Precision swiss army knife: the SMU (source-measure unit)

Have you ever read the datasheet for a semiconductor device and wondered “how are these graphs actually generated?”.

Datasheet characteristic curves for FDMC86450 – ON Semiconductor

On a production floor, companies may use large curve tracer mainframes that are designed to carry out sequential tests on lots of devices at once. The cost… well, lets just say you would rather be driving a well equipped luxury car.

Check out this beast! The Agilent 1505A… not exactly a bench instrument.

For a bench setting, these giant curve tracers are rather “overkill”. But what if you wanted to unlock insights beyond the datasheet at your bench? Lets say that a characteristic curve is not available in literature, or the part is not performing as expected (datasheets can be “optimistic” at times, and parts may vary from batch to batch). How can you do your own tests and really understand the exact properties of components that you have in hand?

Fortunately, companies such as Keithley, Keysight, and Yokogawa understand this need and make small versions of these power analyzer instruments (known as SMU’s – pronounced “smoo” – which is short for Source-Measure Unit) that fit right in with regular 1/2U sized bench instruments.

The Keithley 2400 – released in 1992 – an absolute classic, and most popular bench SMU

Great! I can have a ‘SMU’ on my bench. But what exactly can it do?

Well, a SMU is best described as a “combination” instrument. It features all of the following abilities in one box:

  • Precision voltmeter and current meter (often 6.5 digits)
  • Low noise, precision DC power supply
  • Precision DC electronic load

This means that a single SMU channel can operate in 4 quadrants of the power spectrum.

Operating boundaries for the legacy Keithley 2400 SMU – from user’s manual

The dynamic range and precision of a SMU also pairs well with its ability to digitize source and measure values with high fidelity, thanks to the careful mixed signal engineering that is necessary to achieve both analog precision and fast operating speeds. Most modern SMU instruments feature embedded computers than allow for all sorts of dynamic testing (including arbitrary waveform generation) while also performing data acquisition of source and/or measure parameters.

So what’s the downside? Why don’t I see SMUs in most labs the same way I see a scope or a regular power supply?

Cost. Source measure units combine precision, power sourcing and sinking, and high speed digital control and acquisition all in one instrument. They are remarkable pieces of engineering: a fusion of grey beard analog wisdom with new school application processing. Let’s take a peek under the hood of a few SMUs and savor the art within:

Keithley 2400 – Image from :
Later model Keithley 2450 channel board – Image from –

Agilent/Keysight B2900 series – Image from David Jones –

Take a look at a few of the part numbers on the ICs and other components. SMUs are not instruments built down to a price: the high level of integration and precision means that only top shelf parts and textbook design practices will make the cut.

And of course, Dave Jones of EEVblog has several informative SMU teardown videos, with plenty of insightful discussion about both design and capabilities:

Shahriar Shahramian of The Signal Path has also published several informative videos regarding SMUs, including a number of experiments to demonstrate the practical capabilities of these precision instruments:

As with most test equipment, taking your time to bargain hunt on eBay from reputable sellers is the way to go. I was able to find a Keysight B2902A which was sold as not working due to a failure to boot. I opted to buy this instrument as it is a 2 channel SMU, meaning that you can indeed use it to characterize 3 terminal devices such as transistors. It was also a 2017 model year instrument, with no physical damage to the outside. I was able to score it for 1/5th of the new price.

The instrument arrived in “as described” condition. I first plugged it into a fused outlet and tried powering it on. The fan began whirring at full speed, which was an important first step towards diagnosing the problem. For a 4 wire PWM controlled fan, if no PWM signal is present then it will go to full speed when power is applied. This told me that the power supply is working, but the control signal was not. I had a peek inside and noticed that the P500 CPU module was screwed down in a warped state… the board was literally bowed across the mounting screw holes. It looks like the metal shield underneath the CPU card was not properly fitted to the SMU channel board… and instead of removing the entire channel board to fix this, whoever put it together just jammed the CPU card into its slot on the front panel and ratcheted down the mounting screws. If I had to guess why the CPU failed, it is because it was installed in a physically stressed state eventually leading to failing solder joints and/or traces after a few years of being thermally cycled.

