Three different ranges, often confused
Scene range. What the light in front of the camera actually does. A sunlit exterior with shadowed interior can span 20 stops or more.
Capture range. What the sensor records. Modern cinema cameras reach 14 to 15 stops; phone sensors considerably less, though computational stacking closes some of the gap.
Delivery range. What the encoding can carry to a viewer. This is where SDR and HDR diverge, and it is the only one a conversion can affect.
A conversion operates on the third. It cannot widen the second, which is why footage shot with clipped highlights stays clipped: the information was lost at the sensor, before any file existed.
Where the standards draw the line
ITU-R BT.709 fixes SDR's reference white at 100 cd/m² and the range below it. ITU-R BT.2100 defines both HDR transfer curves, PQ and HLG, and it is the document that made HDR delivery a specification rather than a set of vendor choices. The difference between the two documents is, in practice, the difference this whole site is about.
Stops, and the "ten times" claim
Each stop doubles the light. The figure often quoted for HDR, ten times the headroom, comes from comparing SDR's 100-nit reference white to a 1,000-nit mastering target. That is a little over three additional stops at the top end. It is a fair description of the delivery format and not a claim about what any particular display will show you.
Questions people ask
How many stops does my camera capture?
Modern cinema cameras reach roughly 14 to 15 stops, mirrorless bodies somewhat less, and phone sensors less again before computational stacking. Manufacturer figures are optimistic because they count stops down to a noise floor most people would not call usable.
Does converting to HDR add dynamic range?
It adds delivery range, not capture range. Conversion widens what the file can carry and cannot widen what the sensor recorded. That distinction is the whole reason clipped highlights stay clipped after conversion.
Related: the numbers behind the headroom.