Sensor heat and darks: when to start?

[Demonperformer]

I have recently purchased a ZWO ASI224MC (uncooled) camera, primarily for lunar/planetary imaging, but the specs also say it can be used for some DSOs.
Now this has got me thinking about sensor heat and darks.
If I take 3000 frames at 1/10s, that is five minutes that the sensor is running in a period of fractionally over five minutes (USB3, so download time does not add a great deal to that).
If I take 5-min exposures of a DSO with an uncooled camera, the popular wisdom would seem to be to take darks.
Now I am no expert on camera mechanics, but it seems to me that, if a 5-min sub involves enough heat being generated by the sensor to require darks, by the time I get to the end of my 3000*1/10s run, the sensor is going to be running just as hot. But I don't see people recommending taking darks for that!
So it would seem that there is, somewhere between 3000*1/10s & 1*300s, a point at which darks become worth doing and below which they aren't. I guess that there is no "definitive" answer to where that point comes, but I would like to guage the "accumulated wisdom" of the community. At what point would darks become worth doing? 2*150s? 20*15s? 300*1s?
(Any what is your reasoning?)
Thanks.

[drjolo]

There are two main noise sources in the camera sensor - read noise and dark (temperature) noise. Read noise is the same for each frame regardless the exposure time. Dark noise increases with exposure (and also with temperature). And darks become profitable at the moment, when dark noise effect becomes significant fraction of total noise. It will also depend on sensor/ambient temperature, but in my experience it is not worth to calibrate with dark for exposures shorter than one second. But there are definitely many things to consider. 

[ollypenrice]

The popular wisdom is now much divided on darks with many imagers, both DSLR and CCD, not using them. I'm one. With DSLRs the problem is not knowing what the real temperature is. Many would rather subtract a master bias as a dark and dither guide in order to average out the hot and dead pixels, ideally in Sigma clip. I use a master bias as a dark and a defect map, along with a hot pixel filter. The results are better than with darks and the process takes a fraction of the time. This, by the way, is on very noisy Kodak chips.

I'd try with, without and with master bias as dark.

Olly

[wimvb]

If you can see amp glow in your images, you need darks. The only problem is that for those to work, they need to match the light frames in both temperature and exposure time.