Cables for mesuring high ohms resistors

[jfphp]

Hello,
I am trying to mesure directly resistors 1 Mo and above (Fluke5450 calibrator, ESI standart resistors...) on a 3458A in 4 wires configuration with the well known Fluke cables 5440-7002/4 ; I am very surprised by the noise on the readings at full resolution (not coming from the 3458A : cal and certificate of conformity one week old), no neon bulb above the cables, no vibration and no HF noise coming from a SA, synthesyzer... Any idea ? Thank you.

[HighVoltage]

How big is the noise.
And the shield of your Fluke cable is connected on both ends?

I had a similar problem before and it definitely was RF noise from a neighbor.
Since then I am testing my environment each time before I do some ppm-level testing
and keep smartphones and DECT phones far away.
 

[Awesome14]

And the shield of your Fluke cable is connected on both ends?

I always connect the shield on just one end. Otherwise you can get looping. It only needs to be grounded once.

[VintageNut]

Two sources of error are AC noise and DC current leakage. Cables that shield against noise do not necessarily help reduce leakage while good cables that reduce or eliminate leakage probably will also shield against noise.

DMMs use a small constant current which is not a great way to measure a large resistance. Large resistances are best measured with a fixed voltage. Electrometers use a voltage source for a reason.

How much accuracy do you require?

My Fluke 5450A was calibrated with a Measurements Internation bridge. The low resistances use a constant current bridge while the large resistances use a constant voltage bridge.

You can create a bridge with a collection of well-known standards (Keithley 5155) and some ordinary instruments.

[e61_phil]

And the shield of your Fluke cable is connected on both ends?

I always connect the shield on just one end. Otherwise you can get looping. It only needs to be grounded once.

You should connect both ends to use the full shielding inside the units. To avoid loops you have the guard switches on the 3458A and the 5450A (and all the other calibrators)

[dl1640]

i'd say flunctuation is below 1ohm as i tested with wavetek1271 and 1mega standard resistor, cable is diy

how tight is your application need?

[Moon Winx]

Use two wire measurement, not four wire... use low-noise coax if available, ground shields at one point, ground resistor case, increase test voltage so current level rises above floor. Check resistor for cold solder or loose joints.

[e61_phil]

Do you use offset compensation? Perhaps you schould increase the delay.

[VintageNut]

Do you use offset compensation? Perhaps you schould increase the delay.

Offset compensation is for low ohms. The offset(s) that are being eliminated are on the order of microvolts.

For example, for a KE 2750, on the 100M ohm range, the test current is 0.7 uA in parallel with 10 M ohm, so the voltage will be around 7V. A few micovolts of offset is smaller than the uncertainty of the measurement range and offset compensation has no value.

For measuring 1 ohm, the test current is 10mA. Voltage developed will be 10mV. A few microvolts of offset will be much larger than the uncertainty of the range and therefore must be eliminated to make an accurate measurement of the resistor.

[Dr. Frank]

Hello,
I am trying to mesure directly resistors 1 Mo and above (Fluke5450 calibrator, ESI standart resistors...) on a 3458A in 4 wires configuration with the well known Fluke cables 5440-7002/4 ; I am very surprised by the noise on the readings at full resolution (not coming from the 3458A : cal and certificate of conformity one week old), no neon bulb above the cables, no vibration and no HF noise coming from a SA, synthesyzer... Any idea ? Thank you.

For these high values of resistance, you should use 2W Ohm only.
4W gives no precision benefit @ R >= 1M, but increases measuring time and noise.
A PTFE cable assembly with guard / shield is recommended.

The 3458As accuracy  specification is very poor in measuring 10M and above, but gives a rough idea about the value of the 5450A, for example.
See my article about the 5450A and measuring errors on high resistances.

OCOMP does only work up to 100k (on all instruments), so switch that feature off deliberately on the 3458A.

Frank

[e61_phil]

Hello,
I am trying to mesure directly resistors 1 Mo and above (Fluke5450 calibrator, ESI standart resistors...) on a 3458A in 4 wires configuration with the well known Fluke cables 5440-7002/4 ; I am very surprised by the noise on the readings at full resolution (not coming from the 3458A : cal and certificate of conformity one week old), no neon bulb above the cables, no vibration and no HF noise coming from a SA, synthesyzer... Any idea ? Thank you.

For these high values of resistance, you should use 2W Ohm only.
4W gives no precision benefit @ R >= 1M, but increases measuring time and noise.
A PTFE cable assembly with guard / shield is recommended.

The 3458As accuracy  specification is very poor in measuring 10M and above, but gives a rough idea about the value of the 5450A, for example.
See my article about the 5450A and measuring errors on high resistances.

OCOMP does only work up to 100k (on all instruments), so switch that feature off deliberately on the 3458A.

Frank

Sorry, of course it was bullsh... with the offset compensation at that range.

@Frank: Why does a 4W measurement increase the measurement time? I agree with you, that it has no benefit and may also induce some extra noise. But the time schouldn't be affected, imho.

@jfphp: You could try to use another well known resistor as shunt (for example 100k for the 1Meg measurement) and apply 50V (if I remember correct, that is the limit for the 5450A, but double check it before) to the series.
This way I measured some 20Meg HV resistors. And if I havn't done a mistake in my calculations, the uncertainty (1year) of the 3458A together with the 5440B gives below 15ppm uncertainty. This already included the measurement of the shunt resistors with the 3458A.

[Dr. Frank]

@Frank: Why does a 4W measurement increase the measurement time? I agree with you, that it has no benefit and may also induce some extra noise. But the time schouldn't be affected, imho.

You're right, my fault of thinking.

Frank