RoadTest: Keithley 2450 SourceMeter SMU Instrument
Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: None
What were the biggest problems encountered?: Some issues in the graphing mode and difficulty in loading instrument type into the Kickstart program. Some of the scoring categories don't apply but I gave a 10 to not sway the rating.
Keithley 2450 SMU Review Part 1
Disclaimer, I have no previous experience with a SMU other than studying their use and functionality and dreaming of having one!
If you are not familiar with the basics of a Source Measure Unit (SMU) here is Keithley’s What is a SMU?
The Keithley Low Level Measurement Handbook PDF. on pages 1-8 and 1-9 show why Keithley is calling this a SourceMeter® SMU Instrument
since it has both “Source Measure Unit” (SMU) and “Source Meter® Instrument” features. The 2450’s Source Meter® Instrument attributes are;
directly displaying resistance and power and only having a >10GΩ sense input impedance (a SMU typically has >100TΩ sense input impedance).
The 2450’s SMU attributes are the very low current measuring capability since it has extended low ranges with 10nV and10fA measuring sensitivity.
SMU’s are more typically used in automated testing applications since the tight integration of the precision sourcing and measuring functions allow
simpler and more accurate test setups. SMU’s are also used as Lab bench instruments but the 2450 touch screen interface has made
its standalone bench use much friendlier and more efficient. With that in mind I will be centering my review on looking at the 2450 mostly
as a standalone bench instrument that can provide the functionality listed below.
Although not an inexpensive piece of gear at ≈$5400.00 it could be argued that it might be a less expensive
more integrated alternative to the separate instruments it can function as:
0.015% 6.5 digit DC voltmeter with 10nV resolution and >10Gohm input impedance
Picoammeter with 10fA resolution and a < 100µV burden voltage on all ranges
20W Precision current/voltage source/sink with 500fA and 500nVresolution (100mA at 200V and 1A at 20V)
Precision 20W electronic load.
6.5 digit Ohmmeter with 1µΩ resolution and >200GΩ.
Flexible real time graphing/data logging and extensive standalone programmability.
Semiconductor curve tracer.
I will expand on some of these in part 2.
Here are some links to additional SMU information:
Electrical specifications 2450 Data sheet
2450 Users Manual Sorry, you will have to register or log in to download.
2450 Reference Manual Sorry, you will have to register or log in to download.
General overview video Keithley2450 product tour Video
Comparing SMU vs power supply
Comparing SMU vs DMM
Great resource on additional SMU details Agilent Parametric measurement handbook PDF Sorry, you will have to register or log in to download.
The video segments contain more detail than the written review. Each section below will have its related video segment links after the heading.
The whole index is here for reference.
Video segment index:
The shipping box is high quality heavy double layer cardboard with heavy custom foam inserts. You could drop kick the package
with no fear of damage. It came with a power cord, usb cable, Ethernet cable, test leads, safety interlock connector, Software/manual
disks, 2 quick start guides, calibration certificate, and other documentation. The instrument case is very solid construction of what
appears to be16Ga steel in a one piece rectangular tube with a lap seam on the bottom. The rubber bumpers at each end are nicely molded
quality material and fit snugly. Tilt bail works as expected and is needed for a decent view and working access of the touch screen.
Viewing angle is pretty good up to about 45 degrees off axis. The screen is effective but some of the text/background color choices on the
smaller display items could use a little more contrast IMO. The back lit buttons and indicators are very effective and the unit can easily be
used in very low light conditions. The buttons and navigation wheel/button have a good quality feel amd response. The rubbery buttons are a very close
fit in the panel holes and slightly off center on my unit and will occasionally stick in. But it does not cause a functional problem in use.
The front panel jacks are very solid feel with gold plated machined jacks. Triaxial BNC connections are on the rear panel with a driven guard
for minimizing leakage currents. Overallvery nicely executed and fit and finish are high quality. Boot time is about 12 seconds.
Front Panel with the default home screen showing.
The upper display can be set to measure voltage, current, resistance, or power. The lower display shows whatever is being sourced, voltage or current.
