Open collector comparator output with no Vcc

[Siwastaja]

A rather basic question.

Consider a traditional jellybean open collector comparator, with single supply. For example,
LM393 ( http://www.onsemi.com/pub/Collateral/LM393-D.PDF ),  or let's say
LM339 ( http://www.ti.com/lit/ds/symlink/lm2901.pdf ).

The simplified schematics in the datasheeets do not show any unwanted parts at the output transistor collector. But the ESD structures are not well defined. Are they schottky/TVS type, or diodes connected to the supplies?

Is it possible to connect such output to a voltage higher than Vcc, including the case that the comparator is powered off at Vcc=0? If I do this, should I expect considerable leakage (say, over tens of uA) at the output pin, while Vcc is below operating specification?

[Benta]

I've used both LM339 and LM393 as level shifters with no problems. Mostly running at 5 V and the output collector at +12 or +24 V. Removing V_CC creates no issues, it's a "pure" open collector.

[slugrustle]

It's helpful to note that for both parts, absolute maximum output voltage is specified as 36V and is listed as a different parameter than supply voltage. I think this is datasheet-speak for: You can put up to 36V on the output independently from Vcc. That said, it's not the clearest way to express the concept.

Your questions about leakage current and ESD protection are good ones. I wish those things would be explicitly stated in every datasheet...

[slugrustle]

I looked at the LM393 datasheet again. Max output leakage is 1uA with 30V on the output, from 0C to 70C ambient, with Vin- = 0V, Vin+ >= 1V.

I would guess that the same figure holds when VCC = 0V, since there is no voltage to turn on the base of the output transistor. You may even see lower leakage than the spec. The Vin- = 0V, Vin+ >= 1V in the leakage spec is probably there to put Q15 (in Figure 1 in LM393 datasheet) into saturation, in which case the output transistor base voltage is still higher than 0V.

The only way to know for sure is to try it, probably at a higher temperature.

[Siwastaja]

Indeed, the leakage is specified at 30V on the output and Vcc at only 5V as only 1uA over the whole temperature range. That's a great hint that there can't be any structures there that wouldn't work similarly at Vcc=0. In theory, there could be some effects when Vcc is, let's say, over 0.6V, but still under the operational minimum, so that the output transistor would be driven to the wrong state or leaks more while the supply is ramping, but I guess they are irrelevant in this case.

Thanks for confirming the assumption. You saved me from some extra worrying or an extra pre PCB design test.

[Zero999]

I've used both LM339 and LM393 as level shifters with no problems. Mostly running at 5 V and the output collector at +12 or +24 V. Removing V_CC creates no issues, it's a "pure" open collector.

Yes that's true. Also look at the LM311 and LM211, which have output transistors rated to 40V and 50V, respectively.
http://www.ti.com/lit/ds/symlink/lm311.pdf

[Siwastaja]

Thanks, this is great, you just saved me an extra transistor .

I need to pull down the worst case of 38V, which is just irritatingly over the recommended maximum rating of 36V on almost all jellybean comparators, and parametric searches don't seem to include the output transistor voltage rating, only Vcc as a parameter, so obviously I missed these guys.