I am working on a case where total trip time(time delay, relay time, breaker time) is lower then 100ms(about 80ms) due to a busbar protection system. Is it reasonable to just use 100ms?
For extremely short times, the possible error of each time(breaker, relay etc) would be significantly part of the total time?

Anyone?

I hardly ever go below 100ms. IMO, there are almost too many variables to account for so I leave some room. Kind of like including CTI on a electro- mechanical relay

I would use the 80 ms number. Given that incident energy is directly proportional to time, rounding up to 100 ms increases it by 25%. 80 ms actually sounds pretty long. And don't forget that the worst case quoted delays are affected by phase angle. So you can figure there is 8 ms of "slop" in each layer along the way.

TimMol comments:
I wonder under what conditions the stated “total trip time(time delay, relay time, breaker time) is lower then 100ms(about 80ms) due to a busbar protection system”. Assuming you are considering a feeder breaker , 3 cycle interrupting time 50 ms on a bus fed by a main 3 cycle breaker source, protected by modern solid state relays and all in excellent system maintenance/operating condition, assuming zero CT and relay errors no Coordination Time Interval (main – feeder) and no safety factor the total clearing time might in theory be 50 + 8 – 58ms
However the minimum Coordination Time Intervals (which may be somewhat subjective), specified by IEEE 242 will vary from 120ms to 200ms under these best case conditions. In the event you have a 5 cycle ( 83ms breakers, electromechanical relays etc ) the total clearing time will be greater.

Looks very reasonable to me. It's 87 relaying so there's no inverse time curve It's all definite time, so CT errors and the like only apply in terms of pickup, not delay. Coordination time wasn't mentioned and probably wouldn't be used with 87 relaying anyways since the zoning is based on two sets of sensors instead of communication between breakers. The rest of your considerations would be very appropriate for an induction disc relay but not a microprocessor based one. I can see possibly adding one cycle if it still uses an old electroswitch relay as an extra layer or as you said 5 cycle vs. 3 cycles.

As to 120-200 ms, what in the world? I've tested lots of breakers and the only time I've ever seen anything that ridiculous is on old breakers that haven't been serviced in a long time so the grease hardens up and they are audibly very slow, or on induction disk relays like say CO-9's where the calibration is all way out. Those numbers (not sure where you got them...I checked IEEE 242 and didn't see them anywhere in there) might be reasonable for a very old breaker using electromechanical relay protection that is in pretty bad shape but not for anything built and installed in the last 20 or 30 years.

I would also add 0.5 cycle for LOR.