Author: Gough Lui
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?: Discrete solutions based around comparator/op-amps or the use of transistor-based circuits.
What were the biggest problems encountered?: No output current specification for PTCGOOD output.
Toshiba Thermoflagger TCTH021AE RoadTest Review
By Gough Lui – November 2023
Temperature monitoring for overtemperature protection can be achieved in many ways, with thermistors being a very simple, reliable and cost-effective solution. However, the use of thermistors requires some analog circuitry, especially when interfacing to the digital domain. This circuitry can be somewhat finnicky to get right and is sensitive to component variations, thus requires redesign when voltages or thermistors are changed. It also takes up a bit of board area and can consume more current than absolutely necessary. Toshiba’s Thermoflagger is a way around this problem, comprising of a monolithic IC that tries to solve all of these pains.
Thanks to element14 and Toshiba for sponsoring this RoadTest opportunity. This is a particularly straightforward review and so was delivered express to ensure I had enough time to complete my other RoadTest and Design Challenge commitments. Despite this, the review did not compromise on quality and instead even delivered a RoadTest video which summarises most of the points and provides a visual demonstration of the board in action. The review document would have been very boring without it! As always, I’d appreciate any questions or comments you may have, and if you would leave a like or share this with someone who may find this review useful.
Below is the RoadTest video which provides a quick overview of the Thermoflagger and demonstrates it being tested in various ways. I highly recommend you watch this video as it will provide answers to most questions about the Thermoflagger.
Direct Link: https://youtu.be/i2BfYk8ESoc
The text version of the review, which is somewhat more detailed, follows. This is especially useful for those looking for more details or are not able to watch the video.
The Toshiba Thermoflagger is a small integrated circuit containing a fixed current source, voltage reference, comparator and output driver stage with logic. It was announced on 15th May 2023, making it a very new product.
It is intended for use in interfacing PTC thermistors to digital inputs to form an overtemperature monitoring solution. Such solutions can contain series-chains of PTC thermistors to monitor multiple locations with a single Thermoflagger, however, does not allow for distinguishing which location is causing the alarm. It should be noted that in this arrangement, the system also cannot distinguish between multiple sources of heat and with more thermistors in series, the system becomes more sensitive to heating of multiple thermistors due to the superposition of the “hockey-stick” characteristic curve.
Compared to circuits comprising discrete parts, the Thermoflagger requires less board area, fewer components, is tolerant to varying input voltage and changes in thermistor parts without redesign and may even consume less power. The below comparison was supplied by Toshiba in their catalog.
The Thermoflagger comes in eight different types which are combinations of PTC output currents of 1µA or 10µA, latching or non-latching flag signal and push-pull or open-drain output.
Oh, look – another Newark cardboard box from the UPS guy. It can only mean one thing …
Inside, packed in paper, is a small bubble-wrapped packet containing a bag with a PCB in it. There is literally nothing else in the box in the way of packaging or documentation. I would presume this is not a retail evaluation board at this stage (or if it is, it’s missing its outer box). As far as I can see, the board does not appear listed for sale anywhere, so it may be a special “sample” for testing.
The board design is rather cute, featuring a silkscreen which tries to explain the operation of the board schematically, although the traces themselves don’t necessarily take the silkscreened route. The ID-1 card form factor is very well understood and the board is designed such that it has traces on just one side. The edges are a little rough, seemingly V-scored rather than routed along the long edges.
The board is based around the Toshiba TCTH021AE Thermoflagger, a 10µA, non-latching, push-pull output type.
Two PTCs are used in series, one designed to alert at 80°C and the other at 105°C to illustrate the possibility of sensing with multiple PTCs on one Thermoflagger.
The PTCGOOD output flag status is used to drive two MOSFETs which each control one LED and series current-limiting resistor to show the status of the board.
It is much improved from the original shown in the pictures on the RoadTest page and in some of the documentation by incorporating a second Thermoflagger, broken out, for ease of testing.
There are no traces on the back at all.
