RoadTest: Become a Tester of the Toshiba e-Fuse IC Evaluation Kit EVB-TCKE805NA
Author: Gough Lui
Creation date:
Evaluation Type: Semiconductors
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?: eFuses, Load Switches, Current Limiters from a number of other vendors including ADI, NXP, TI, On Semi, LittelFuse, MPS, Reed Semi, Semtech, Vishay, STMicroelectronics, Alpha-Omega Semiconductor.
What were the biggest problems encountered?: Design of evaluation board was not ideal for easy evaluation of item, requiring SMD soldering to perform certain evaluations.
Detailed Review:
Toshiba TCKE805NA eFuse Evaluation Board RoadTest Review
By Gough Lui – October-December 2024
Fuses are an integral part of most electronic devices, employed as the last-line of defense to avoid catastrophic outcomes in case of a fault. Rather than attempting to save the attached circuitry (as it is often misunderstood), they are usually designed with a higher current level than the operating circuitry could consume under all circumstances with some margin for safety to avoid nuisance trips. It is simply there waiting for a situation that, whether due to component failure or due to overstress due to power supply problems, excessive current flows. By interrupting this current flow, it prevents fire and excessive smoke to improve safety, reduces stress to the upstream power supply network to minimise disruption and enhance repairability. If it does avoid collateral damage, that’s usually a bonus.
Fuse technology has been dominated by the cartridge fuse, an integral unit containing a specific type of wire that has a known blowing characteristic (e.g. fast, slow) and current-time relationship (I2t curves). These are inexpensive and fairly reliable, but can only be used once. Resettable thermal-magnetic circuit breakers have been used in higher voltage/power applications where resettable protection is required, for example, in a distribution board. But for electronics circuitry with lower voltage and power, polyfuse (PTCs) are the primary self-resetting alternative. Unfortunately, while these have definitely seen their uses (e.g. in computer equipment, USB hosts and devices), they often have higher resistances causing power loss and an inconsistent trip value (which depends somewhat on prior trip behaviour).
That being said, the joke in the electronics repair trade is that “semiconductors blow to protect the fuse”, alluding to the fact that fuses are slow, while silicon usually gets damaged quickly. Then, it might seem on the face of it, that an electronic fuse (or eFuse for short) based around semiconductors might seem a silly idea at first. But in reality, what if we could use the speed and consistency of silicon to produce an even better fuse? Clearly, Toshiba thinks so, so let’s try the TCKE0805NA eFuse solution and see how it behaves.
Toshiba’s TCKE8xx series is a range of eFuses that feature:
Based on this description alone, it can be ascertained that the eFuse appears to be targeting common electronics/IT applications, especially with the 5V and 12V clamping versions that are common voltage rails in computer power supplies. The design of the solution that integrates the forward FET minimises the amount of supporting components necessary – mainly a forward current programming resistor, a soft-start programming capacitor, and the necessary ESD protection diodes (as this is a semiconductor device) along with bypass capacitances. Its 18V input is perhaps a slight limitation, especially as consumer products move towards higher DC voltages (e.g. USB-C PD standard voltages include 20V, EPR includes up to 48V).
From looking at the datasheet, some of the key advantages of such an eFuse solution includes the relatively tight repeatability of the trip current setting, the speed at which the trip occurs (fast-trips at 1.6x Ilim of 150ns specified in the datasheet) is much faster than conventional fuse solutions, the possibility to program different current limits using the same part by changing resistances may make BOM management easier. Furthermore, the EN input with its accurate voltage threshold can be used as both an undervoltage lock-out protection and also to turn the eFuse into a load switch, which can be useful in case of rail sequencing for example. Additionally, the soft-start slew-rate control helps to control inrush current, especially useful in case of hot-plug scenarios to avoid rail brownouts when charging up input capacitances. Finally, the external FET driver allows for reverse polarity protection when used with an external N-MOSFET, which adds additional value as such requirements may exist in certain circuits to avoid back-feeding especially when multiple power sources are available with slightly different voltages (e.g. a battery and DC power device). While all the FETs will add some level of resistance, this can reasonably be in the order of <100mΩ depending on the chosen FET, which is less than most polyfuse PTCs and could be less than even a wire-based cartridge fuse especially for lower current ratings. The voltage clamping feature is also interesting, limiting overvoltages to a sensible non-lethal level where possible.
