Does little jumper has a friend? (YES CONFIRMED)

[mimmus78]

This week I build others revision of my references and now I have a total of 4 LTZ1000 references.
Now comparing the four LTZ1000 measurement charts I noticed a little jump up and down of 1uV that repeats every few minutes in a pattern that is present with all LTZ1000 I have. This is really annoying and no matter what I do this pattern keeps repeating in every design I do.

Now with all 4 LTZ1000 references having the same annoying problem I started to think that the internal reference inside the 3458A could be the source of this jump. With this in mind today I compared the one year old KX reference, back to back with my just built ovenized LTZ1000 and results seems to confirm that the jump is internal to the 3458a.

I attach here some sample of my ovenized LTZ1000 (HBDR107A) vs 3458A and of KX back to back with my ovenized LTZ1000. I select this 2 of four because they exhibit the lower noise among the four.

PS: in second chart there are 3 little jumps at offset 3.5, 4, and 5.5 hours those a due to automatic autocal not due to references. In the first chart there is only one autocal at 12.25h. All measurements are at 100 NPLC.

What do you think?

[Andreas]

Simple test:

use 2 of your LTZ-references and do a differential measurement in 10V-range of your instrument.
If the noise stays the same or even doubles sometimes -> your references are affected.
(Double check that its not your DMM buffer amplifier by shorting the input).

If the noise is gone (smaller by factor 10-1000 depending on difference voltage) -> your instruments reference is affected.

with best regards

Andreas

[Jorn]

Try substitute the reference with a 7V battery. Keep cable layout the same. Does the noise disappear with the battery?

I also noted that in your back to back measurement there are jumps that indicate minor EMI problems. (I'm not referring to the acal jumps).

Try using ferrite beads (clamp on type) on every cable in your setup. This includes the power supply cable on your 3458A.

I use small RF Toroids (if i remember correct the Amidon FT50-43) on both the negative and positive wires from reference pcb to the output terminals on my references. In addition ferrite beads from power supply to reference pcb.

Use twisted pairs everywhere you can to minimize magnetic coupled noise.

Not likely the reason but check for sensitivity to line voltage variations. 

Kind regards
Jorn

[mimmus78]

Noise at 10V - 100 NPLC with shorted input seems ok. I went up after the automatic autocal, but than it went down to 80nV RMS average.

I'll try to bodge the ratio python script if I do not fell asleep on the desk ...

[TiN]

  Looks like jumper A9 in 3458A indeed . If you have another meter, sensitive enough, you can hook it to 3458A A9's output and test it out that way.
Jumps with shorted input wouldn't happen, because reference level is invisible there.

[Andreas]

Noise at 10V - 100 NPLC with shorted input seems ok. I went up after the automatic autocal, but than it went down to 80nV RMS average.

I'll try to bodge the ratio python script if I do not fell asleep on the desk ...

Hello,

So you have checked that its not the input buffer of the DMM. The voltage reference(s) can be still affected.

Why do you need a ratio script?
You can use 2 of your LTZ references with ground pin connected and measuring the difference in 10V range.

With best regards

Andreas

[mimmus78]

Hello,
So you have checked that its not the input buffer of the DMM. The voltage reference(s) can be still affected.
You can use 2 of your LTZ references with ground pin connected and measuring the difference in 10V range.
Andreas

Well this is how was obtained first chart.
Reference where running on insulated power supply (KX on a mains powered with transformer, HBDR107A on batteries).

- Ground pin of both ref connected.
- The two 7V connected one to HI and one to LO.

I attache here also this night run with ratio measurement.

Noise in ratio mode is around 250nV RMS with some spike, a little bit more than when it was back 2 back but still I cannot see any multiple minute 1uV jump.

I was expecting ratio mode to have similar noise than back to back ... I have to think at this.

[Andreas]

Well this is how was obtained first chart.
Reference where running on insulated power supply (KX on a mains powered with transformer, HBDR107A on batteries).

- Ground pin of both ref connected.
- The two 7V connected one to HI and one to LO.

Hello,

overlooked that. So it really seems that the DMM-reference is the guilty.

with best regards

Andreas

[TiN]

If you using 100mV range for back-back test, then your reference noise impact is also reduced x100 times?

[Andreas]

If you using 100mV range for back-back test, then your reference noise impact is also reduced x100 times?

Hello,

At 45mV difference voltage (45% of range) it should be around factor 150 against the 10V range (71% of range)

With best regards

Andreas

[Jorn]

I'm still not convinced that it is the DMM reference that is the problem.

Why? Because when measuring 45 mV difference in the 100 mV range the ADC is at 45% of full range.
When measuring directly 7.2 V the ADC is at 72% of range. The periodic nature of the "jumps" should be visible in both cases but they are not.

A direct battery measurement would reveal if jumps are present or not....

Are your references buffered? If not a small capacitive load might cause the jumps in a direct measurement and not in a back to back measurement.

Finally, if you dare, you can use one of your references as 3458A reference .

[Alex Nikitin]

I'm still not convinced that it is the DMM reference that is the problem.

Why? Because when measuring 45 mV difference in the 100 mV range the ADC is at 45% of full range.
When measuring directly 7.2 V the ADC is at 72% of range. The periodic nature of the "jumps" should be visible in both cases but they are not.

On 100mV range these "jumps" , if originated from the HP3458A reference, should be about 7nV each (for 50% input).

Cheers

Alex

[Jorn]

I'm still not convinced that it is the DMM reference that is the problem.

Why? Because when measuring 45 mV difference in the 100 mV range the ADC is at 45% of full range.
When measuring directly 7.2 V the ADC is at 72% of range. The periodic nature of the "jumps" should be visible in both cases but they are not.

On 100mV range these "jumps" , if originated from the HP3458A reference, should be about 7nV each (for 50% input).

Cheers

Alex

You're right. My bad. Please disregard post.

[Kleinstein]

.......
Well this is how was obtained first chart.
Reference where running on insulated power supply (KX on a mains powered with transformer, HBDR107A on batteries).

- Ground pin of both ref connected.
- The two 7V connected one to HI and one to LO.

I attache here also this night run with ratio measurement.

Noise in ratio mode is around 250nV RMS with some spike, a little bit more than when it was back 2 back but still I cannot see any multiple minute 1uV jump.

I was expecting ratio mode to have similar noise than back to back ... I have to think at this.

The short spikes in ratio mode also support the noise comming from the 3458 internal reference. During a jump in the reference voltage the two measurement used for the ratio can use a different reference and thus give a single value that is significant off, either high or low, depending on the direction of the jump. It may happen at a time when not effecting the readings, so no need for a regular sequence of positive and negative spikes.

[mimmus78]

So let put a nail in it.

I hoked up a couple of LIPOs, waited a few hours for the LIPOa to stabilise and this is what i got.

It's not jumping as the TiN one, but still pretty annoying with more than 1uV at 8V (it will reach 1.25uV at 10V).

PS: T INT REF is temperature of another reference and has noting to do with 3458a.