Perhaps my main source of frustration is the fact that the Thermoflagger is an inexpensive part (US$0.476 individually, or US$0.175 in bulk at the time of publication) that comes in eight types. Why didn’t they just replicate the break-out and supply the other types for testing as well? The types with latching capability can be especially useful for slow or sleeping microcontrollers, while the 1µA PTCO versions could potentially appeal to those on a very strict power budget. I think such a change would make the board potentially more useful to an evaluator by allowing them to test the whole range with just a single evaluation board. In fact, the latching feature could also be demonstrated too with the addition of a momentary push button switch for the reset with appropriate resistor.
The board accepts USB-C power input, which is easy to find nowadays. No cable is supplied with the board, as a result. Use from a power bank is not advised due to the low current draw, which may cause the power bank to go to sleep. The board itself has an obvious function – namely, it lights one LED in case of temperature normal …
… and the other in case of over-temperature.
The video demonstration above shows this functioning as expected. However, there are some other tests that can be performed with the separate broken-out Thermoflagger and by modifying the board.
Available documentation includes their:
Overall, I found the documentation quite readable and generally sufficient for design purposes. There was one metric I found missing in the datasheet – the maximum PTCGOOD flag-signal current in source and sink, which only seems tested at 4mA by the specification table.
In order to run further tests and ensure no interference, I used the hot air pencil to remove the first PTC and swing it around such that it is an open circuit. With a soldering iron, I was able to mount a force-sensitive resistor that I had on-hand.
Unfortunately, this FSR is a bit long, so it had to face the opposite direction to that intended by the board designer.
Because of the resistance of the FSR being nearly open-circuit when un-pressed, the output state is inverted.
However, the binary output is still easily shown to be functioning. A similar change can be made to the board for a light-dependent resistor (CdS) but I could not demonstrate this as I didn’t have one on hand. The process is straightforward, the principle is quite similar and is covered by the documentation.
To make the board more friendly for the performance testing, I also fitted 2.54mm header pins to the broken-out Thermoflagger for connection to test equipment.
But there was an additional hack I decided to perform. The GND1 and GND2 pins should be connected together to the system ground, according to the datasheet. Testing with a DMM seemed to show that both pins weren’t internally shorted together inside the package, so I decided to use a solder-blob on the back to bridge the pins together to ensure the grounds are kept at the same voltage. This also has the advantage of making testing easier, by doubling-up the ground connection.
Tests were performed using the Keithley 2450 SourceMeter SMU, Rohde & Schwarz NGM202 Two-Quadrant Power Supply and Rohde & Schwarz MXO4 Oscilloscope (review still in progress). Tests were performed at 25°C room temperature only.
Note that current consumption tests were made with the PTCGOOD flag signal open-circuit as any current consumed by the downstream input taking a signal from PTCGOOD will depend on the design of your circuit and is a variable that does not represent the self-consumption of the Thermoflagger and its connected PTCs.
With the PTCO open-circuit, the self-consumption of the chip ranged from 1.43µA to 1.55µA depending on Vcc. With PTCO shorted to ground, the total consumption ranged from 11.42µA to 11.78µA which matches well with the typical figure of 11.3µA and maximum figure of 14.7µA listed in the datasheet.
The PTCO current will depend somewhat on the voltage at the PTCO terminal. When getting close to Vcc, the IC is unable to push current out of the PTCO terminal. It takes about 0.6V difference for the PTCO to reach near the rated current, from which the output is then regulated.
Zooming into the current axis around the set-point, we can see at minimum Vcc of 1.7V, the current is very much “spot-on” at just shy of 10µA. With the maximum Vcc of 5.5V, the current is a bit higher, but not much so, at around 10.13µA for the most part, ignoring noise, with a slight increase near zero volts to about 10.2µA. This is quite good regulation performance and meets the datasheet claimed range of 8.0 to 12.2µA across the Vcc = 1.7V to 5.5V range.
The amount of current that can be sourced or sunk by the PTCGOOD output is not indicated in the datasheet. I did some I-V curves to see what the characteristics of the output are.
