The 34470A is not the first 7 1/2 digit instrument (from HP / agilent).
Before that, the 34420A, nV/ µOhm meter already had 7 1/2 digit.
It is based on the elder Multislope III A/D, similar to the 34401A, and had slightly better linearity than this new DMM.
The 34420A features DCV and Ohm only, and has an LM399H built inside.
The LTZ1000A reference is even stable to 8ppm/yr, as the 10V DCV range has 8ppm 24h uncertainty, and after one year 16ppm, therefore 8ppm drift.
It's a pity, that they did not improve that stability above the 3458A, by also running the LTZ1000A on 95°C. (15k/1k divider for oven)
They could have reduced that temperature to 75°, so that the elaborate selection process would have been easier, as that lower temperature would give stabilities of 2..4ppm/year very easily.
About the monitoring process, see HP journal 4 / 1989, for the 3458A.
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf Although the circuit is very similar to the one in the 3458A, it only features 1ppm/°C temperature stability, whereas the 3458A has 0.15ppm/°C. Maybe that's due to the usage of all SMD resistors for the oven and reference circuitry, (instead of leaded metal foil ones), which mainly influence the T.C.
The reason, why the 10V DCV range is the most stable on mostly all bench DMMs, is the fact, that this range relies on the stability of the reference plus one or two resistive amplifiers for the A/D conversion reference voltages of about +/-10 or +/- 12V. These two amplifier circuit will also be Autocal'ed for T.C.
All other DCV ranges have additional resistor dividers implied, which create these 5ppm/°C.
That is the 100:1 high voltage divider for 100, 1000V, and a x10, x100 amplifier for 1V, 100mV.
Latter one is also used on DCI, so creating an even higher T.C. as the current shunt resistors add up here.
The 10k VH102Z resistor is used as a real reference resistor, max. 0.8ppm/°C T.C. and (per datasheet) 2ppm/6 year stability.
The Autocal function transfers its uncertainty and stability to all other Ohm ranges, to some degree.
The datasheet of the 34470A therefore is very conservative regarding Ohm 1 year stability, although this reference resistor may probably be much better than the one in the 3458A!
Frank