All of the communication interfaces are included as standard
Related video segments Front Panel Touch Screen Interface
To me the 2450 is very intuitive but it really took a while to get all the nuances of where everything is and all it can do. I found the reference
manual necessary to find/understand every nook and cranny of all the features and menu screens. The help screens are effective and fairly
detailed but if you don’t realize the help only works on the highlighted top levels before you get to a selection level you might think the help
button rarely works. The engineering of the entire menu/display is very well done. There is a lot of functionality on many of the screens yet
none feel cramped or cluttered. There is nothing that jumps out as needing improvement as far as navigation goes. The touch screen response
is good on all the general tap and sweep motions. But the response on the graph screen needs some fine tuning. Keithley tech support has been
extremely responsive to the bugs and improvement suggestions I have reported so far. They definitely display an open "we want to make it even better"
attitude. The navigation control knob works well but I rarely have the urge to use it over the touch screen.
Below is an example of the help screen for the home screen. The 5 views are of the same screen just scrolled down through the content.
Home screen with the default source display swipe screen
Help screen for home screen
More of same same help screen
The Settings swipe screen with many of the measure parameters settable here.
The Statistics swipe screen uses the buffer set on the data buffer page not the graph data page
User display swipe screen. as is says content is remotely programmed
Data Trend swipe screen, an up swipe will give the full graph screen shown in the next section below.
Related video segments Navigating the General Use Screens
I consider "general use" to be voltmeter, ammeter, ohmmeter, power supply/E-load, data logger, data trend chart, and system functions.
This is the menu screen and I will focus on the Quickset under Source Measure, Source Settings, Measure Settings, Graph and Sheet under Views, and System.
These are all the screens that you would need to use the 2450 as the above instruments.
This is the Quickset menu which allows setting the source measure functions, Resolution,
and the One-touch Voltmeter, Ammeter, Ohmmeter, and Power supply setups
This is the Function screen where you can set sourcing to voltage or current and then pick what will be measured including power.
This screen is also directly accessible with the dedicated function button
The Source settings screen shows the current source/measure setting in the yellow box (source voltage measure voltage).
All of the other settings are accessed here. I like how the range of values is shown for each setting without having to access the function.
The Measure settings screen shows the current source/measure setting in the yellow box (source voltage measure voltage).
All of the other settings are accessed here. Again I like how the range of values is shown for each setting without having to access the function.
The View Graph screen will show whatever tab was last accessed which in this case was Graph. Here you can pan and zoom using finger gestures. This is very functional but needs some work on the gesture response and it can show erroneous trace shapes. (watch the video segment)
The Data tab allows which buffer to use and to clear the buffer. What each Axis is using as input is also set here.
I can't tell any difference in the marker modes on the screen output. I spent half an hour convincing myself the dual linear sweep function didn't work.
Then I realized that when the X axis is set to source the up sweep and down sweep are on top of each other.
When I switched the X axis to time I could see both the up and down trace on the graph.
The Scale tab allows setting the scaling and position for both axes and linear or log on the Y axis. If you end up lost in space with the zoom and pan on the touch screen you can come back here and set auto for both axes and find your missing trace!
The Views Sheet screen brings up the data sheet showing the buffer contents.
The System column of the menu page has the following:
Event Log- Logs all system events and allows save to a usb stick. Setting available for what is logged and what warnings are shown.
Communication- Settings for all the various communication protocalls
Settings- Various setting for backlight, password, key sounds, command set, etc. etc.
Information- Serial number, firmware version, cal. dates, and number of cals.
Manage- Firmware up/down, System reset, demos, etc.
Related video segments Navigating the Advanced Use Screens
These are all the screens and functions not covered above and are Filters/math, Sweeps, Triggers, Scripts, and related configs.
Filter and Math screen where filter count/type and math functions/values can be entered.
The other items can be changed here and on the settings swipe screen
Sweep settings screen where a source configuration list will automatically be generated. Set all the sweep parameters then you must press generate.
Beware if you generate a dual sweep that you must use time for the x axis not source on the graph data screen if you want to see a graph that looks like the picture.
This is an example of the source configuration list generated automatically.
Here you can edit and manage source configuration lists.
Measure configurations can be set by setting all the measurement parameters you want on the 2450 and then come to this screen and use system to list and it
will make an entry into the active list. All the measurement parameters can be seen in on the details of each entry.
