SENT protocol output unit is LSB. What does this mean?

I have read couple of datasheet of devices which uses SENT protocol. They define the output between 193 and 3896. The unit is LSB. I think that this does not mean least significant bit?

• Why do you think that? Jun 4 '18 at 11:21
• because LSB is not an unit? I don't know
– Kono
Jun 4 '18 at 12:19

It does in fact mean LSB or, alternatively, the measurement amount corresponding to 1 LSB. The specs you're looking at are measurement ranges. The numbers given are the highest and lowest readings you can expect. Note that they're 12-bit readings, meaning they have a potential (positive) range of 0 to 4095. 4095 what? Depends how it's calibrated, but to skirt that quagmire, they call it a LSB and spec those ranges accordingly. It's kind of like using "percent of full scale" as a unit, but the fact that it's already digitized makes LSB a natural unit.

• I did not understand your explanation. LSB means least significant bit or byte. Yes indeed, the 12-bit data represents a value between 0 and 4095.
– Kono
Jun 5 '18 at 16:05
• Example: 2500 LSB means 2500 least siginicant bit or byte. That does not make sense..
– Kono
Jun 5 '18 at 16:05
• 2500 LSB is interpreted as a numeric value of 2500 times the amount and units represented by 1 LSB. For example, if you have a range of 0-3.3V using 12 bits, 1 LSB means 3.3v/4096, or 0.0008057V. Jun 5 '18 at 16:26

At one company we've used internal terms "A/D count" or "A/D tick" for this. The meaning was the incremental amount of physical value1 which corresponds2 to an increment of 1 in the digital value3.

1 Usually voltage or current, but could be other physical values too.
2 The correspondence is usually linear.
3 From 0xA2 to 0xA3, for example.

Why use arbitrary units such as LSB, A/D counts, ticks instead of using physical units of measurement? I can see two reasons.

• The exact relationship isn't always known.
For example, a relationship between input voltage and the A/D reading depends on the reference voltage. The authors of the datasheet at the time of writing can't know what the reference voltage will be, and it may also vary dynamically.
• The A/D values are always integer, and it often makes sense to stay with integer calculations. In some embedded controllers, for example, floating point math isn't practical.

I also use "epoch" as a unit of time. A sampled system usually has a refresh rate, and one epoch is the refresh period. Why not use normal units of time? The period may be an odd-looking number when expressed in engineering units of time (how about 17.39ms). More importantly, I may decide to change the refresh period during the project.

Sometimes I end up with [ticks/epoch] as a unit of measurement. Why not.