Thank you,SRF. Very detail. I understand most of you said.
But I still couldn't understand why the structure didn't function when AVDD was connected to power supply.
For example, AVDD is connected to 5V power supply, when I/O pin is 2KV( HBM), I think the snapback beakdown could still happend. Am I right ?
Thank you.
SRF Tech wrote on Oct 14th, 2007, 1:29am:Hello xwj623,
So this is interesting for an I/O protection. Rajeee understood the operation correctly and stefan correctly noted that the circuit will only function when there is no power, or at least as an active device, passive snapback will still be in effect (which is the case with most ESD protection circuits), now here are a few additional insights i would add.
This is an attempt at using an active clamp (sometimes referred to as FET or BigFET clamp), only these approaches are generally the exclusive domain of power supplies and here is why. Reasonable ESD currents can be anywhere from ~1.3A (~2kV HBM ideally) to 10A (largest 500V CDM event for most semiconductors). In order for a fet device to actively clamp such currents in its normal operation, it needs to be a BIG device. Most power supply clamps are generally ~1000-5000um wide and have multiple clamps placed on each domain (as a power domain it is feasible to place such large clamps). The idea is if the clamps can properly conduct the ESD currents under normal operation then everything is good. Now the I/O are usually protected by a diode to the power supply. The diode easily handles these large currents and then the current flow down the power bus to the Big FET clamps and get shunted to ground. Normally placing a BigFET directly on an I/O is prohibitive because of area issues.
With the design you have here, someone figured if they removed the diode and put a big clamp on the I/O itself they will be saving a voltage drop and simplifying the ESD architecture. Unfortunately what they forgot is that this device as you have it sized (I am assuming with 8 fingers, that are say 50um long at the widest, we are looking at a clamp of maybe 100-400um wide), it will never conduct proper ESD level currents without seeing voltages far beyond its normal range of operation, in which case what we now have is a LV triggered snapback device.
In otherwords even though it looks likes it may be an active device conducting ESD current, because the device is not large enough to conduct full ESD currents it will actually snapback as if it was a grounded gate nmos, only the actual snapback voltage and current will be lower than normal because it will be biased at the gate, similar to a gate-coupled NMOS ESD device. This is what is likely providing most of the protection.
Hope this helps,
Stephen
SRF Technologies