TM-902C: cheapest dedicated k-type thermocouple thermometer review

[electronupdate]

So I was getting tired of having to move my multimeter around to measure temperature at various point of a prototype I am working on.  The solution, of course, would be a thermal camera.... but they are still too expensive.

The next best thing would be some dedicated meters which would accept standard k-type thermocouples.  A quick web search shows up the "TM-902C" as the go-to meter at the bottom end.  Wonderfully uncomplicated (there is only a power switch), it works as advertised.

It would have been nice, however, if there was a way to calibrate the meters.  One of them is different than the other 3 (still within spec from what I can tell).. the PCB hints at test points for calibration and switching between fahrenheit and celsius modes  however, there is no documentation.

[mariush]

Like a video comment says, those thermocouples can only do 250-450c ... the isolation on the cable and the blob in the tip can't handle much higher temperatures.

For 1100c you need probes that cost 15-30$ a piece, or even more. For example see this one, with stainless steel, insulated sheath etc etc : http://uk.farnell.com/labfacility/kmi-1-5x150-iec/probe-thermocouple/dp/7078146 

[Vgkid]

Thanks for the review, I'm comtemplating getting one of these.

[Gyro]

Hi,

Sorry to resurrect an old topic, but given that these things are still available so cheaply on ebay I though it was worth it.

Only my second post, I've been lurking as a guest for a while but thought I would wait until I had something to say.

Anyway I purchased one of these a while back from a UK seller. On inserting the battery I was greeted by a burning smell and rapidly heating battery. I contacted the seller and apparently I was one of several. After receiving a refund threw it into a dark corner. A few weeks later I decided to investigate. It was immediately clear that the SOT223 3V regulator on the board was blown and the source of the smell. No idea why, it should have been well within voltage rating and I hadn't reversed the battery. Probably a case of 'contents not matching the label'. Curiously the power switch was wired AFTER the regulator too, meaning low continuous quiescent drain on the battery.

I removed the regulator and bridged the footprint and powered the rest of the circuit from a bench supply and found that it was quite happy. What's more it was quite happy down to about 1.6V, ie. end voltage of a pair of 1.5V cells, when the low battery indicator came on! There was no change in display over the normal expected voltage range.

I have no idea of the logic of powering a chip designed for 3V operation from a 9V battery via a non-zero cost regulator - I wondered if it was a case size constraint, but no - a wire ended double AAA holder fits perfectly in the battery compartment, with sufficient space for a bit of foam too.  Amazon reviews say battery life is long, but AAA cells are cheaper and much higher capacity. Wasting the best part of 6V from a low capacity 9V battery is just plain silly.

Hopefully this will be helpful to others looking for longest possible runtime. BTW a small section of plastic cut from a blister pack sandwiches nicely inside the front of the case to protect the otherwise bare LCD.

[Johnny B Good]

Hi,


Only my second post, I've been lurking as a guest for a while but thought I would wait until I had something to say.

Anyway I purchased one of these a while back from a UK seller. On inserting the battery I was greeted by a burning smell and rapidly heating battery. I contacted the seller and apparently I was one of several. After receiving a refund threw it into a dark corner. A few weeks later I decided to investigate. It was immediately clear that the SOT223 3V regulator on the board was blown and the source of the smell. No idea why, it should have been well within voltage rating and I hadn't reversed the battery. Probably a case of 'contents not matching the label'. Curiously the power switch was wired AFTER the regulator too, meaning low continuous quiescent drain on the battery.

I removed the regulator and bridged the footprint and powered the rest of the circuit from a bench supply and found that it was quite happy. What's more it was quite happy down to about 1.6V, ie. end voltage of a pair of 1.5V cells, when the low battery indicator came on! There was no change in display over the normal expected voltage range.

I have no idea of the logic of powering a chip designed for 3V operation from a 9V battery via a non-zero cost regulator - I wondered if it was a case size constraint, but no - a wire ended double AAA holder fits perfectly in the battery compartment, with sufficient space for a bit of foam too.  Amazon reviews say battery life is long, but AAA cells are cheaper and much higher capacity. Wasting the best part of 6V from a low capacity 9V battery is just plain silly.

