Datron 4000A output impedance issue?

[essele]

I've been slightly concerned about my 4000A's output impedance for a little while, but have worked around it for most of my use cases ... but, having just fixed it yet again (a board I missed when re-capping) I thought it might be worth exploring.

Firstly I'd like to check if I'm reading the specs correctly...



The third line down is the 10V range for the 4000A ... so I read this as an output impedance < 0.1mOhm and capable of 25mA.

So if I were to load a 10V output with a 1K resistor, then I should see a 10mA current (well within it's capability) and at 0.1mOhm I should see a maximum drop of 1uV ... is that right? Or am I missing something?

I see a drop of 260uV into 10K, and 30uV into 100K. So something isn't right?

I have the 4000A attached to a meter with normal banana leads, then at the 4000A end I put a resistor across the output, so the current through the resistor is not on the meter leads.

Before I go digging around, does my theory and approach sound reasonable? Or am I doing something daft?

Cheers.

[Mickle T.]

Low output impedance is for 4-wire connection only.

[guenthert]

I sense some miscommunication.  It would help if you'd describe exactly what you're doing and what you're trying to accomplish.  Perhaps also a picture of your set-up.

"The third line down is the 10V range for the 4000A .."  Well, we take your word for it, but the screenshot was badly cropped making it impossible (well, short of downloading the spec ourself and pattern matching the image you provided) for us to follow.

[essele]

I sense some miscommunication.  It would help if you'd describe exactly what you're doing and what you're trying to accomplish.  Perhaps also a picture of your set-up.

"The third line down is the 10V range for the 4000A .."  Well, we take your word for it, but the screenshot was badly cropped making it impossible (well, short of downloading the spec ourself and pattern matching the image you provided) for us to follow.

Hi guenthert,

I'm sorry for the spec image quality ... the spec for the 4000A (in all the versions I can find) runs across two pages and so you do have to match up range and values, all I did was crop off the white space ... but I don't think it changes the problem at all.

My issue is that I've noticed a marked drop in output voltage in a number of different situations when using my 4000A (the one that really springs to mind was when testing a Fluke thermal transfer standard (when you switch it in an out -- it's a relatively low impedance) therefore I wanted to see if my unit was within specs or not.

My reading of the specs suggest there should be an almost imperceptible voltage drop due to reasonable changes in load impedance (<1uV for 10V into a 1K load) however I'm seeing several orders of magnitude greater than that.

I will draw a picture once I'm back at home, and also test the 4-wire setup as this may make the whole thing moot.

Lee.

[guenthert]

      No worries.  I was just a bit slow on the uptake (and thrown off by the incomplete spec).  It seems Mickle T. and TiN understood you better.  You'll need to use both drive and sense pairs of connectors (there will be a significant resistance in series with the sense connectors), not unlike a lab PSU to which a low impedance load needs to be connected.

[essele]

Ok. I *think* I've found the issue.

I was experimenting with the guard terminal and saw some very strange behaviour where the output voltage actually increased when using the guard terminal and adding a load, that really didn't make sense to me, so I began to look at the output switching to see what was going on.

With remote sense off the 4000A should connect the sense lines to the output lines, however I wasn't seeing continuity on the low terminal (high was ok), and looking at the DC PCB schematic it looked like one of the relays (RL12) was most likely the culprit.

However on further inspection it was actually a fuse (F5) which had blown. This is actually in the power output path so it's surprising that there's any output at all, so there must be some path via GU or GU TERM (which I'm hoping explains the previous weird guard behaviour.)

I don't have any 1A PCB fuses and I didn't like the idea of using a 5A one just in case there was another issue, so I bodged in a 500mA cartridge fuse on some longish wires just to test it.

Initial results are good, in remote sense mode, I see absolutely no drop with a 100K load and just a few uV into a 10K load, a 1K load still shows a fair drop (about 100uV), but I do have the long wires and am seeing a significantly higher resistance through the wires and cartridge fuse than the datasheet for the pcb fuse shows, so hopefully this will be better with the proper part.

So I'm now waiting for the part to arrive and will then test again.

On the four-wire vs. two wire difference ... given the setup, the only real difference is internal to the 4000A, so wiring from the terminals via the motherboard, to the DC board and relays ... is that what you are talking about? If so, then that makes sense to me (no pun intended!)  Externally the meter is parallel to the load and doesn't share any of the path (other than through some banana plugs for the purpose of this test) so I don't see how that could make so much of a difference.

[essele]

All is now perfect.

The fuse was part of the problem (tripled the magnitude of the error) ... but there was also a significant element of user stupidity when I was testing the 4-wire mode ... as I'm using banana leads (for ease/speed) there is actually a relatively high resistance between the power terminals (I+/I-) and the sense terminals (Hi/Lo) and as I had transitioned from the 2-wire mode (Hi/Lo only) I had been attaching my load resistance to the Hi/Lo terminals ... which of course is incorrect.

When I attach the load to the power terminals (I+/I-) I see brilliant results ... not even a uV of difference, even at 1k/10mA.

And clearly banana leads are not appropriate for this kind of device ... interestingly I see about 60uV of difference if I switch the banana plugs for the load and the meter around ... which doesn't seem unreasonable at 10mA (6mOhm?)

[guenthert]

[..]
And clearly banana leads are not appropriate for this kind of device ... interestingly I see about 60uV of difference if I switch the banana plugs for the load and the meter around ... which doesn't seem unreasonable at 10mA (6mOhm?)

     The beauty of the 4-wire ("Kelvin") connection is that resistance of the wires doesn't matter at all (as long as the compliance voltage of the current source isn't exceeded and the internal resistance of the sense amplifier is much larger then the one of the sense wires).  The difference you are seeing might be caused by thermal EMF on the sense wires.

[TimFox]

Nickel-plated banana plugs can have a substantial thermal emf against copper and/or gold, and are not ideal for critical measurements.
See the .pdf file linked in this Fluke page:  https://us.flukecal.com/literature/electrical-calibration/watch-out-those-thermoelectric-voltages-cal-lab-journal-reprint