3458A ADC section innards

[diegogmx]

I've been checking out the schematics of HPs 3458A multimeter today, it really caught my interest since it is a device designed in the end of the 80s (before i was born actually), yet the accuracy it reaches is remarkable even by today standards.
when checking out the adc section i must say (with some embarrassment) that i got quite lost.
first of all U180 the ADC or at least part of it seems to be a custom part, so i really don't know what it does with a decent level of certainty, i would appreciate if someone could illuminate mi a little there
i've read in a cern paper that this meter as well as other super accurate ones use dual slope adcs, which if im not wrong in this case it's not made from a single ic, the integrator seems to be external and at least to me surprisingly convoluted, so does the comparator, and there is a slope amp that looks just like a precision rectifier, not yet sure of it's purpose

in short Help
i would really appreciate if someone could guide me through this a bit

*cant attach the schematics, the file is too big
yet it can be downloaded from here
https://www.schematicsunlimited.com/h/hp-agilent/hp-3458-8-digit-multimeter-schematic-and-clip

[Micke]

The 3458A design is discussed in great detail in the April 1989 HP Journal 
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf
Hope it helps!

[Kleinstein]

U180 combines a very good quality resistor array and matched analog switches (presumably JFET) and switch drivers.

The ADC is a classical multi-slope design. So in principle not that special, though running relatively fast, e.g. with a rather fast integrator.
The other point seems to be a good implementation: e.g. very good resistors inside U180 and little parasitic effects.

While the principle circuit is relatively easy to reproduce (with modern parts), the implementation with little extra effects form something like capacitive coupling or supply coupling is still difficult today. So getting the same linearity is the difficult part, while the noise level should be relatively easy.

[diegogmx]

The 3458A design is discussed in great detail in the April 1989 HP Journal 
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf
Hope it helps!

i'll take a look at it

U180 combines a very good quality resistor array and matched analog switches (presumably JFET) and switch drivers.

The ADC is a classical multi-slope design. So in principle not that special, though running relatively fast, e.g. with a rather fast integrator.
The other point seems to be a good implementation: e.g. very good resistors inside U180 and little parasitic effects.

While the principle circuit is relatively easy to reproduce (with modern parts), the implementation with little extra effects form something like capacitive coupling or supply coupling is still difficult today. So getting the same linearity is the difficult part, while the noise level should be relatively easy.

yes, the linearity for what i've read is about 0.1ppm which is really insane, for what i've seen analog devices highest resolution adc today has about 0.5ppm peak in full scale

[tggzzz]

The Art of Electronics III, section 13.8.6 "Designs by the masters: Agilent’s world-class “multislope” converters"...

With these analog-switch applications in mind, we’re in good shape to understand the “multislope” techniques used in instruments like the Keysight 34420 7 1/2-digit and 3458A 8 1/2-digit multimeters.

[dacman]

U180 used to be a fairly flat IC with pins on all four sides.  It is now a daughterboard and I believe the designation is A13.

[TheSteve]

U180 used to be a fairly flat IC with pins on all four sides.  It is now a daughterboard and I believe the designation is A13.

U180 has never changed. It is the gate array ASIC that was changed and now uses a daughterboard.