Hopefully this will be helpful to others looking for longest possible runtime. BTW a small section of plastic cut from a blister pack sandwiches nicely inside the front of the case to protect the otherwise bare LCD.

Hi Chris, I realise I'm responding to an almost 7 year old post but I thought you might like an update on the latest versions being sold in 2021/2022.

 I was looking at these K-type thermometers the end of January and noticed they had replaced the on/off slide switch with a momentary press on/off button and were now using a 2xAAA cell battery in place of the more expensive 9v block battery (PP3) so ordered one from the cheapest ebay seller (a Chinaman, of course!).

 This was as pictured (red push button and 2xAAA cell battery holder) but was DoA (PM me for details ). I got a refund and tried again. The second working unit was also as pictured and AAA cell powered. Out of curiosity, I measured the consumption and was amazed to see that it only drew 0.35mA (after an initial startup surge of just over a mA for a second or three).

 I then decided to buy another one from another seller offering the "or best offer" option and got it a little cheaper. When it arrived, I was a little dismayed to see that it used the slide switch rather than the red on/off button of the current models. However, it also used a 2xAAA battery pack and likewise drew a mere 0.35mA.

 The first working one with the push button had the seemingly unselectable auto-shut off time out annoyance (10 or 15 minutes) which the second slide switch version hadn't been cursed with. After some calibration testing with my best cheap thermocouple (mini-K plug upgraded to a chromel/alumel pinned plug), this proved to be the most accurate of the two meters (within 0.1K of melting ice and showing 99.9 deg C in boiling tapwater at an absolute barometric pressure of 1017mBar suggesting an error of about -0.3 deg C at this calibration point). Almost certainly just a matter of "pure dumb luck" rather than 'by design'.

 By using a shorting link to see what these meters thought their reference temperatures were, the second (good) one matched to within .2 deg C of what the two room thermometers were showing under stable early morning room temperature conditions (the only time they'll agree to within 0.3K of each other, largely due to vastly different thermal time constants). The first working one under-reads by about 2K (shorted input or compared using the same thermocouple) - I haven't bothered to test that one with melting ice and boiling water.

 Today, I decided to open the DoA meter to see whether there might be some way to recalibrate these meters and with a 5.1M resistor and a 100K with a 1K on the board alongside of the mystery CoB, I suspect there well may be a means to adjust the high gain settings involved to adjust for the 40.7uV/K required for temperatures at the high end and just possibly also a means to adjust the reference offset to get an accurate melting ice value within 0.1K.

 However, I've only opened up the DoA unit so far. I'll need to open up the working push button (sacrificial) unit to test whether they can be recalibrated with subtle changes of resistor values. Anyone else thinking of doing the same experiment, please be aware that the tiny screws holding the board against the display panel's zebra stripe have been tightened to breaking point and care should be taken when reassembling it to clean any plastic off the threads before carefully trying to avoid cutting a fresh thread into what remains of the plastic.

 IOW, turn those self tapping screws anti-clockwise with minimal pressure to detect when it (hopefully) drops into engagement with the existing thread. Don't be like one of those clumsy klutzes that do such teardown videos which make me wince every time I see them blundering straight into tightening self tapping screws back into already damaged plastic without any attempt to engage what remains of the original thread    

 Just prior to checking out the DoA unit, I'd been taking a closer look at the display during its short switch on initialisation sequence to see if I could spot any interesting but inaccessible options such as Celsius/Fahrenheit temperature scales (it defaults, sensibly imo, to Celsius) and some means of disabling the annoying auto-shut off feature (at 0.35mA and a couple of cheap alkaline AAA cells, totally superfluous).

 After experimenting with button press sequences, I eventually discovered how to disable that annoying auto-shut off feature. This proved to be a remarkably simple process, just a matter of the usual momentary press and wait for it to complete its initialisation and declutter the screen before pressing and holding the button down until the timer icon, top left vanishes after about 5 or 6 seconds. Useful to know for anyone who has bought any of these late version models.