The trigger system can use specified source and measure configuration entries to generate automated testing applications. This includes using external inputs and generating outputs. The flow tree can be generated and edited right on this screen without any remote access needed.
Scripts are a way to store all of the source, measure, and system settings to a named script.
Then a script can be selected and run and all the settings are restored the script value.
This allows rapid changes to commonly used setups.
Scripts are lost when power is lost so they need to be saved to a usb stick.
This menu allows saving and retrieving scripts.
This is where you create script names. You have to set all parameters manually
with all the appropriate menus then save to a script name.
In part 2 of my review I will compare the 2450 performance to a 24ppm 6.5 digit DMM and a dedicated Micro Ohmmeter on voltage, current, and resistance.
This will be both paper specs and actual measurements. Also compare performance/ease of use as a precision bench supply vs a HP6115A precision lab supply. Emphasis will be made on when and where the 2450 excels or not.
Keithley 2450 SMU Review Part 2
Setting your zoom level to 200% will allow seeing more detail on the following charts
This will cover the 2450 as a Power supply, Voltmeter, Ammeter, and Ohmmeter. I show how the 2450 specifications compare with the Agilent B2901A SMU, my Fluke 8846A and my IET LOM510A micro-Ohmeter. The last two are the only calibrated instruments I have to do accuracy comparisons with. The Agilent B2901A is a close match at $500.00 more than the 2450 and is higher performance in several areas but is less accurate on paper. All the specs have been charted in a way that graphically show where each instrument is most accurate. The measurement comparison charts have a spec ratio column that show which instrument should be the reference and how much more accurate it is. I include the ancient HP6115A power supply in the power supply testing since its specs are very close to the 2450 and even better in some parts of the ranges. This 6115 is as received from fleabay with its last calibration in 1992. There are lots of things a metrologist would pick apart about my comparisons but I am just trying to give a general feel about how this unit performs. 2450 readings were done at 10NPLC and a repeat filter of 10 and then stat average using the number of readings show on each chart. The 8846A was set for 100NPLC and a stat average of the number of readings show on each chart. This was done to minimize noise and give stable readings at these high resolutions. 8846A accuracy is the same on 10 and 100NPLC
I have used the touch screen even more in doing this testing and I really love this interface. I can’t think of a more efficient system for entering settings. The milli, micro, nano, x10, and /10 buttons on the input screen are absolutely brilliant. I have found I rarely want to use the navigation knob even though it works well. This screen pops up for both voltage and current settings/limits.
Use as a power supply in Voltage mode and as a Voltmeter:
This is the stat screen of sourcing 1mV for 101 readings (roughly 6 mins). Notice the 10nV resolution and a span of 170nV and SD of 31nV. This is pretty good stability in my book.
This is the stat screen of shorted input and zeroed with relative function for 658 readings (roughly 36 mins). Notice the span of 210nV and SD of 37nV. Only slightly higher span and SD than the 6 mins above.
This is the response of changing the voltage value from 1V to 5V and back to 1V with no load. Nicely behaved
This is the output switch on and off with a 5V setting. Also well behaved. Notice the approximately twice as fast rise time compared to the voltage change above.
Some points on all the spec charting:
These are plots of the 1 year specs of maximum error as a percent of the measured value. This puts all specs on an even playing field for comparison. The jagged data jumps are the natural effect of (% of reading + offset or % of range) at the range changes where percentage and/or offset values may be different. The other subtle data kinks are the effect of the value data not being the same step in value from decade to decade. What appear to be discontinuities of data from one chart edge to the next is a range jump happening right on the chart edge.
The SMU’s are very close and better than the 6115A in this range. The 8846A is over 40x more accurate than the other instruments in this range up to 10mV and progresses to 2.2x more accurate at the 200V level. This makes it the reference for all measurements in all ranges on the voltage testing. Unfortunately it won’t be that clear cut on current and resistance comparisons.
The 2450 and the 6115A are significantly more accurate than the 2901A in the 2 to 20V range.
The 2450 and the 6115A are significantly more accurate than the 2901A in the 20 to 200V range. The 6115A even beats the 2450 measure spec in the 20 to 80V range.
