I ran a second test round to determine the effective number of bits (ENOB) of some oscilloscopes that cought my interest because of advertising claims regarding high definition and extreme resolution values, the RIGOL HDO4000 and the R&S MXO44. ENOB values published are either very modest (8Bits for the RIGOL) or absent (MXO44).
Properly measuring ENOB is a challenge by itself, but the results are interesting. Neither scope can even reach the nominal number of bits specified for their ADCs, which is 12 in both cases.
The details are here:
Electronic Projects for Fun 
ENOB can be easily misinterpreted. It doesn’t account for any DC offset errors, and only measures a single tone. So, if the scope has phase non-linearities (most low end scopes don’t correct for phase), ENOB can look good while signal representation can be significantly distorted. ENOB is just one metric…better than none, but isn’t all inclusive.
I like to think of ENOB as a series of curves for scope: one curve for each vertical scale and BW. Vary the BW and you get another curve. Vary the scale and you get another curve.
Nice description of the source being a significant source of ENOB measurement difficulty. I’ve encountered this, and the need to use a very narrow notch filter on the source.
Why characterized ENOB @ 10 MHz, instead of the full scope BW? I don’t understand why some scope vendors do this…other than just to get a higher value. Is there some other reason?
LikeLike
Hi,
I think the reason why 10MHz is a common ENOB test frequency is twofold:
– its somehow entrenched as an industry standard. That does not say that better oscilloscopes also measure ENOB at higher frequencies.
The 110GHz UXR from Keysight has ENOB values up to the maximum bandwidth (5Bits).
– For the cheaper guys, they know that ENOB will always get worse with increasing frequency. At 10MHz, the dropoff is small for most ADCs used in oscilloscopes. So they dont publish values that are not so impressive.
Its clear that ENOB values will vary with scope bandwidth settings, scale factor, … A good choice is a bandwidth at least 10 times the signal frequency, and the smallest “true” scale factor (ignore “zoom” stuff and the like). Normally ENOB is tested in AC mode, to get rid of offset errors that can be in order of magnitude of the scale factor (e.g., 1mV). Including this would make ENOB values useless because DC will completely dominate, but ENOB is meant to give a measure of dynamic range rather than DC accuracy.
regards
Wolfgang
LikeLike