 Getting back to the possibility of displaying temperature in Fahrenheit units (why would anyone want to inflict this on themselves?), from my examination of the DoA's circuit board, there seems to be a solder blob jumper option for this choice of temperature scale (currently un-blobbed).

 There's also another jumper close by marked "SW", presumably to handle the difference between a momentary button press and a permanent on state from a slide switch which might also offer a means to disable the auto-shut off feature by default.

 Interestingly, there seems to be another set of connector holes which look like a PT100 or PT1000 or even a 3 wire PT100 sensor option (the power on display test hints at such a possibility). One other remaining 'interesting' feature is the board version apparently being at Rev1. even though it is totally different to the early and later improved slide switch versions. It seems so far to date, that there are at least 3 different versions that have been or are being sold (quite possibly even more).

 I've attached a couple of photos each of the two variants that I acquired (the DoA unit is exactly the same Rev 1. as the working Rev1. unit btw) for comparison against other internet published images of this meter's PCB.

[Gyro]

Hi John,

It's interesting to see that they've even changed the circuit of the slide switch variant from the ones I have. It looks as if they have taken the temperature sensor on-chip on both slide and push button switch versions now. From your description, it sounds as if, when you get a good one, their accuracy is still very respectable.

This thread has quite a lot of discussion of the various types, including, in Reply #11, what they used to look like in the 1980s. It really shows the extent of the 'value engineering'!

https://www.eevblog.com/forum/testgear/tm902c-quality-and-accuracy/

There was some discussion of whether the connector holes that you refer to might be some form of comms port of calibration purposes or maybe even serial, but as far as I'm aware, nobody ever successfully followed it up.

It looks as if the revision I have might have been the last one with a discrete diode temperature sensor on the PCB and footprinted analogue calibration components. I bought a second one not long after the first and modified the fixed resistive divider by adding a preset pot and resistor at the appropriate locations. With careful adjustments (over a long time - it takes a long time for the thermals to settle) I was able to get them to consistently match to 0.1'C over a respectable temperature range. I ended up adding a lot of copper foil and as little Kapton tape insulation as possible to keep the connector and sensor diode (diode connected transistor) as isothermal as possible. Re-reading the other thread, it looks as if I intended to try physically moving the diode to the connector itself, but I never seem to have got round to it!

It used to be that the way to differentiate the on-chip sensor and the external diode was whether it had a slide switch or push button. From your images, that is clearly no longer the case (scanning through the web images, there may be a couple of other slide switch variants too). The photo below (not mine) shows the board revision that I have in my units (in its original 9V state)...

[Johnny B Good]

Hi Chris,

 thanks for responding so quickly to my "necro-posting" I'd been searching for more information on these these K type temperature meters after opening up the DoA one to discover that there were at least three different versions if not more. This and the other eevblog topic thread, <https://www.eevblog.com/forum/testgear/tm902c-quality-and-accuracy/>, were the only threads I could find (DDG or Google do a much better job at tracking down eevblog topics than eevblog's own search engine ).

 I only skimmed through the other thread which seemed to have drifted a little OT by the end so this thread seemed the more appropriate to post my own observations into and maybe elicit some more up to date information. Having only just had a proper read of that thread, I see that the only new information I seem to have added to the pool of knowledge was in regard of the chimera version (3v powered slide switch version, properly secured to the pcb I was happy to discover) and my thoughts about those mystery P2.4, P2.5 and GND tphs on the first one (marked P21, P22 and P23 on my chimera version) possibly being for a PT sensor socket (hinted at by the letters PT appearing under the minus sign during the screen initialisation/test startup sequence).

 What staggered me was the immediate interest my attached images had aroused (5 views between my posting and reading it back to make sure the attachments hadn't gotten mislaid by the eevblog server with one count being my own viewing, all in the space of less than 5 minutes!). Initially there was me, one other named member and 8 guests watching the thread. I was quite surprised there could still be so much interest in these K-type meters after all this time!