In all, all of these advantages may mean that such an eFuse could replace a load switch, a UVLO circuit, a polyfuse (or ordinary fuse), a reverse-polarity protection diode (or ideal diode), a soft-start circuit and a TVSS voltage clamp/crowbar (and perhaps even do the job better). This makes it quite a desirable solution especially considering the tiny space this chip occupies.
On the other hand, the trip behaviour appears to be related to an internal limiting causing heat generation, relying on a thermal process to initiate the trip. This makes the over-temperature protection essentially “free” in such a device, but also may lead to concerns about lifetime as silicon degradation at high temperatures is a known phenomenon. As fuses are rarely ever called upon to do their job, I don’t expect this to be a major issue in practice. However, if the fuse is an auto-retry type or one with voltage clamping, it could stay “hot” for long periods especially if the product continues to function normally or in some capacity despite the fault. This is something designers should be careful of, along with its ESD sensitivity as a semiconductor device. As a result, while it is very capable, there may be instances where the eFuse is used as a secondary protection instead, backed up by an ordinary fuse in case the eFuse were to fail. This negates some of the benefits, but still retains the flexibility and other functions provided by this particular eFuse. Another potential complication is that the eFuse requires good power to operate and may not operate correctly on reverse current whereas a real fuse doesn’t care either way.
However, it’s important to note that this is far from the only eFuse on the market. From my quick look around, I was able find many other eFuse series from various vendors, some of which are summarised below:
As a result, it would seem that designers looking for eFuses may well be spoiled for choice. Many of these have very specific application targets – e.g. computer 5V, 12V buses, 12V automotive, industrial, etc. As a result, it may take some narrowing down to find the one that has the mixture of features that you may need for a given application. Sometimes they are hidden as functions of a power switch, hot-swap controller, current limiter, etc so searching for eFuse on its own may not find all solutions. Because applications and products are so diverse, I’ll have to leave the detailed comparison of products as an exercise for the reader.
The board arrived in a white thin-card box. It would seem from the labels that the board has probably been in stock for around a year before this RoadTest. It is apparently Assembleed [sic] in Japan.
Inside, the item is ensconced in large pink static-dissipative bubble-wrap.
The inclusions are the evaluation board inside a blue static-dissipative bag, a printed Japanese manual and a printed English manual. This is printed on a large piece of paper, consisting of 10 pages of documentation. The size of the paper makes it a little difficult to navigate, but the small print could be a bit difficult for some to read. Unfortunately, there are no terminals or header-pins included in the package.
The board appears designed in a “Japanese way”, with the input on the right and output on the left. This is the opposite of convention in many western countries, which go left to right. That being said, there is also a misspelling of the word “sense” as “sence” in the silkscreen. The board is designed as a rectangular 66 x 46mm two(?)-layer board of unspecified copper weight. The board design itself doesn’t appear to be available.
Design features include exposed copper areas for attachment of wires. This is a good idea as it allows for better testing with high current sources. However, the use of terminal blocks may be preferable for convenience. These connections appear doubled-up using thermal-relief vias for connection to 2.54mm header pins (not included). Provision of sense connections improves measurement accuracy, which is a nice addition for those willing to go the extra-mile with a four-wire connection to accurately characterise the chips’ losses.
However, the design is a little strange. The EFET and EN/UVLO jumpers need to be connected with jumper shunts (not included) to allow the EFET gate signal to pass through, or for the EN signal to be sourced from the EN terminals or the voltage divider for the UVLO feature, which isn’t actually populated on the board. The other header connections are actually paired (i.e. connected together), which seems strange especially for ILIM and dV/dT as they don’t allow for overriding the existing mounted values, instead, at best we can just parallel another resistor or capacitor. So it seems surface-mount soldering is a necessity for using this evaluation board fully. Had they designed it such that multiple common values can be jumpered in or out, or the actual connections for the resistor or capacitor were broken out on both sides such that an external value can easily be wired-in, that would simplify evaluation.
The eFuse IC itself is pretty small, especially considering the chosen bypass capacitors. The board has the necessary ESD protection and output-side reverse polarity protection components installed as well.
The current limit programming resistor, as provided, is a 35.7kΩ resistor corresponding to a current limit of about 2.98A.
