I was the first to publish my review of the Keithley 2450 SMU in July this year, however, the review itself was couched in a major caveat. This was the fact that the instrument was not performing correctly under certain circumstances and appeared to have some quirks which were still being investigated.


I have maintained contact (where possible) with Tektronix representatives since my first contact on 22nd May, providing further information where possible and also examining some of their responses. However, the review was not updated in the interim until now (in late-August) since there were no major noteworthy developments that would warrant a full-blown posting.


This post will look at some of the things I have been doing with the SMU in the interim (knowing it is still faulty and a bit “temperamental” at times) and what has been brewing on the Tektronix side of the fence as well.


CR2032 Battery Discharge Tests

One key experiment I wanted to conduct with the SMU was a battery discharge test on CR2032 coin cells. These popular cells have been used and abused in all manner of electronics including keyring LED torches and Bluetooth beacons to name a few. While they are ubiquitous, their performance can vary substantially and the current use cases fall well outside their intended use scenarios.


Most CR2032 cells are intended for use in memory backup and RTC power applications where the continuous draw is in the order of 5µA. A majority of these cells are specified for continuous discharges of 0.5mA at most. In many modern applications, actual discharge current can reach into the 10s of mA which means that the cells are being pushed hard.


I wanted to understand how much of an effect this might have on the usable capacity of a cell, thus I wanted an experiment that would discharge a cell at 4mA (a rough average of modern heavy-load applications) and at 0.5mA (which is the approximate maximum continuous load specified by some CR2032 cells). I thought of conducting this experiment a while back, but I did not because the B&K Model 8600 DC Electronic Load and the Rohde & Schwarz NGM202 have current setting accuracies with uncertainties of 0.05% + 1.5mA and 0.02% + 2mA respectively. This would mean that we have very little control and certainty over the current! Thankfully, source-measure units are much higher precision instruments and in that range of current values, the expected error was 0.02% + 1.5µA at worst (10mA range), making it worthwhile to conduct the test.


My first test was to discharge at 4mA. Two reputable CR2032 cells were chosen, a Panasonic and a Varta. Each test run was performed using a four-wire jig which I made from an old PCB holder and some gold-fingered battery spring contacts. Slight jumps in voltage may be due to vibration affecting the contact stability. Each run took approximately 34 hours.

Compared to the commonly quoted datasheet values for CR2032 cells which are around 225-230mAh, the cells delivered much less at the 4mA current. For a 2.0V cut-off, the Panasonic lasted slightly longer than the Varta, but the delivered capacities were in the range of 94-100mAh. Slower bumps and variations were due to room temperature changes which were uncontrolled predominantly affecting the battery, but remained within accuracy limits for the instrument.


To run the same test for a current of 0.5mA would take much longer, around 19 days. As a result, just for verification purposes, I performed the test with the Panasonic cell only.

The cell is specified for a capacity of 225mA, with the test showing it delivered about 236mAh which is slightly above the claimed capacity. This clearly illustrates that the cell definitely can deliver the specified charge, but only at lower rates.


For all of these tests, I performed data capture using the onboard features alone, dumping the buffer to a USB memory device. The 2450 remained stable, despite running continuously for a total of 24 days without any issues.


Epishine Light Energy Harvesting Module Tests

Recently, I did a review of the Epishine organic-PV light energy harvesting evaluation kit through direct contact from element14. The Keithley 2450 SMU was definitely a key player in running the I-V tests in the second review posting, being able to measure microamps with clean, smooth precision despite the sub-optimal set-up.


Unfortunately, while things appear all nice and smooth, my experience with my particular 2450 was a bit perplexing. With the PV panel exposed to a fixed source of artificial light, running an I-V curve using my own SCPI code (as my KickStart 2 trial has expired) for the first time resulted in strange “bumpy” curves with values that were initially below the expected value, suggesting that on certain ranges, the unit really wasn’t measuring what I was expecting. Was this a quirk with the auto-zeroing on the unit, an intermittent contact or something else entirely?

I couldn’t get a firm answer on this, but to say that if I re-ran the test three times, usually the last result actually looked right. Perhaps it was a contact issue somewhere that settled on its own or by vibration or slight movements. If I didn’t want to just keep trying until it seemed okay, rebooting the unit restored the right output first-time. I really don’t know whether this was because of the unit or some external influence, but it really did bug me that I couldn’t understand why it was happening. The panel was connected with the supplied test-leads which had a very “springy” retracting hook setup which I would have expected to be sufficient for the currents involved.


New Firmware Release – More to Come!

