Hello,
in your headline you write about linearity and 752 vs. 3458A.. that's the wrong comparison parameter.
The 3458A is the most linear analogue comparator, that allows to make best transfers within an order of magnitude, e.g. from 10V reference down to 9, 8, 7, ..., 1V. Only the quantum based JJ array is better.
Therefore concerning linearity you have to compare the 3458A with the Fluke 720A KV divider, where we have to compare < 0.02ppm vs. 0.1ppm INL.
Transfer uncertainty is another aspect, where additionally other parameters come into play, what Conrad and 2N3055 have correctly suggested already.
You obviously mean 10:1 and 100:1 transfers, when comparing 3458A and 752A.
The 752A beats the 3458A in two aspects.
At first, the 0.02ppm INL of the 3458A computes into 0.2ppm @ 1/10 F.S., plus other sources of uncertainty makes about 0.3ppm 10:1 transfer (see hpj 4/89)
The 100:1 transfer uncertainty then accumulates to about 0.6.. 0.7ppm. But these are theoretical limits only, maybe only accessible with certain tricks.
But if you look into the official specification of the 3458A, the range calibration is not better than 1.8ppm for 10:1, and not better than 2.5ppm for 100:1, where you only might substract a 0.5ppm reference uncertainty. The official Transfer accuracy of (0.05 + 0.05)ppm will yield 0.55ppm for e.g. a 10V => 1V transfer, and 0.55 + 3.1ppm down to 100mV.
The 752A in contrast is officially specified 0.2 and 0.5ppm of output, i.e. 3 times and 7 times better, for this example.
The 2nd aspect is the High Voltage case, where the 752A by design is far superior.
These 0.5ppm for a 1kV => 10V transfer (or measurement at these levels) include its self heating effects, whereas the 3458A has up to 12ppm additional error.
All other DVM and also calibrators like the 5440B or 57x0A show this effect, but mitigate it by calibration of this self heating effect, or by heating the divider.
Therefore, please keep both instruments under all circumstances, especially if you ever want to measure/calibrate at voltages between 100V and 1kV.
One last remark, Conrad stated that the Fluke 752A is a fundamental (standard) and can be checked with .. simple procedures.
The first statement is true, i.e the 752A is a primary transfer standard 'by design', see elaborate error consideration in the 752A manual. All its uncertainty is derived from basic principles, and specification upper limits are derived from calculations only.
Therefore, if e.g. the switch resistance deteriorates over the years (it's compensated in a tricky manner inside the 752A) then you will not notice that.. This error source and its mitigation is not explicitly mentioned in the manual, though.
The 2nd statement is not that trivial.. as there really is no simple method to check (calibrate) its uncertainty.
You would need an intrinsically better transfer standard for that.
Only the JJ array can be used for that, for 10V or less.
For High Voltages, there is no such quantum standard available, so you need to design a different divider, as well estimate its sources of error, and compare both.
You also might check the self heating effect by dynamic methods.
Frank