Keithley 2450  Zener diode test

Hi Rob,

Nice graphs! It is showing an incredible amount of detail. Clearly some very nice analog circuitry in the box. The log one is correct, without having an unusual/ fake y-axis - the easiest way to see this is to notice that as the values get smaller, it would need a very long axis (infinitely long) to reach zero amps.

When I initially saw the roadtest, I was wondering if the built-in screen could do sufficient justice to the information it could capture, or if it would need to be permanently tethered to a PC. At first glance these screenshots show that the built-in display and graphing routines are highly competent.

I am saying that they should just fold the log scale over around Y zero and have logarithmic negative current shown properly. If I am interpreting the current log scale properly it should fold over at the -7.5 V where it touches the Y axis zero.  Without having already seen the linear version of the same data you have no idea that the current shown as positive to the left of -7.5 volts is really negative.   I consider this a firmware bug until someone educates me as to why I am wrong (which is very often)

Hi Rob,

It is leakage current, but because of some offset inside the test tool, the measured current has the wrong polarity (which is why it is shown to flip at -7.xV rather than 0v). In a way you're trailblazing at testing out this, because it's rare to have the capability to go down to this incredible resolution. If it wasn't for the high-ish voltage it could even be dissimilar metals I suppose :-). I don't know what p.d. they could generate; some physics expert may know. These are just some ideas to help start to explain it, because I suspect it takes some different than normal thinking when the capability exists to measure such tiny values. I think there is a chance it may correct itself by increasing the time to maybe half a minute or so for the same number of steps (so that the box has time for a more accurate reading)  - not sure if that will help or not.

That quick test was done with the standard test leads from the front panel connections just to get familiar with the unit. So the first step is to re-do the test with triaxial cables from the rear so that the driven guard can minimize leakage currents in the cables. Also switch to 10NPLC and add some filtering. I also did not have the diode in a metal shield box during the test. It was not a glass body diode so I don't think that there were any photoelectric effects going on.  I have to make up some triaxial leads first.

I don't think you are wrong here, the negative current should be shown below the X axis - the graph as plotted shows two different voltages for each current level which is obviously nonsense. With a zener diode the user can work out what it really means but with some other devices this might be really difficult.

I'm also slightly concerned that there is no indication (on the graph) that the current offset error dominates the current readings below about 1nA - it might be interesting to devise some tests to explore this in more depth.

MK

Thanks for concurring on the log behavior. Most of that apparent current offset is probably from not using the traxial connections with the driven guard. The worst case measuring accuracy spec at the lowest resolution of 10fA on 10fA is +/- 50pA which is about half of the offset in the graph.  Also the fact that the trace exactly touches the Y axis before it goes the wrong direction leads me to believe that it is the leakage current of the standard test cables I used for the test added to the diode leakage. 
RR

Since the instrument is allowing you to take measurements with forward and reversed current on a single graph, the implementation could have gone several ways:

1/ flag an error and not allow you to plot the complete graph in log mode

2/ Display a graph with a broken y-axis

3/ Display two graphs

4/ Display a single graph as it currently does (as shown in the screenshot above)

5/ Display a single graph as in 4, but perhaps with two colors, one for each leg

I think 1, 4 or 5 is the best implementation - option 1 and the others have these disadvantages:

1 - Restrictive

2 - (a) May be open to misinterpretation, and (b) means less detail can be shown on a given sized LCD

3/ Less detail can be shown on a given sized LCD

But, everyone may have a different choice. Which one would you have chosen, or some option 6? Just curious, in case there is a better way of representing the data that I've not thought of.

I'm curious because at one time I was planning to design a home instrument to do a similar thing (cut-down of course...) and I would have drawn a graph in two colors, i.e. option 5 (and also curious because I'm always interested in graphs :-).

Since the instrument is allowing you to take measurements with forward and reversed current on a single graph, the implementation could have gone several ways:

1/ flag an error and not allow you to plot the complete graph in log mode

2/ Display a graph with a broken y-axis

3/ Display two graphs

4/ Display a single graph as it currently does (as shown in the screenshot above)

5/ Display a single graph as in 4, but perhaps with two colors, one for each leg

I think 1, 4 or 5 is the best implementation - option 1 and the others have these disadvantages:

1 - Restrictive

2 - (a) May be open to misinterpretation, and (b) means less detail can be shown on a given sized LCD

3/ Less detail can be shown on a given sized LCD

But, everyone may have a different choice. Which one would you have chosen, or some option 6? Just curious, in case there is a better way of representing the data that I've not thought of.

I'm curious because at one time I was planning to design a home instrument to do a similar thing (cut-down of course...) and I would have drawn a graph in two colors, i.e. option 5 (and also curious because I'm always interested in graphs :-).

IMO it is very simple, it just needs a negative log scale below zero.  The fact that technically you cant plot negative values on a log plot is irrelevant. just take the points that are <0 that they have plotted as positive and show them inverted below the x axis on a "negative" log scale. I have contacted Keithley tech support about it.

Here is a better setup for measuring the diode that almost eliminates the cable leakage issue

Much better now only about 18pA offset at the voltage reversal point.  Its current measuring accuracy is only +/- 50pA at this level so it is doing a pretty good job.