I carefully disassembled the instrument and inspected both channel cards. Thankfully, the 2 identical channel boards (where all the SMU magic is) were unmolested. However, getting the P500 CPU module online for the B2900A series is nearly impossible, not to mention the front panel board with the damaged connector. I performed a careful inspection and cleaning of the entire instrument, reassembled it correctly (unlike the previous person), and had it sent over to Keysight for repair. Working with Keysight was a pleasure, and they installed a new front panel board, CPU module, and performed a calibration. Surprisingly, even with my disassembly, cleaning, and reassembly, the instrument was well within its calibration spec. Just goes to show that the instrument’s precision is truly built into the design at the lowest level.

I’m ecstatic to now have a (like new) 2 channel sourcemeter in the home lab for 60% off of the sticker price! I already have a few experiments lined up, including some lipo battery testing, and low power bluetooth SoC firmware development. As the SMU can source up to 210 volts, it is also great for nixie tube I-V characterization and testing high voltage boost converter components. Of course, if one wanted to make measurements on the femtoamp level (10^-15 amps!) which this instrument is capable of, you would need to use guarded test leads (such as triaxial cables) and shielded test fixtures to minimize leakage and external fields between the DUT (device under test) and the SMU front panel. I don’t plan on needing this ultra low current resolution any time soon, so I will just make my own set of high quality shielded kelvin (4 wire) test leads with beryllium copper banana plugs, 10awg coolflex45 leads, and high quality copper grabbers from E-Z Hook.

SMU’s are a fantastic multirole precision instrument, especially if you can find one at an affordable price point. Both Keithley and Keysight make excellent bench source meters; I also have a bit of experience using a late model Keithley 2635B at work and I can assure you that it is an excellent performer, especially when combined with a computer for setting up test configurations and data recording. You cannot go wrong with a SMU from either brand!

Update 9/29/2020

When it comes to precision power instruments like SMUs, high quality test leads become even more important for good measurement integrity. You’re paying for each digit of measurement resolution, so why skimp out the actual connection between the DUT and the instrument?

For ultra-low current measurements, triaxial cables and shielded fixtures are needed to guard band your test signal from lower potential circuits, across which leakage will occur (recommended for measurements < 1 nA). Keysight Technologies has drawn a very simple and effective pictogram to illustrate this phenomenon:

Triax vs coax and measurement integrity at low current – Keysight Technologies

Most of my planned use of the SMU will be for power devices, so ultra-low current leakage is not of concern. However, voltage measurement resolution can and will be affected if performing 2 wire measurements at higher current levels. Test leads can have non-neglegable resistance depending on the desired voltage measurement fidelity, which will result in a voltage drop between the connection at the DUT and the measurement at the front end of the instrument. The solution is to use a 4-wire (also known as Kelvin) connection. 2 wires are used to deliver power, and voltage is measured across the other 2 wires with a very high input impedance at the instrument, minimizing voltage drop across the sense leads.

After doing a bit of research, I could not find a set of 4 wire (kelvin) test leads that really impressed me. Pomona does make a few different sets of kelvin leads (Pomona part #5940 and #6303), but they are not grabber-style as I would prefer. I decided to cook up a custom set of my own design. The ingredients are as follows for one (1) 36″ long, shielded, high current 4 wire banana to grabber test lead:

(2) EZ-Hook XCH copper blade grabbers: EZ-Hook website

12ft Coolflex45 10awg stranded, silicone insulated copper wire: Digikey

(5) BU‐3110410 nickel plated beryllium copper banana plugs: Digikey

3ft EPS300 3/4″ adhesive lined heat shrink – Digikey

3ft of 3/4″ Tinned copper metal braid sleeving: Amazon

3ft of 3/4″ PET expandable braided sleeving – Amazon

Some pictures of the first (2) sets of 36″ leads:

Building these leads takes time and patience, but the end result is a very high quality set of connections that should last years and preserve the validity of your measurements!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s