0

Keithley 2400 Verification Procedure Errors

I think the verification for the Keithley 2400 series SMU's has a calculation error on the upper and lower test limits during the DC Current Measurement test. Has anyone seen this?

 

 

 

21 comments

Date Votes
0
Avatar
Michael J.

Hello Jim,

My guess is that the result is being guardbanded, so the reported limits are reduced according to the guardbanding method used. At the top of your results, is there a guardbanding method mentioned?

0
Avatar
Nathan Ryder

Correct me if I'm wrong but the reported limits should still reflect the actual specs, The guardbanded limits only effect the pass/fail statments as in it could be a pass inderterminate or a fail interterminate...

 

I could be wrong here though

0
Avatar
Michael Schwartz

The report should be able to reflect both limits.  So the technician and end user know why the UUT is failing. 

This whole 4 to1 Ratio and Guardbanding is about managing risk in the measurement process.  

In my opinion guardband, indeterminate pass-fail and risk calculations should all be done post calibration, not in the automated script. 

0
Avatar
Nathan Ryder

Yes it should, The end user should be able to make the risk decision as to use it or not with all the information they have 

0
Avatar
Michael J.

The fields that appear on the test report are determined by the format of the Crystal Report used to generate the report.

If you want your test report to show both the original and guardbanded test limits, alter the report to include both sets of fields.

If you don't want the procedure to guardband to determine the reduced test limits, turn off guardbanding.

0
Avatar
Jim Vaspoli

I did the caluculations for the 1µA range and it doesn't seem to be right to me. I'm attaching a better document to show all the ranges and tolerances.

Attachment not imported: test_errors.pdf
0
Avatar
Jim Vaspoli

The tolerance for that range is ±0.029% +0.0003 units

0
Avatar
Jim Vaspoli

The test results show a lower tolerance of 0.949403 and upper as 0.950554.

I came up with a lower tolerance of 0.94882 and an upper of 0.9499718.

0
Avatar
Michael J.

Hi Jim,

Gold Procedures use a default guardbanding method of Root Difference Square (mentioned at the top of your test report that you attached there). I'm finishing up an article for the Fluke Support knowledge base on the subject, but you can change the default setting to match your local quality policy. If you don't want to guardband the results and instead want to use the direct test limits, open your gb.ini file (it will be located in your Shared Directory) and change GB=RDS to GB=OFF. Once you make that change, all the Gold Procedures you use will not use guardbanding.

0
Avatar
Michael J.

It looks like the 1 uA test point is using a system actual value of 0.949979 uA to calculate its limits, which is the value measured by the DMM. Based on that, the test limits, without guardbanding, would be 0.949029 and 0.950929 (0.00095 from the system actual). The procedure is guardbanding those limits with the Root Difference Square Method to reach to calculated limits on the report.

0
Avatar
Jim Vaspoli

I think the test has the 2410 source the current and report what it thinks it's sourcing and uses that as the nominal. Then takes a reading with the system DMM and compares the two.

0
Avatar
Nathan Ryder

We guardband in every single procedure we run here and use the more conservitive method of MU (done to satisy our auditors) and every procedure we have validated we check the reported limits are correct and they always tie up with the actual published specs...As in they are not being adjusted to suit the new GB'd values, I'm thinking maybe his report template might be looking at the wrong thing.

As per MJ's reply... "If you want your test report to show both the original and guardbanded test limits, alter the report to include both sets of fields."

Have you actually checked the procedure to see whats going on at these points?

0
Avatar
Michael J.

Jim opened a ticket on this with the software support group, so I went through the code with him there. The system actual value is being pulled from the DMM, so that's what's being used to determine the test limits, and then as expected, it's guardbanding the limits.

0
Avatar
Nathan Ryder

Yeah that makes more sense..But is it reporting the modified GB'd limits or just the limits based on what the DMM read?

0
Avatar
Michael J.

It's reporting the GB limits. The default test report that comes with MET/CAL is set up to show the guardbanded limits if it's guardbanded. 

0
Avatar
Jim Vaspoli

The problem occured in the DC current measurement function. During that test is the system actual pulled from the meter or the UUT?

0
Avatar
Michael Schwartz

Yes,, This is something that will drive you NUTTS when validating a MET/CAL procedure.

% of Reading is not the same a % of Nominal. 

Mike

0
Avatar
Michael J.

In general, when the device under test is a device that generates a signal, the system actual is referenced to the value measured by the standard.

0
Avatar
Michael J.

It is, the difference comes in what value is the reference for the calculation.

That's why the VSET/TSET TOL_REF parameter exists, though.

0
Avatar
Nathan Ryder

Well technically speaking the DMM is the actual so MET/CAL is operating correctly

Just out of interest does the procedure use MEMC/MEMCX for this test point?

0
Avatar
Michael J.

It uses MEMC. This is the recommended choice for tests where the DUT is generating a signal that is being measured by a standard, generally speaking.

Please sign in to leave a comment.