 I guess close to a dozen members must have clicked the email notification button over the past 7 years because the count had reached 12 views when I last checked about an hour ago. I see the first image has now been viewed by 22 other members and guests as of 00:38 GMT this Monday morning. There does seem to have been a fair amount of interest in this topic (there is no one else currently viewing this topic right now).

 With only about 1mW dissipation by the COB, I should think an integrated temperature sensor would suffer very little isothermal temperature offset error. The whole thing will be pretty well at ambient. The internal sensor could easily be within +50mK of the PCB temperature (I'm guessing ). In this case, no real need for placing a thermistor right onto a chromel/alumel K-type socket which can be a more suitable base metal (brass spring contacts?) anyway.

 Using a dedicated K type, very low power consumption thermometer COB in this case, is a win win. The only fly in this ointment being the effect of rapid ambient changes and the very long thermal time constant between ambient and the COB's reference temperature sensor. It's all a far cry from its original 1980s incarnation!

[Cerebus]

Anyway I purchased one of these a while back from a UK seller. On inserting the battery I was greeted by a burning smell and rapidly heating battery. I contacted the seller and apparently I was one of several. After receiving a refund threw it into a dark corner. A few weeks later I decided to investigate. It was immediately clear that the SOT223 3V regulator on the board was blown and the source of the smell. No idea why, it should have been well within voltage rating and I hadn't reversed the battery. Probably a case of 'contents not matching the label'. Curiously the power switch was wired AFTER the regulator too, meaning low continuous quiescent drain on the battery.

I had one of these die on me a couple of years back after a couple of years good service. In my case, like yours, the regulator had died. I didn't investigate too carefully after that and got some replacement Holtek 75xx-1 LDO regulators of the appropriate voltage. Turns out I was unlucky and after replacing the regulator it became clear that something downstream had died too leaving me with a pile of surplus cheap regulators which have turned out to be useful. Those regulators have an incredibly low quiescent draw of 2.5uA typical, 5uA max and live up to their specs. So the rather odd placement of the power switch after the regulator isn't quite as weird as it first seems as the LDOs quiescent current draw is on the same order as a 9V PP3s self discharge.

I've bought a couple more TM902-Cs since and those have been fault free (so far). Useful things to have around, they'll run continuously for months on a single PP3 and all the ones I've had have been accurate to better than 1.0ºC at 0ºC and 100ºC

[Gyro]

It still makes me chuckle that they would have put down a die on the board that was fully intended got 3V battery operation and then, without realizing, gone to the expense of adding a low quiescent regulator and 9V battery). The photo reminded me that they also put primitive reverse diode protection on the battery terminals when they had more than enough headroom for a (or half a dozen, come to that!) series diode. The original regulators were obviously 'knock-offs' which were very marginal at 9V (unlike the genuine Holteks). I never got around to testing the battery life with AAA cells, but I suspect that they would leak before they went flat (makes mental note to go and check them!).

[Johnny B Good]

 My cheap 9999 counts Mestek DM91A meters take just 2mA from their 2xAAA battery packs as does a BSIDE ACM91 DC clamp meter (bought purely for its 1mA DC clamp meter resolution) for everything other than DC clamp meter mode when it draws a whopping 20mA!

 I reckon I was seeing around 400 hours worth out of a pair of cheap alkaline cells with the Mesteks (I'd been using them to monitor EFC and C field voltages and was in the habit of cancelling the auto shut-off feature for each day's use).

 The notice of impending auto shut-off beeps remained active which is both a blessing and a curse - luckily, rather quiet compared to the more strident of the BSIDE's similar warning beeps when its auto shut-off feature is likewise disabled (but since it drew ten times more current from the battery when measuring DC current, it seemed a reasonable precaution to make sure the warning wasn't missed or so easily ignored ).

 By that 2mA / 400 hours (per cell terminal voltage circa 1.3v) standard, I've estimated a conservative run time some six times longer off the same cheap brand of AAA cells, some 2500 hours worth! I've just checked Duracell's "Plus" AAA datasheet which shows 750 hours to 1.3v at 1mA constant current and get a similar, if slightly lower 2250 hour estimate (those cheap JCB branded cells from Home Bargains aren't quite so inferior to Duracell's overpriced offerings after all! ).

