The maximum permissible disconnection time is 0.4 s in TN system. Why and from where this value (0.4 s) is obtained?

As I understand it, the 0.4s disconnection time was an outcome of a working group within IEC that set about developing a maximum touch voltage curve for use in installation design so that maximum disconnection times could be set. That curve itself was derived from a line plotted in the AC-3 zone of the graph AJ posted above (the body time/current curves) and certain other assumptions. Amongst the important assumptions made was the stipulation that the curve was relevant for normal dry conditions. Since we all have different body impedance it was decided that the body impedance would be representative of a value hand to foot exceeded by 95% of the population. That value is 1000ohms. To allow for two hand contact to two feet the value was reduced by half to 500 ohms. A value of 1000 ohms was then added back in to account for footwear and floor resistance. Body impedance is affected by many things including the value of voltage. So for example, for the same person at 50vAC the total assumed impedance is  taken as 1725 ohms while at 220VAC it is taken as 1500 ohms. 

So the same person making two hand to two feet contact at 50v would suffer a body current of 50/1725 = 29mA and at 220v it would be 220/1500 = 147mA. Reference to the line established on the body time/current current as noted above, the latter would require disconnection within 180ms whilst there is no limit on the former. 

Similarly, at 100VAC, the body impedance is indicated as 1600 ohms which would result in a body current of 62mA and a required disconnection time of 400ms. This value of touch voltage Is taken to be a likely magnitude touch voltage that would exist where the nominal voltage is 230v but assumes that the relationship between R1 and R2 is equal in terms of resistance. This is not the case in the UK where reduced csa is used for twin and earth cables which, when using the same calculator for prospective fault voltage, would result in higher values and, in consequence, shorter required  disconnection times. It would paper that we are happy to stick with the 0.4s anyway.

As I understand it, the 0.4s disconnection time was an outcome of a working group within IEC that set about developing a maximum touch voltage curve for use in installation design so that maximum disconnection times could be set. That curve itself was derived from a line plotted in the AC-3 zone of the graph AJ posted above (the body time/current curves) and certain other assumptions. Amongst the important assumptions made was the stipulation that the curve was relevant for normal dry conditions. Since we all have different body impedance it was decided that the body impedance would be representative of a value hand to foot exceeded by 95% of the population. That value is 1000ohms. To allow for two hand contact to two feet the value was reduced by half to 500 ohms. A value of 1000 ohms was then added back in to account for footwear and floor resistance. Body impedance is affected by many things including the value of voltage. So for example, for the same person at 50vAC the total assumed impedance is  taken as 1725 ohms while at 220VAC it is taken as 1500 ohms. 

So the same person making two hand to two feet contact at 50v would suffer a body current of 50/1725 = 29mA and at 220v it would be 220/1500 = 147mA. Reference to the line established on the body time/current current as noted above, the latter would require disconnection within 180ms whilst there is no limit on the former. 

Similarly, at 100VAC, the body impedance is indicated as 1600 ohms which would result in a body current of 62mA and a required disconnection time of 400ms. This value of touch voltage Is taken to be a likely magnitude touch voltage that would exist where the nominal voltage is 230v but assumes that the relationship between R1 and R2 is equal in terms of resistance. This is not the case in the UK where reduced csa is used for twin and earth cables which, when using the same calculator for prospective fault voltage, would result in higher values and, in consequence, shorter required  disconnection times. It would paper that we are happy to stick with the 0.4s anyway.