Since the review, a new firmware version 1.7.3 has been released which fixes a number of issues and adds some improvements. The changelogs from the release notes are as follows:


I was keen to apply this update to see if it would make any changes to the issues I had experienced with the SMU. The process of updating was straightforward and identical to the last, taking several minutes and was ultimately successful.

Unfortunately, the overheating problem persisted (which is no surprise) but also I experienced a one-off glitch where the unit booted up thinking it was under remote GPIB control (with nothing attached) and the output became uncontrolled. Attempting to take local control caused the unit to lock-up. Fortunately, this issue did not persist after a reboot.

But perhaps the biggest good news to come out of this firmware update was an acknowledgement of an issue with using high-impedance mode and/or four-wire sensing mode where readings would stop when the output was turned off but not restart when the output was turned on. I received confirmation that this is not the intended behaviour and instead this firmware fixes it for resistance mode but I had picked up on the other two cases that were not patched. The bug was apparently introduced in version 1.7.0 – with such a complex instrument, the software is quite hard to audit!


As a result, we expect a fix for this issue in an upcoming version 1.7.4 firmware release, which is good news!


Further Quirks?

Tektronix were helpful enough to investigate my reports of unusual or unexpected behaviour to try and replicate or explain them. Where they were able to provide advice, it has definitely been enlightening.


One observation that I had made was that the output power-up time seems to fluctuate between a fast and slow speed. It was explained that in auto-ranging mode, the supply may have reached compliance (i.e. current limiting) in a low range as it charged up any capacitances resulting in a “slower” ramp to full voltage or incurring some delay due to the auto-ranging operation. In this case, I did not expect or understand the intimate relation between sourcing and measuring ranges nor how auto-ranging delays could cause measurable differences in the output (as most power supplies don’t have many ranges to speak of), so in this case it was a clear case of mea culpa!


Unfortunately, perhaps due to the sheer number of queries and their limited resources, I have not had any clear explanation for “Overflow V” being recorded on auto-ranging or I-V Characterizer runs, nor for parameter error messages in I-V Tracer, the trial-date message for I-V Tracer or the voltage regulation tightness on a load-ramp test which seemed a little wide. My personal suspicion may be that these issues could be particular to my unit or test setup as I observed this perplexing result as well in a range that usually works just fine –

A limit of 100mA is selected and sourcing 0.196V is requested, but the unit hits a limit sinking 0.109mA (which implies it’s actually in 0.1mA range rather than the requested 100mA range) thus the voltage sits above the requested voltage.


I’ve also encountered strange incidences where toggling the range back and forth while re-setting the same current limit each time results in significantly different readings – it may limit when changing one way, but repeating the change provides the right result without a limit annunciation. I suspect there may be a bug or issue with my unit here and have forwarded a video along with .tsp setup file by e-mail, but it has not been examined as of yet. Other times, strange readings occur but after rebooting, it works just fine (e.g. in the I-V curve testing of the OPV cell above).


Unfortunately, the process of actually trying to follow-up on these issues by e-mail is taxing and the round-trip-time can range from days to weeks. I feel that perhaps the issues may have been handled in a better and more systematic way with responses communicated on a query-by-query basis rather than “ad-hoc” based on whoever has some idea as it stands now.


At this point, given the uncertainty about the sanity of this unit, I’ve discontinued use and hope that the remedy of an RMA will have it fully fixed, calibrated and tested.


My 2450 Prepares to Depart for Surgery

As the observed overheating messages while using the 200V range were rather permanent and concluded to be hardware related, the unit would have to be returned to a service centre for repair.


We were fortunate to receive an explanation from the representative as to the cause, which I have paraphrased. The cause of the fault was due to a change in the parts of the ranging circuitry aimed to improve instrument response. It caused unintended oscillation when operating on the 200V range at particular voltage and current levels which caused the instrument to falsely believe it was overheating. This was not detected as the routine auditing used fixed combinations of voltage and current that did not coincide with the issue. A fix has been developed and validated which uses passive components to correct this without affecting the specifications of the instrument and improved auditing procedures which cover all ranges and quadrants have been instituted to ensure such issues would be caught in the future.


Unfortunately, as I did not purchase the unit, the process was rather complicated and involved a number of staff at Tektronix performing internal processes to clear a path for the unit’s return and repair which took a few weeks. Furthermore, I would have to ship the unit to Singapore (presumably at my cost), whilst locally sold units may have the support and backing of our local distributor and avoid such hassle.


