as wären sozusagen bessere trefi-schritte abseits von z.b. 54000 geht, 55000 aber nicht mehr, und das dazwischen ausloten ist praktisch zeitverschwendung?
REFI depends on tRET @ target voltage @ target temperaturecase & refresh method.
Even has PCB layers as variable.
REFI 54000 has no binary step for the controller. Neither 55000
Input becomes invalid.
Memory while running in picosec, its not really on our worldclock.
ns targets are silly and only done @ 1.1v where everything scales cleanly (decimal)
Only for QA specification, but still not the way memory operates at. Neither the controllers have such fine granularity.
Even here, there is no way to set anything finer than that.
"nice looking numbers" you see are straight-out wrong. They have no mapping. Controller will not accept them.
Also those values need (value-1) as input, given decimal to hex to binary conversion. Value 0-65535. 0 is then a value.
32767 & 65535 herby are correct values. Same as 40959 & 49151.
Only for QA specification, but still not the way memory operates at. Neither the controllers have such fine granularity.
To paint a perspective
You know how little 1ns is
1 sec = 1e12 ps & that value is then rounded/aligned with/by Herz.
It cant be thaat fine-granular
It doesnt move in our perceptiveness & even docs can't display that.
Its soo much faster and has to follow a logic.