This isn't my usual job, I'm not an EE, but I want to give it shot...
This question concerns developing an interface board to connect a device (DUT) to a test system in a manufacturing scenario.
The signals I have access to are a bunch of 1.8V cmos logic lines from the GPIO pins of an embedded CPU, a 5V power rail, and GND (see diagram below). Everything is in a fairly tight space, I have control over all power/ground etc.
The 1.8V GPIO signals (s1,s2,s3...) eventually get read by a opto-isolated digital input device (happens to be a NI-6525). The opto-isolated digital inputs need at least 3.2V to sense a HIGH so I must level shift these, moreover, the current input per channel is given as "3 mA MAX"-- too much for 1.8V cmos I think?
To perform level shifting, I was thinking about the MAX3000E because it is designed for things like cellphone cradles, smart card readers, etc. It has ESD protection and it can handle 8 signals and I can use it to level-shift from 1.8 to 5.0, all I have to do is provide voltage rails for 1.8 and 5 from my test system. No problem, I thought...
BUT looking at the datasheet for the 3000E it seems that the output current I can provide from the 5V (Vcc) side is only 10uA max! That isn't enough to drive the diodes in the digital input device.
So now, I am thinking about adding mosfets to each output of the level converter so that I can deliver enough current to the opto-isolators.
Am I even using the right level converter for such an application? I get the feeling I am doing this wrong. The test system uses opto-isolators but it is a good pick for other reasons.
I thought about ditching the level translator entirely and just using mosfets to drive the optoisolators from the 1.8V signals. But then, there's the problem of ESD protection, losing the nice "enable" pin, plus I don't feel confident with with just naked mosfets. Is this easy to do with mosfets in this scenario where DUT's are going to be constantly plugged and unplugged into the system? Should I invest the time in trying to do this with just mosfets? Will it be reliable in a factory system?
Looks like the moral of the story is that "self-configuring" bidirectional logic translators can only deliver a very small amount of current on the output side (~uA's). In my scenario, I am using the output side to drive opto-isolator inputs on the system side which require ~mA's of current. This requires using an additional buffer of some kind, or, selecting a different logic level translator. Since I don't need to change direction in-situ (never be tx'ing and rx'ing on the same line), selecting a translator that has fixed or settable direction solves the problem because these don't seem to have the extremely small max current spec on the output.