The process was not as smooth as expected, for a number of reasons related to the Australian Government regulations, our postal services and the information provided. My first hurdle was simply the declared value of the parcel which exceeded our postal service’s thresholds and required me to apply for an account with our border force to obtain permission for export, with everything geared towards commercial entities and requiring a lot of work. The easier route appeared to be eschewing the postal system entirely and sending the unit via a courier service. This is when I discovered the postal code provided was incorrect, which took a day to clarify. Even then, certain bits of information such as the HS Tariff Code were not provided and I had to obtain it from the original shipping documents. The RMA was missing an RMA number, which I had to request from them, adding another day to the process. Granted this is an exceptional circumstance, but I expected things to be a bit easier … I guess it’s common to underestimate the challenge!


Just when I thought I had found a service that would economically bring my parcel to Singapore, it turned out that after I keyed in my card details, it would fail at the last step because the package was overvalued, but the limit was not documented anywhere. I tried again with a lower value, and it “failed” while claiming to have succeeded. My card was not charged and no confirmations were sent – but Couriers Please was a division of Singapore Post, so I expected things to just work.

I had to contract a different courier (DHL) to carry my parcel – at least they accepted my order and generated a waybill, but ultimately I did not opt to insure the parcel because the cost with insurance was about AU$550 (~US$400), but the cost without was about AU$330. That’s definitely more than I’ve ever spent on a RoadTest on out-of-pocket expenses – in fact, prior to the pandemic, I could literally hop on a plane and fly to Singapore and back for a lower price on a sale fare!

The unit was safely packed, taped up and weighed. It was then unsealed, as DHL requires the unit to be open for inspection and they will seal the consignment. Bubble wrap was added to ensure extra protection for the screen, while padding a little slack space because of the box expanding slightly due to prior shipping.

As of now, the unit still resides at my premises, awaiting pick-up on Monday to begin its long journey. Assuming all goes well and there are no issues with paperwork or customs, the unit should arrive in Singapore where Tektronix will forward the unit through to a service centre in the USA for it to be repaired, recalibrated, tested and returned to me. Hopefully this will permanently resolve the observed quirks and issues experienced during the RoadTest.


But perhaps the biggest anxiety I have now is whether it will arrive in one piece, whether it will be properly fixed and whether I’d have to pay inbound duties for the returned item. If I do, that would be another AU$700-or-so out of pocket.


While this is related to the RoadTest, I don’t expect element14 to foot the bill for this one, since they have (technically) fulfilled their obligation of delivering the item to me for review and likewise, I have fulfilled mine in providing a comprehensive review as close to the proposal as possible within the timeframe given. In this case, I am pursuing repair as part of my intent to perform further experiments that have been curtailed but also to understand how a properly functioning unit would perform. Because of this and the value the equipment provides to me, I have decided to cover the cost of returning the unit myself. At this point, I am not too fussed what happens to the unit either way – my review has been completed and in its current state, it’s really not that much use to me as I don’t have that much confidence in the results from the unit unless I carefully massage the ranges and test its response prior to running experiments. On the upside, Tektronix have extended their appreciation and goodwill, promising to supply full KickStart 2 licenses and some accessories with the repaired unit.



In this follow-up, I was able to present the results of a long-term test looking at CR2032 battery capacity. The SMU showed its benefit in its high measurement precision for current, making this test a possibility compared with more conventional instruments. The results show the detrimental effects of using CR2032 batteries above their rated current, resulting in a marked decrease in available capacity.


Testing was also performed with the SMU on some organic PV energy harvesting modules, where the low-current measurement capability also proved itself to be valuable. Unfortunately, during testing, there was some unexpected glitches in readings which may have been due to the unit, test setup or something else entirely. It seemed to resolve itself when the test was re-run a number of times or if the unit was reset.


Between when the review was first released and now, one firmware had been released to resolve a number of issues. One issue where the SMU does not return to continuous measurement after toggling the output in 4-wire sense or high-impedance mode was deemed to be a bug and related to one that had been fixed with a future release on the cards to correct it.


Replies from Tektronix have been helpful, providing insights into some of the issues which were mea culpa, primarily due to my unfamiliarity with the instrument’s behaviour. Other issues, sadly, have not received the systematic attention I had expected and were perhaps not reproduceable. The protracted back-and-forth of e-mails is rather taxing and I no longer see any good reason to try pushing some of these quirks as the unit itself may be atypical.


Most importantly, an RMA has been set-up and the unit will be soon returning to the manufacturer for corrective service to repair the issue causing the 200V range to trigger overheat shutdowns. The process of returning the unit, sadly, will cost me a significant amount of money and the process was complicated by issues from postal agencies, governments and incorrect/incomplete information from Tektronix that took some back-and-forth to clarify. I hope the unit reaches its destination safely and is returned fully fixed … preferably without me having to shell out even more to cover the customs duty.


I suppose it will still be some time yet before this story comes to its final conclusion, but at least something significant has happened. In the meantime, I'll be working away on other RoadTests ...