Why is the Hi-Z (>10GΩ) impedance mode on DC only available for low voltages?

[6SN7WGTB]

Did search and although plenty of threads explain meter impedance effects, which I understand, none appear to address:

Taking an Agilent bench DMM as an example, why is the Hi-Z (>10GΩ) impedance mode on DC only available for low voltages?

I suppose what puzzles me is that the current drawn from the CUT goes up as the voltage goes up, so why can't the Hi-Z function work at higher voltages than the often-cited 10V?

[BillyO]

Check this thread:
https://www.eevblog.com/forum/metrology/how-a-precision-dmm-have-a-high-input-impedance-in-the-gohm-rang/

[TimFox]

The higher voltage ranges on DVMs use a resistive voltage divider to reduce the high input voltage down to a voltage that is within the capability of the Hi-Z mode, which is usually lower voltage than the battery or power supply in the voltmeter ("naked" op amp input).  10 Megohm is a typical value for the input resistance of such networks.
Fancier equipment, such as the Keithley electrometers, use more complicated electronics to "float" the amplifier guts up to a common-mode voltage of a few hundred volts.

[David Hess]

TimFox said everything I was going to say.

On a multimeter, high input impedance is only available as long as the input voltage is within the input range of the high impedance input buffer, which is typically +/-10 volts maximum, but some extend to +/-15 or +/-20 volts.  Above this, a resistive input divider is needed which limits the input resistance.

Electrometers which support a high input resistance at high voltages bootstrap their high impedance input buffer so that the supply voltage of the high impedance input buffer follows the input voltage, and then the high impedance input buffer drives the high voltage resistive divider.

[Someone]

Further to the posts above on floating/bootstrapping the measurement, SMUs can do this too without needing a dedicated for purpose electrometer instrument. Guaranteed input impedances achievable in the gigaohm range and up (limited by leakage current).

Actually making use of such high input impedances is challenging! small currents and capacitances really get in the way.

[Kleinstein]

For a voltmeter it does not hurt having high impedance. Usually the signal source is still not that high in impedance. As an example a 1 kohm signal source can get a possible noticable loading effect from a 10 M meter. With a > 10 Gohm meter one can usually ignore the effect, though it depends.
The loading is less an issue with lower resolution (e.g. 3.5 or 4.5 digits), but it becomes an issue with higher resolution, like 6 or 7 digits. So the higher higher resolution meters usually have the high impedance modes.

[TimFox]

Of course, some voltmeters for electrician's use had to add a mode for lower input impedance on AC volts because the contact protection capacitors on relays and switches could give too high a voltage with an open switch driving only the voltmeter input, while the old analog Simpson 260 VOMs had a moderate input impedance on AC volts.

[Kleinstein]

For the high resolution DMMs it is not su uncommon that the AC ranges have a low impedance, like 1 Mohm instead of 10 M for the higher DC voltage. In parts this is a compromise with noise vs source loading.

[TimFox]

Some Fluke DMMs (again, for electrician's use) have a low-Z selectable mode with 3,000 ohms input impedance on AC.

[bdunham7]

I suppose what puzzles me is that the current drawn from the CUT goes up as the voltage goes up, so why can't the Hi-Z function work at higher voltages than the often-cited 10V?

A more abstract way of thinking about it than the previous explanations would be to consider that the >10G spec is highly variable and really isn't a true resistance.  There is a small bias current, typically less than 100pA, and then a variable current that depends on voltage which may be very, very small (the >10G may be >>10G) and then there is leakage to ground and the environment.  So you can't just take a high-impedance +/-10V input and add on a 90G series resistor for a 100V range as your results would be highly unpredictable.  To get stable and accurate results, you need your source for the 10V input to have a reasonably low impedance and 1M is a good number for that--it results in a fairly low error from the input currents but still allows you to use a 9M series resistor for 10M overall on a 100V range.  I suppose you could use a 99M series resistor for the 1000V range, but typically they don't do that and just split the lower 1M into 900K and 100K. 

[TimFox]

A concrete way of thinking of this is to remember that the input amplifier (with the low bias current and other leakage currents discussed by bdunham7) is connected almost directly to the input in High-Z mode, with no input voltage divider in the path.
You can't safely apply greater than, say, 20 V to an integrated-circuit op amp with normal supply voltages.

[David Hess]

Some old oscilloscopes supported ultra high impedance inputs, and had the same limitations.  The Tektronix 7A13 vertical amplifier has a switch to disconnect the internal 1 megohm termination allowing "infinite" input resistance at 50mV/division and below.  The Tektronix 7A22 also has provisions to disconnect its 1 megohm termination.