Meanwhile, the provided soft-start inrush-limiting capacitor is 120pF, which is expected to provide a speedy ramp-up. It is unusual they should choose a capacitor with such a large footprint for this value, especially when it doesn’t have to handle the full supply voltage (the datasheet implies 3V is all that is applied).
The underside does not have any componentry, although the exposed paddle vias do show some evidence of solder flow-through.
Further documentation is also available through the RoadTest page and also on their website. The documentation is acceptable, although there are places where remnant Japanese exists in the English documentation and occasionally, sentences can be a little difficult to understand.
To prepare the board for testing, it was important to attach some header pins to the board for attachment of measurement test leads and to attach wires for the main power input and output. For a more challenging test, I opted to use thick 2.5mm² cables and a high-current power supply (50A) to ensure that any overshoot (if any) would be observable.
After preparation, the board looked as above.
To test the over-current trip behaviour, I decided to supply the circuit from an Aim-TTi QPX750SP power supply from 4.4V to 18V with a B&K Precision Model 8600 DC electronic load sweeping the load from 0-5A. This allows me to build a surface plot, indicating the output voltage as a function of load current.
The above plot shows that at the expected operational voltage of 5V, the actual current limit is about 3.08A. The output voltage tracks the input voltage, but reaching about 6.14V, the circuit clamps the output voltage, preventing the load from being exposed to overvoltage. This is a little above the 6.04V typical quoted, but not significantly so. Because the circuit is acting as a linear regulator, the current available until thermal shutdown tapers down as a function of input voltage and is dependent on the board design and thermal dissipation. For 5V circuits, exposure to 6.14V is usually not immediately fatal and ordinary operation is sometimes possible, and if the load current is low enough, it seems the circuit may continue to just operate (albeit, dissipating power as heat).
Plotting the data as a function of output current, we can see that the tripped regions are not exactly zero in current. This is because, once tripped, the TCKE805 being an auto-retry type, the eFuse will rapidly cycle between tripped and soft-start until reaching the current limit, falling into another trip until the fault clears. This lets through short bursts up to the current limit, resulting in 0.1-0.4A in this test depending on voltage input.
Observing in the time-domain with the MXO4, we can see these cycles as frequently as 300µs (or ~3.3kHz). The trip is so rapid that no appreciable current overshoot is visible.
After each trip, the board appears to go through its soft-start cycle, which in the case of the provided capacitance, appears to have a rise time in the 30-40µs range.
Toggling between 1A and 2.8A at 50Hz, the board did not trip, which is the correct outcome.
Sweeping to 3.2A, the output does trip and cycle. Prior to this, the output voltage seems to decrease as the chip goes into “current limit” (i.e. pre-trip) state at about 2.9A (approximately, as oscilloscope values from the B&K 8600 I-Monitor output are not 100% accurate). This matches the datasheet’s description of the operating principle of the eFuse.
As previously mentioned, the process of tripping, current limiting and voltage clamping are all processes which are thermally-related. In the case of tripping, it occurs through current limiting first, creating heat which brings the package up to thermal shutdown. Similarly, clamping occurs in a dissipative fashion, similar to a linear voltage regulator, which also creates heat – enough of which can lead to thermal shutdown.
As a result, it should not be a surprise that, in a tripped state, the chip is hot. The maximum temperature recorded by the Topdon TS001 Thermal Camera was a toasty 127.5 degrees C. While this won’t immediately be harmful to the chip, prolonged operation at such temperatures will shorten the lifetime of the eFuse. This is why I believe the eFuse should not necessarily be relied upon as a form of primary (or sole) protection unless the design can assure that the eFuse isn’t stressed continuously to the point it could fail.
While it may often be considered that fuses don’t “consume” power, traditional fuses do waste power through their resistance. Conversely, eFuses can waste power through their Ron, but this is not usually any greater than traditional fuses.
However, eFuses also have power consumption required to power the internal comparator circuitry. As a result, I measured the power consumption of the TCKE805NA Evaluation Board using a Keithley 2450 SMU in both the enabled and disabled state.