A little guidance on this mess. All the blue columns are 1 year specs in ±volts for each value. The yellow column is how much more accurate the 8846A is compared to the 2450 measure for each value. The light green columns show how much percentage of the spec is used if the 8846A had no error. To go to the other extreme, in the magenta column I ignore the extremely good 8846A Cal. data and add the full 8846A spec error to the instrument error and use that to calculate the worst case percentage of spec used. In both cases the percentage of spec used is impressively low. The 6115A was last calibrated in 1992 and is still in spec (also very impressive). I used 2 short coax cables with 4 coax to male banana adapters to connect with remote sense right at the 8846 input terminals. I think the outer shields of the coax being connected to the 8846A measurement common helped the noise be lower than using 4 independent banana jumpers.
Use as a power supply in Current mode and as an Ammeter:
This is a stat screen of sourcing zero current thru Force high and Force low triax cables that are connected at one end The measure reading was zeroed out using relative. A span of 110fA and a SD of 19.7fA over 62 readings is amazing. You can’t get here without the triax and driven guarding of the rear terminals.
With current things get a little more complicated. The 8846A is only equal or more accurate in the 1A to 100mA range. After that the 2450 is progressively more accurate as the measurements get lower. At 10nA and below the 2450 is approximately 500X more accurate than the 8846A. Another area where the 2450 excels in current measurement is its very low burden voltage of ≤100µV on all ranges. This is an example of how difficult spec comparison would be without this chart style. Look at how the 2450 source and 8846A 10NPLC cross over each other multiple times on which is more accurate.
Clearly the 2450 excells here and is more accurate than the 2901A in the 110pV thru 1A range. The 2901A does not have separate source and measure accuracies. On the 2450 with source readback on, the sourced value is measured at measurement spec accuracies before it displayed and used in any calculations such as resistance or wattage measurements.
The color coding here and chart explanation is the same as the similar Voltage chart above. This is the accuracy measurements where the 8846A is the reference. Obviously you have to take the spec ratio into account as to how much weight you give to specific values. This was done with Force high and Force low triax cables directly feeding the 8846A in current mode. Considering the last Cal data of the 8846A it is reasonable to use the “error as percent of spec assuming no error on the 8846A” column as a general measure of accuracy. Very nice results with a high of only using 8.6% of the allowable spec. Even if you use the magenta worst case column it is still well within spec.
These are the accuracy measurements where the 2450 is the reference. Again you have to take the spec ratio into account as to how much weight you give to specific values. But the 2450 is clearly the reference here. The 1.2 and 1.3 ratios seem to be anomalies but they are cause by the 8846A having 20% over range where the 2450 has a 5% over range. This is most noticeable where the spec ratio is less than 1.5 but exists in all the ratios. The offset or % of range part of the specs has the least effect at the top of the ranges. The measurements at 100nA and below were done with both the 8846A and 2450 offset currents removed with relative or zero. The 2450 worst case source error as a percent of spec is all within tolerance even when adding the full measure spec error to the measured value. The 8846A also does very well with very low percentage of spec numbers especially below 100µA.
Use as an Ohmmeter:
Quickset gets you to ohmmeter and allows setting 4 wire mode. For maximum accuracy source readback needs to be on and offset compensation needs to be on for low resistances to minimize the effects of thermal emf’s. This chart shows the LOM510 is clearly the reference for 2 ohms and below. The 2450 has better resistance specs than the 2901A in all ranges.
Here we see the 8846A is clearly the reference from 20Ω to 10MΩ
All the 2450 resistance measurements were done using using full 4 wire triax kelvin connections to the resistances. Source readback was enabled along with offset compensation. The <2Ω values were done sourcing 1 amp. The 8846A was also 4 wire kelvin connections and offset compensation on the low values using my offset compensated ohms adapter. The IET LOM510A is 4 wire kelvin connections and used the offset compensated ohms mode. The resistances from .1Ω to 20MΩ were my .1% decade box. The three lowest resistances was a piece of wire with the sense connection spacing adjusted to give the nominal value as measured by the LOM510A. The spec ratio of the LOM510A on these values is certainly high enough for this to be a valid method. My connection system allowed me to connect each instrument to the setup without disturbing the exact spacing of the 4 kelvin clips. This chart is strictly a general comparison using the best spec instrument as the reference for the relative accuracy of the other two. The best spec instrument is dark blue background with the second best spec instrument in light blue background and the worst spec instrument with no background. All I can say here is all the instruments agree pretty closely and generally do much better than their one year spec. I finally realized the 2450 has a fan that kicked in when sourcing the 1A on the low resistances. It’s audible tone says variable speed and its seems to track the demand closely.
