measurement application
that the user would not normally see indicated on the SMU display. In addition, the fineness of adjustment of the programming resolution (10µV) is overwhelmed by the inherent error of the source, so this level of resolution is unrealizable.
In contrast, for the Keithley Model 2400 SourceMeter instrument, note that the readback voltage closely tracks the actual voltage measured at the output terminals. You’ll also see that the readback voltage differs from the programmed voltage. One would expect to see a difference, given the source’s accuracy specs. These kinds of results should give you confidence that the voltage actually being delivered to the DUT is that which is expected. In addition, with the Model 2400, the source error does not overwhelm the programming resolution, as it does for the non-Keithley SMU. That means users can have the confidence to take full advantage of the fineness of adjustment of the programming resolution.
As this comparison shows, an SMU instrument’s programming resolution specification is not a good indication of its stability and overall performance. It also shows that the source readback results can be highly questionable. Therefore, when evaluating an SMU for your application, be sure to do some testing for yourself.
Measurement settling time, offset error, and noise can have a big impact on an SMU instrument’s performance, particularly in low current applications. The example illustrated in Figure 11 shows the results of two SMU instruments sourcing 200V with nothing connected to the input terminals while measuring the resulting current using each instrument’s built-in ammeter feature. This comparison offers a good indication of each instrument’s fundamental low current performance, and it’s an easy test to recreate on the test bench.
Note that the non-Keithley 61 ⁄2 -digit SMU (the blue
line) settles to its specified offset error of 50pA in about four seconds. The “bumpiness” of the data curve indicates measurement noise. In contrast, the Keithley Model 2636A (the red line) settles to its specified offset error of 0.12pA (120fA) in about half a second.
The smooth data curve indicates a distinct lack of measurement noise. So, based on the data, it’s obvious the Model 2636A will deliver a better measurement faster. In fact, at the point when the Model 2636A is settled and capable of providing in- spec sub-picoamp measurements, the non-Keithley SMU still has nanoamp-level errors. In addition, if you were to take a series of measurements over
time, the Model 2636A would provide more consistent results due to its fast, flat, and noise-free settling.
Note that, in either case, when measuring low current, the settling times drive overall test time. This is due to R-C time constants inherent in the overall architectural design of any SMU instrument. Therefore, an ADC running at sub-line cycle integration (for example, at 0.001 NPLC) won’t provide a faster measurement. Low current performance is very important for many semiconductor and optoelectronic applications, as well as in materials research applications such as nanoscale devices, graphene, etc. To understand the true measurement performance of an SMU instrument, it’s important to look beyond “headline” terms like 61
⁄2 ⁄2 digits or 10fA resolution. Figure 12
offers another comparison of the low current performance of the Model 2636A with the non- Keithley 61
-digit SMU.
Figure 12. It’s important to understand the difference between an SMU instrument’s actual measurement performance and its “headline” specifications. The table lists specifications from the data sheet; the diagram explains the offset accuracy.
The non-Keithley SMU is specified as having 61 ⁄2
digits and 10fA resolution. However, a closer look at the manufacturer’s specs shows that its bottom current range is 10nA and its offset accuracy is 50pA. The total accuracy of most instruments is calculated as the gain accuracy plus offset accuracy. Gain accuracy is typically given in % of signal, and offset accuracy is usually a fixed amount. The Model 2636A is specified as having 1fA resolution. The spec table in Figure 12 shows that it has a 100pA range and 120fA of offset
Issue 2 2012
www.siliconsemiconductor.net 29
Figure
11.Comparison of measurement settling time,offset error,and noise
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40