunner-II in this regard?
XAPP 429 goes into great detail on the gate oxide, but doesn't make any
mention of clamp diodes in the input pad structures. Did the CR-II
really not have clamp diodes on the inputs? Or were the CR-II clamp
diodes rated for significantly less current than those of the S-3?
Without some understanding of why such different advice is given for the
CR-II and S-3, I'm reluctant to use the series resistor approach.
Eric
Spartan 3 has clamp diodes that always prevent the voltage on the pin
from being more than a diode drop above Vcco (or below ground).
So, if Vcco is 3.3V, and the diode drop is ~ 0.5V at the current thru
the 270 ohms, then the max V is 3.8V, well below the abs max limit of
4.05V in the data sheet.
But, take care, as if you have Vcco = 3.45V, then you are getting
uncomfortably close to the 4.05V abs max limit.
As it turns out, the characterization work we have done has shown that
the nmos gate is very robust, and the pmos gate is the one limited to
the no greater than 4.05V restriction (in .25u IO used by V2P, S3, and
V4). In order to stress the pmos gate, you would have to have a voltage
below ground, AND a high Vcco! So for the 5V and a resistor case, even
if you had 3.45V Vcco, it still does not represent a reliability issue
with the nmos gate oxide.
I have stated here before, that the pmos stress >4.05V results in a
leakage of ~ 10uA over time, and a weakening of the pmos output
transistor by as much as 10%. This is not affected or accelerated by
temperature. If the IO is always used as an input, then you don't care
about the driver becoming weaker over time. If the input is thru a 270
ohm resistor from a 5V CMOS output, you probably also don't care about
10uA of input leakage. But, the leakage is an indication of gate oxide
damage of the pmos pullup, so we can not recommend operating it under
these conditions.
Austin
And CoolRunner-II do not?
Thanks,
Mark
If I want to drive a Spartan 3 input from 5V TTL (not CMOS), such as
an SN74LS14, do I need a series resistor, and if so, what value? The
data sheet gives a typical output voltage of 3.4V at -0.4 mA (which is
the recommended maximum Ioh). They don't give a maximum Voh. Since
it's bipolar, it's obviously less than the supply rail. I suppose I
could size the series resistor as if it were a 5.5V CMOS output, but
I'd prefer to use a lower value if it can be guaranteed not to exceed
the maximum current of the S3 clamp diode.
Thanks,
Eric
I would run a spice or IBIS simualtion of the driver into an open.
Then you would know.
Austin
Is 74LS still available? Why would you use it and who still makes it?
In theory, you are probably OK electrically.
In practice you need to watch the production to make sure no one
does a 'Heck, 74LVC14 (etc) are the same as 74LS14', replacement :),
plus LS14 have to be getting near EOL....
-jg