I am very impressed with the 2450 performance and ease of use. Entering or changing values is really nice and very quick. The graphing capabilities are very useful as is and I believe will be improved even more with the next firmware release. Even if you could buy separate programmable instruments of equal resolution and accuracy for the same money, I don’t see how it could be as effective as the 2450 with all the built in integration of source, measure and graphing. Since complete test routines can reside on the 2450 it can be a stand alone test and measure unit that can even control external equipment such as binning of components. If you don’t use a SMU for a living there is a lot to learn to really be aware of all the capabilities that are packed into this unit (I am not there yet!). The setup scripts are an important tool to have correct repeatable setups. There are a lot of variables to be set for both source and measure and they don’t get remembered when you turn the unit off unless you store them as scripts and reload them from the usb. This will probably be my most used instrument on the bench because of its extreme versatility.
I will post a video link of general use functionality of all the above source and measure functions in the next couple of weeks,
Thanks again to Element14 and Keithley for such an awesome opportunity. I feel very blessed to have a unit like this in my lab.
Thanks for your response; I didn't expect to see your correspondence, John mentioned that you had made some good recommendations along with other users, so I was curious as to how much user feedback is implemented in future firmware releases, such as the forthcoming release at months end.
I'm also quite impressed with Keithley's customer service, but it will be interesting to see how far Keithley will go in providing their customers with what they have asked for, and how quickly those changes are implemented.
Anyway, thanks again for your response, it's much appreciated!
Part one of the review was trashed by the editor as I added part 2. I still have to fix that.
I had a lot of suggestions and bug reports. It would be difficult to enumerate all of it here and I don't think it would be appropriate for me to post all my email correspondence with Keithley. My focus was on using the 2450 as a standalone instrument so I was beating up on the graphing heavily and I found some things that were not behaving as they should. I was also focusing on performance not speed and many of the bugs/improvemenst I found were only noticeable when at 10NPLC and higher repeat filter counts. I am very impressed with the attitude of all the Keithley people I have communicated with. It will be interesting to see how much of "my input" gets into this firmware release. I say "my input" very loosely since there is nothing saying I was the first person to bring these bugs or improvements to Keithleys attention so I cant take much credit.
Firstly, great review! You done a thorough job there, although the formatting can make the text hard to follow in places.
Anyway, I have been in contact with John Tucker and he mentioned that in the forthcoming firmware release, there will be some product included which came as a recommendation from you, along with improvements and bug fixes. I was interested in knowing whether you could share your recommendations with us, as it would be interesting to see how much user feedback made it into the next firmware release, as well what we can look forward to.
I'll just cross-link here to Robin's comments in another Test & Measurement thread which discussed ways of making video more educationally effective and useful than a single monolithic lump of video generally provides. As a result, Robin has made an already very good video presentation even better, and I think he should be commended for it.
It would be nice to see a good use case applied in the review - some kind of low-power test/verification which would seem what the SMU is targeting, I had the perfect project to test an instrument like this - a design around an ultra low power wrist watch.
I am focusing mainly on using this as a standalone instrument. I don't have a background in controlling instruments by remote connection so my opinions would be pretty useless. Hopefully one of the other 2 reviewers will be able to evaluate that properly. I do plan to hook it up and learn but not for the review. The spec sheet link and manuals should tell you if this thing has the functionality you need. I will post a blog about a quick accuracy check I did.
Nice start - I didn't look at the videos because I'm interested in how well it works - I'll worry about the interface if it can do the basic job. I liked your initial blogs about it. Are you going to test out remote control - I can't imagine using an instrument like this other than connected to a computer to at least log the data but more likely control the whole experiment.