Because of the higher resistance of P-MOS, I would expect sourcing current to be more difficult than sinking current. At a glance, it seems that with a Vcc of 1.8V, the chip can be expected to source up to 5mA before hitting the knee point of the curve, without considering whether this is sufficient voltage for digital logic. At a Vcc of 3.3V, the knee point is closer to 23mA and the voltage at 4mA is about 3.1V, which is a little higher than the datasheet specified 3.03V minimum, which is good. At a Vcc of 5V, the knee point is not seen up to the test limit of 30mA (as I didn’t want to overstress the chip). It appears that the source impedance at a Vcc of 5V is about 40Ω which suggests that (in theory) 61mA might be available for a power dissipation of 150mW (chip limit) or 89mA at 320mW (chip limit when mounted on board) assuming the circuitry doesn’t itself dissipate any power. I didn’t test it to this limit, however.
Sinking is usually easier, and this holds true here as well. At a Vcc of 1.8V, it seems the knee point of the curve is about 7mA. At a Vcc of 3.3V, the knee point seems to be above 30mA as is the case at 5V. The output impedance is about 25Ω at a Vcc of 5V, suggesting much more current can be sunk into the chip – theoretically 77mA for 150mW or 113mA for 320mW. Again, I did not test the chip to its limits and the “linear” resistance relationship will break down at higher currents. Best not to stress the chip out – but good to see that currents greater than the tested 4mA in the datasheet is a possibility.
Testing of the minimum pulse length required to activate the PTCGOOD output showed that a 18µs pulse was insufficient, being filtered out.
A 20µs pulse, however, was sufficient. This matches the “operation MASK time” specification in the datasheet, suggesting there is the capability to perform deglitching. Under this state, the output was low for 33µs, which is a little shorter than expected.
Using a longer pulse of 100µs, the input to output propagation time is measured at 20.5µs. This is longer than the 17µs typical for tDET1 in the datasheet, but not significantly so, and the test conditions are not identical.
The release time was averaging around 226.5µs with some jitter visible on the persistence display. This is not far from the 214µs typical listed on the datasheet for tDET2.
Testing of the threshold was based on a 5V peak-to-peak triangle waveform, offset by 2.5V to sweep PTCO through 0-5V at a rate of 10Hz.
On the PTCGOOD upward transition, I saw a threshold of 320mV.
On the PTCGOOD downward transition, I saw a threshold of 510mV (very close to the nominal 500mV). Looking at the datasheet, the detect voltage should range between 420mV to 580mV, therefore, this is a pass. The hysteresis voltage is listed as typical 100mV – in this case, it is implied as 190mV. However, the hysteresis could be waveform and frequency dependent.
Testing hysteresis with a sine wave instead at 1Hz, I saw that 110mV was enough to toggle states reliably but 100mV was not, regardless of the offset, therefore implying that the 100mV hysteresis claimed by the datasheet is likely to be accurate under certain conditions.
The hysteresis is effective in suppressing false triggers with this triangle waveform bouncing between 0.5V and 5V.
Because the chip requires so little power, I decided to power it from the MXO4’s signal generator directly, with the PTCO shorted to ground, so as to see where in the ramp waveform it powers up and shuts down.
It appears UVLO functions at around 1.333V in both directions. This is noticeably less than the 1.5V typical as claimed in the datasheet, but it is good to see that the chip is able to tolerate power cycling without glitching.
The Toshiba Thermoflagger seems to do what it says “on the tin” and does it rather well. As an integrated circuit integrating a current source, reference voltage, comparator and output-stage with logic, it does ease the implementation of PTC-based overtemperature detection for digital inputs. It simplifies the need for circuit redesign due to changes in thermistor or input voltage and can handle a range of input voltages just fine. The chip is physically small, saving on board area; while current consumption is very tame with a consumption ranging from 11.42µA to 11.78µA, which helps save on power. The output current is very well regulated (from 9.99µA to 10.2µA observed), the detection threshold/hysteresis seem accurate and the response is fast (20µs). Best of all, the chip is also relatively inexpensive, making it a potential winner for those looking to implement an inexpensive and simple digital logic overtemperature signal.