At the intended 5V operational voltage, the consumption is about 910µA when enabled and 38µA when disabled. The datasheet values claim 460µA and 33µA respectively, which is a significant discrepancy for enabled state and similar values for the disabled state. The discrepancy is likely down to differences in test condition and implementation – a lower Rilim value than datasheet and higher Ven (tied to Vcc) than datasheet. Nevertheless, this consumption is something to keep in mind, especially for battery-operated and power-constrained equipment.
As mentioned in the last section, power is also lost as it traverses through the eFuse due to the “Ron” or more traditionally, the Rds-on of the embedded MOSFET. This is claimed to be in the order of 28mΩ typical, up to 38mΩ through the whole temperature range as stated in the datasheet. To measure this, a current was flowed through the eFuse from an Aim-TTi QPX750SP power supply into a B&K Precision Model 8600 DC electronic load. The voltage across the input-to-output connections was measured over the sense connections using a Keithley 2110 5.5-digit digital multimeter for maximum accuracy.
According to my results, the Ron was about 30-31mΩ, increasing to 32.5-33mΩ as it warmed up at higher currents. This is within the expected range, but a little higher than typical. This would not be unusual, as this can vary from device-to-device.
The EN input of the TCKE805 is quite a versatile input. As an input that is rated to full Vcc input, it can be simply used as an enable input. But it also has a low, but accurate, voltage threshold allowing it to be controlled from a microcontroller or to be used as a under-voltage lockout (UVLO) mechanism simply through the use of a voltage divider.
The threshold is stated as 1.0-1.2V with 1.1V typical when EN is rising, and 0.89-1.01V with 0.96V typical when EN is falling. To verify this, I decided to run a sweep using the signal generator of a Rohde & Schwarz MXO4 oscilloscope while observing the behaviour of the output voltage.
The EN rising threshold was measured as 1.1085V (in the presence of quite a bit of noise), which is within the datasheet range and exceptionally close to the typical value.
The EN falling threshold is a little more difficult to discern due to the lack of active load in the test (I should have turned on my load, in retrospect). The threshold would be at the point where the output voltage just begins to fall, which I would interpret to be around 0.958V, also in the presence of a bit of noise. This result is within the range and exceptionally close to the typical value. As a result, it would seem the TCKE805NA thresholds are very accurate (based on this sample-of-one test).
To be able to assess some of the capabilities, the board would have to be modified. For this section, I aimed to perform three modifications – the first was to increase the trip current to approximately 5A. The second was to slow-down the soft start by replacing its capacitor with a much larger 100nF capacitor. Finally, I wanted to test the reverse-current blocking capability with an external FET.
Unfortunately, while the board is designed for use with a Toshiba SSM6K513NU (30V, 15A, 8mΩ) as an external FET for this application, the FET is not provided nor mounted. Thankfully, element14 have this part in their catalogue (4415937), so I ordered five (to meet MOQ). They were about AU$1.13 including GST each, which is a reasonable price for the specification.
It should be noted that the board warranty (Article 4) apparently only covers leaded component mounting, thus this part of the review involving surface mount component mounting would void the warranty. While I can understand why the warranty is written this way, I feel that the board would have been better designed if such components were provided, but were switchable, to avoid the need for users to have to perform surface-mount soldering (which may pose significant difficulty to some) to evaluate the reverse current blocking capability.
The MOSFET is packaged in a UDFN6B package, measuring 2 x 2 x 0.75mm. The footprint is pre-tinned from the factory, which makes soldering a little easier. I chose to use flux and heat-gun from the bottom to avoid blowing the component away and to avoid damaging the header pins I already installed. An additional lead also needed to be installed for the output with the FET in series.
All three modifications were successful and the board appears as above.
Changing the capacitor had an obvious effect, now lengthening the start time to a rise time of 167ms for a gentle and linear ramp, whereas before, it was so quick that it was in the 30-40µs range.
The second desired result was also achieved, with the trip current now at about 4.9A. The behaviour of the circuit has changed slightly, with the trip only being released as the DC load current drops to about 1.1A.
Reverse polarity protection is a bit trickier to test. For this, I decided to also employ my Keysight E36103A power supply. For this test, I decided to run the same load ramp as above, but instead, have the QPX750SP supply one voltage (5V) and the Keysight E36103A provide a higher voltage on the output (e.g. 5.2V) backfeeding into the QPX750SP, but limited to a low current (~200mA). As a result, during the load sweep, we go from a situation of the E36103A backfeeding, to a case where the E36103A hits its current limit and the QPX750SP is forward-feeding through the eFuse and back again.