 You have a point regarding the risk of cell leakage occurring before the cells become depleted but with branded, even if only the cheaper JCB cells, I think that risk is vanishingly small unless you leave it to run flat hidden away in a drawer for a couple of years.

 Now otoh, those cheap Chinese squidgy carbon zinc 'heavy duty', aptly named "Bursto", AAA cells that are often supplied with these meters, I'd say it's odds on that they'll leak before going flat.

 Other than using them to briefly test the meters they'd been supplied with, I put them straight into my "Spent Batteries" box. However, since I wanted to upgrade the twisted wire joint hot junction on the end of a 5 metre coil of high temperature thermocouple extension cable to a bead weld joint, I salvaged the carbon rods from that last pair of cells and put them to a good use in making up a heavy gauge 5 metre long K type thermocouple - those AAA cells hadn't gone completely to waste in my case (plus, I'd discovered yet another novel use for my 10A 30V bench supply ).

[Gyro]

... I was happy to discover) and my thoughts about those mystery P2.4, P2.5 and GND tphs on the first one (marked P21, P22 and P23 on my chimera version) possibly being for a PT sensor socket (hinted at by the letters PT appearing under the minus sign during the screen initialisation/test startup sequence).

Interesting. Maybe those later chip versions do have a PT100 capability, I've never seen a PT annunciator on my version. I wonder how it would be selected though. The only solder blob jumpers I can see are the 'C/'F selection and the SW, for switch - unless that's a switch between K-type and PT100. The problem with glob tops is having absolutely no part number to reference.

... I was quite surprised there could still be so much interest in these K-type meters after all this time!

Yes, they're still popular, mainly, I think, because of the ridiculously low cost. they (or at least the case and part number) seem to have had something like a 40 year lifespan! There's no point in faffing around with the thermocouple function on a DMM (or basing a buying decision on its presence), when you probably want to be doing something else with the DMM while you're monitoring a temperature anyway. Having two is useful for a whole range of stuff (including balancing radiators!). I was lucky enough to pick up some thin wire bare RS K type thermocouples cheap on ebay, they are more flexible and much lower thermal mass and conduction than the thicker ones supplied.

With only about 1mW dissipation by the COB, I should think an integrated temperature sensor would suffer very little isothermal temperature offset error. The whole thing will be pretty well at ambient. The internal sensor could easily be within +50mK of the PCB temperature (I'm guessing ). In this case, no real need for placing a thermistor right onto a chromel/alumel K-type socket which can be a more suitable base metal (brass spring contacts?) anyway.

 Using a dedicated K type, very low power consumption thermometer COB in this case, is a win win. The only fly in this ointment being the effect of rapid ambient changes and the very long thermal time constant between ambient and the COB's reference temperature sensor. It's all a far cry from its original 1980s incarnation!

Agreed, there should be very little effect on the on-chip temperature sensor from its dissipation. One thing I did find though, is that the thermocouple effects in the connector do make a big difference to the reading. The stock units probably don't have proper matching contact materials so close thermal coupling between the connector sensor is very helpful. I found that, when matching my units, it was best to make a tiny screwdriver hole in the front of the two units so that I could adjust the trimmers without touching anything else. With the on-chip temperature sensor version, you might find a worthwhile improvement in tracking by covering the whole blank rear of the PCB with self adhesive copper foil and extending this up and around the body of the connector. Irrc, I did this on mine (at least as far as the diode footprint, in addition to the top side foil) but it was a long time ago.

[mawyatt]

 By that 2mA / 400 hours (terminal voltage circa 1.3v) standard, I've estimated a conservative run time some six times longer off the same cheap brand of AAA cells, some 2500 hours worth! I've just checked Duracell's "Plus" AAA datasheet which shows 750 hours to 1.3v at 1mA constant current and get a similar, if slightly lower 2250 hour estimate (those cheap JCB branded cells from Home Bargains aren't quite so inferior to Duracell's overpriced offerings after all! ).

