3458a and usage of guarding and the guard switch

[aronake]

I am trying to figure out how to correctly use the guard terminal and garud switch for voltage and resistance measurement. But there are some conflicting information in different manuals.

3458a Calibration Manual say on connecting 10K resistance:

"Connect the resistance standard to the multimeter's front panel HI and LO
Input and HI and LO Sense terminals as shown in Figure 3-3. If using a Guard
wire (as shown in Figure 3-2), set the Guard switch to the Open position. If not
using a Guard wire, set the Guard switch to the To LO position."

So they recommend leaving the guard floating. I find that rather odd.

Same for voltage calibration:

"Connect the voltage standard to the multimeter's front panel HI and LO Input
terminals as shown in Figure 3-2. If using a Guard wire (as shown in
Figure 3-2). set the Guard switch to the Open position. If not using a Guard
wire, set the Guard switch to the To LO position."

The 3458a user manual say about guarding:

"The measurement connection illustrations in this chapter show the multimeter's
Guard terminal connected to the low side of the measurement source (guarded
measurements). This configuration provides maximum effective common mode
rejection (ECMR) on the input terminals selected by the Terminals switch,
assuming the Guard switch is in the Open (out) position. For non-guarded
measurements, depress the Guard switch (TO LO position) and do not connect the
Guard terminal to the measurement source. In the TO LO position, the Guard
switch internally connects the Guard terminal to the LO Input terminal on the
terminals selected by the Terminals switch. This configuration provides reduced
ECMR."

The for pictures of measuring voltage and resitance with guard, guard is connected to low on voltage source or resistor and guard switch should be open.

The 742b manual say though:
"Make sure the GUARD terminals of all
interconnected instruments are tied to earth ground at one point and one
point only in the system, and all LO or COM terminals are tied to GUARD
at only one point in the system."

Summary:
3458a calibration manual. Guard to Guard, Guard switch open.
732b manual. Guard to Guard. Connect Guard to ground at one point. Connect Guard to Low at one point (i.e. use the guard to low switch on 3458a, or probably better, short ground, guard and low on 742b side)

I like what the 742b manual says most....  But I would think those who wrote the 3458a calibration manual know what they are writing. And they have had like 35 years to update it if something was wrong....

Anyone have any more thought on this? Any experments made?

[Dr. Frank]

Hello,
I've experimented a lot with different voltage standards, with and without proper Guard shielding, connected to the 3458A..

I have a FLUKE 5442A, which has case ground and separate inner shield, a FLUKE 7000, which only has a case ground, and probably you know my DIY LTZ1000 references, which are built with an inner tuner box case, which is the GUARD, inside an outer aluminum box, which serves as a case ground, isolated from the inner tuner box.

I use a shielded 2 wire PTFE cable, and connect this shield cable to both guard shields, leaving the switch open, so the inner shield of the 3458A  is connected to the inner shields of either the 5442A, or my LTZ references.
For the Fluke 7000, I simply do not connect the Guard wire at the reference side, only at the 3458A.
I did not see any difference in reading , when this wire was connected either to the case of the F7000, or to the minus connector of the reference voltage (what some other volt-nuts are doing).
I can only tell, that all my measurements were much less noisy, using the shielded PTFE cable and only connected to the GUARD of the 3458A.

I suggest one practical thing, that you try out the different possibilities of connecting this Guard in the different configurations, and test, if you see any different readings, on the order of 0.1ppm or less is to be expected.

Similar thing I encountered with resistance measurements, I always use the Guard wire at my DIY resistor boxes, which only have a single case shield and connector.

Please do not "trust" the old 3458A manuals much, as HPAK have never updated their original operators and calibration manuals and also not the specifications since 1989. Several specification parameters should have been improved, like a more stable 40k internal reference resistor since ~1995, which has better T.C. and better 1yr. stability.
Ohm transfer stability is missing in the 3458A datasheet, whereas FUKE did specify this, when they offered their 3458A HFL version, which included also that better 2ppm/yr. reference.

The FLUKE manuals however are not specifically intended for use with the 3458A, therefore, they had to make more general proposals, not knowing what would be the best solution for your use case.

So again, just try out on your own.

Remark:
This shielding and guarding problem is also quite interesting, and you can learn a lot, if you are using the FLUKE 720A KV or 752A reference divider in conjunction with the FLUKE 845AR/AB and the 732A to calibrate other equipment. In this very sensitive bridge configuration, it is extremely important to use the correct guarding/shielding scheme, otherwise you will get great deviations and/or very noisy zero readings, by ground and shielding currents.
Please check the manuals, there are errata sheets , how to choose for battery and line operation of the 845A.
That's also important, if you would use a long-scale DMMs, like the 3458A, or the 8508A as a null instrument.

I have described these problems in my threads about the 720A and the 752A repair and operation.

Frank

[julian1]

So they recommend leaving the guard floating. I find that rather odd.

My guess/understanding -

I believe 'Lo' means short the lo terminal input to the internal 3458a guard node.
While 'Open' only means floating in the limited sense - that it creates the possibility for the operator to make this connection with a guard wire to the DUT lo/gnd.