Running this test revealed a few strange behaviours. With no EFET in the path, we see a situation where the backfeed is clearly travelling from Vout to Vin, as expected as the integrated FET’s body diode would do this. But when the load reaches the point where the E36103A current limits and the eFuse goes into forward feeding, the chip appears to start up and go into soft-start (which is now nice and slow). As a result, backfeeding the TCKE805NA could result in unexpected soft-start behaviour it would seem.
But trying to run the experiment with the EFET in the path resulted in strange behaviour. The output mostly stayed at 0V for some reason, never mirroring the EN line in most cases, which I suspected was due to poor soldering. Removing JP7, sometimes it seems that leakage would charge the gate and the Vout would be equal to Vin. After taking three attempts and being very sure that my soldering was fine, I wondered if the remnant flux might be causing issues due to the limited FET gate drive current available (the datasheet suggests 2µA) which may be lost to leakage even before I attached any probes to the circuit. I tried using some isopropyl alcohol, to no avail. I tried re-mounting a final time with a fresh MOSFET, but then, when it seemed to give me output, it seemed the FET gate wasn't being driven hard positive and it was just conducting through the body diode instead. I wonder if my experiments had damaged the EFET output on the eFuse IC?
As a result, my experiments into the EFET reverse current protection was inconclusive. As previously mentioned, it would be nice if it could come pre-fitted but in a design which made it switchable to eliminate such frustrations.
The concept of an eFuse really subverts our ordinary thinking about the “fragility” of semiconductor parts especially when thinking about what happens when traditional fuses blow. By harnessing the speed and low-resistance of modern semiconductors, it is possible to create a solution that repeatably and accurately limits over-current scenarios through a predominantly thermal shutdown mechanism.
The Toshiba TCKE805NA is an eFuse that appears tailored for 5V applications, such as in computer IT equipment, with a fixed 6.04V (typical) clamping voltage. The evaluation board is pre-configured for an approximately 3A current limit with provision for a user-supplied external FET for reverse current protection.
On the whole, I found the solution performed as expected with regards to tripping, being both accurate and exceedingly-fast to the point of being immeasurable. The datasheet fast-trip specification of 150ns was difficult to observe, as repetitive trip-and-reset cycles seemed to accurately limit current to the set level without significant overshoot despite a high-current, low-impedance supply. The voltage clamp was effective, although the measured voltage was slightly higher than datasheet specification.
The downsides of eFuse solutions include the product requiring power for operation (about 910µA when enabled and 38µA when disabled), it only functioning when a voltage of 4.4V to 18V is available limiting applications, it not necessarily functioning correctly in the case of reverse current (untested) and ESD sensitivity as a semiconductor device (requiring protection components). Reliability can also be a problem under repetitive/continuous tripped/clamping applications as the chip would be operating at high temperature (~127.5°C) with an associated decrease in lifetime that may necessitate the provision of upstream primary protection (e.g. traditional fuse) which may negate some of the eFuse benefits.
However, I think the biggest allure of eFuses are not simply as a replacement for a fuse, but as a replacement for a load-switch (through use of the EN line), a UVLO protection circuit (also through clever use of the EN line), a soft-start/inrush-limiting circuit (exploiting the Cdv/dt input), a range of different current limits (through RIlimit programmability) which can reduce the BOM requirements and open more sophisticated protection schemes, and over-voltage protection (substituting for a TVSS crowbar). With the use of an external FET, it can also perform reverse-current blocking, which can be useful in systems with multiple power sources and substitute for an ideal-diode circuit or a blocking diode (although in this review, I was unable to demonstrate this despite attempting to fit the EFET). Best of all, it can do this in such a small space with a minimal number of extra components.
Thanks to element14 and Toshiba for the opportunity to review the TCKE805NA and learn about eFuses. I hope you found this review interesting and informative – please feel free to leave comments if you have any feedback or questions, like and/or share with anyone who may be interested.
Top Comments
Gough Lui Another great write-up. I like that you address things I wouldn't have even thought to ask about. I enjoyed the review.
Good review.
Nice review.
Thanks.
MK