In this configuration, CM currents generated from the 3458a transformer screen/internal shields - are preferentially diverted through the guard wire to the lo/earth of the DUT.
And they don't generate voltage potentials on the test leads, due to lead resistance.

A confusing point from Fig 3. is the implication is that the resistance standard, has a valid earth.
But perhaps that was normal for resistance standards at the time.

I can't see the guard switch shown in the 3458a schematic.
But perhaps that's because its function is merely to short two inputs, so it mostly exists as a panel wiring/assembly item.

[Echo88]

@julian1: The guard switch is shown in the schematic on page 53 middle right section: https://xdevs.com/doc/HP_Agilent_Keysight/3458A/doc/3458A%20CLIP.pdf

[julian1]

Good eyesight. 

p57 suggests W1(1) is floating, to create an Open switch position
Probably GRD_LO is 'guard common' coming from the input terminal mux switch.

But what is the purpose of CR003 - a 6.2V diode (zener?) to ground.
How does the diode make sense when the max input voltage limits are much higher?

Perhaps to rectify CM voltages, so they are always positive wrt gnd/PE ?. 

[Wallace Gasiewicz]

Here is a link to the HP Application note 123  for the HP 3456 about Floating Measurement and Guarding.  I would attach my copy but it is too large.    I know this is old stuff but Maybe this will help someone.
https://www.qsl.net/iv3cve/hp/an/pdf/an_123.pdf     

The most important statement is  "Don't leave the Guard Open"     

[Bill158]

More application notes on Guarding.  Again, always tie the GUARD terminal to somewhere, don't leave it open!  On the other hand (left?) don't exceed the maximum voltage between GUARD and GROUND!  Also don't exceed the maximum voltage between LO and GUARD.
I hope this doesn't add to your confusion.
Bill

[aronake]

Thanks to all for good input here.

Dr Frank always helpful with a long and informative answer. Do you have a gofundme page or something similar to contribute to your test equipment budget as a thank for all insights you are sharing?

To get more clarity i did some test measurements. 3458a connected to 742b. And some different connections.

Cable used is PTFE 2 wire with shield, where the sild serve as guard conductor.

Guard to Lo on 732b - This would be as the 3458a manual recommend connecting for voltage measurements with guard.
Guard to Guard on 732b - This would be as how Dr Frank measure is LTZ1000 that have guard.
Guard no connected to 732b - Guard connected on 3458a but not on 732b.
Guard to Guard short to Ground - 3458a guard to 732b guard, then short ground and guard on 732b. This would be as the 732b manual recommend to connect.
Guard not used - Guard not connected anywhere.

Then for each above I put the guard switch on 3458a to "open" and "to Lo"

1000 measurements at NPLC10. Average of voltage and PPM standard deviation. Also PPM for the different voltage measurements Vs Guard switch open and guard connection to Lo on 742b.

The results:



Except for Guard to Lo on 742b and guard switch to lo (which turns into a 1 to 3 PPM disaster, but not higher standard deviation) I can not with significance see that any of the other alternatives do any significant difference. I had some temperature fluctuation issues in my lab, so that probably explains the up to 0.15 PPM difference for the other measurements. And i may not have given the connections enough time to for temperature to settle properly.

If i do same voltage measurement at 10V range with shorted inputs 3458a #1 show 0.030 PPM standard deviation, and #2 show 0.025 PPM.

I also tested a couple of the different configurations on two other of my 3458a and saw very similar results. Guard to Lo on 742b and guard switch to lo gave 1 to 3 PPM lower measurements and no significant difference on anything else.

Maybe for lower voltages in more noicy environment how to use guard makes more of a difference.

[aronake]

Here is a link to the HP Application note 123  for the HP 3456 about Floating Measurement and Guarding.  I would attach my copy but it is too large.    I know this is old stuff but Maybe this will help someone.
https://www.qsl.net/iv3cve/hp/an/pdf/an_123.pdf     

The most important statement is  "Don't leave the Guard Open"   

This was a very interesting and educative read!

But having read this text, I am even more surprised of the recommendation in the 3458a calibration manual to connect 3458a guard to voltage standard guard and leave guard switch open. This would have made sense if guard and ground was connected on the voltage standard. But just having guard unconnected is explicitly warned against in this HP3456 text, which I think makes sense. Hummmm....

[Wallace Gasiewicz]

Guard "OPEN" means the guard terminal is not connected to LO, so you should connect it to something else. The common use would be  the Guard would be connected to the LO of the source voltage.  Connected to LO (by the use of the switch) is probably the most common use of voltmeters we use. 
These efforts to minimize common mode usually are only visible on low volt settings. Like the 100 mVolt setting.  That is if they are visible at all.

As Dr Frank points out, there are two issues, one is "Guarding" and the other is Shielding. He explains how to shield and guard with the same connections.  You need a good "shield" for this, probably a piece of coax with the inner conductor removed...so you can put your wires thru it.

[aronake]

My mistake when testing guarding on Fluke 732b using 3458a to measure was ofcourse that 732b have an extremely good design with very limited need for extra guarding. When trying some less ideal sources with less good transformers, i see that guarding makes much more of the difference its supposed to do.

[Wallace Gasiewicz]

I wonder what the effect of using multiple twisted pairs for the test leads would have on voltage  readings.Just wondering.......