 You have a point regarding the risk of cell leakage occurring before the cells become depleted but with branded, even if only the cheaper JCB cells, I think that risk is vanishingly small unless you leave it to run flat hidden away in a drawer for a couple of years.

 Now otoh, those cheap Chinese squidgy carbon zinc 'heavy duty', aptly named "Bursto", AAA cells that are often supplied with these meters, I'd say it's odds on that they'll leak before going flat.

Duracell batteries we've used were almost a guarantee of leaking, and they wouldn't stand up to their warranty claim covering battery leak damage

Found the batteries that IKEA offers are good, and don't tend to leak. These are made in Belgium, not China.

Best,

[tunk]

Ikea don't sell alkalines any more.
Some of the Ikea batteries were made in China.
And Duracell has a factory in Belgium, so they may have been made by them.

[mawyatt]

Just got some a few months ago, and these say Made in Belgium right on battery 

Best,

[tunk]

"IKEA to remove non-rechargeable alkaline batteries by 2021":
https://www.ikea.com/us/en/newsroom/corporate-news/ikea-to-remove-non-rechargeable-alkaline-batteries-by-2021-pubb6aee8b0

[jpanhalt]

I just received my TM-902C with single red button and tested it.  Less than 1°C different than my Fluke.

I got the 902C to stick to the outside of my Yihua 853A preheater to give a reading of the actual PCB/aluminum plate temperature.  There are a couple of 5V and GND pins on an unused 6-pin connector attached to the 853A's PCB that I could use for power.  So, I opened the 902C to see whether it has a voltage regulator.  Talk about minimalistic bare bones.  The laminated paper PCB connects directly to the LCD by a zebra strip.  I have had bad luck getting those things back together.  So, I didn't go any further.

Does anyone know whether there is a voltage regulator on the board?

John

EDIT: I just reviewed the pictures posted previously .  There does not appear to be a discrete VR.  Built into the COB?

 

[magic]

I have seen three versions of those meters.

One is the oldschool 9V battery and slide switch.
One is 9V battery and red button. These two work pretty well, although calibration isn't equal on them so at least one must be off some 1°C.

The third one runs on two AAA cells, has a red button, 3-bar battery status display, and it sucks.
Want to know why? It has fairly large input bias currents, which produces meaningful drop on thinner or longer thermocouples. At 100Ω it is already a few °C off.
IIRC, this unit also had some weird issues with very low negative temperatures, but I don't remember what was wrong.

[mawyatt]

"IKEA to remove non-rechargeable alkaline batteries by 2021":
https://www.ikea.com/us/en/newsroom/corporate-news/ikea-to-remove-non-rechargeable-alkaline-batteries-by-2021-pubb6aee8b0

We were at IKEA last Dec, so looks like these are the last cells we'll get from them. Maybe they'll have some quality rechargeables cheap, we have a bunch of NiH enloops but don't use them much.

Best,

[jpanhalt]

@magic

Thanks for the response.

Mine is the 3rd version.  A regular Omega mini TC plug fits and that is what I will be using with a heavier gauge, Teflon TC that I make from bulk wire.  That should help with the bias current.

The 9V version must have had a VR somewhere, but maybe that also accounted for the high quiescent current. I'll just use a 3.3V regulator in the pre-heater case to be safe.

[Johnny B Good]

@magic,

 The two examples that I described (both 3 volt versions, one with the properly secured slideswitch, the other with the red momentary press button) show no strange behaviour checking our 20 year old (still in production!) Elcold chest freezer temperature at -21 deg C so it looks like you've been lumbered with a substandard unit.

 Mind you, I have to agree about the high bias currents since shunting the input pins with even a 10K resistor resulted in a higher than the expected 'ambient' reading and the 920K I'd initially tried gave the minus only indicator of a disconnected sensor!

 Shunting the HI and LO sockets used by bench meters to read the microvolts output from a K type thermocouple when set to their temperature scale with a 1M resistor is normally enough to get the meter to display its internal isothermal reference temperature as per the effect of plugging in a shorting link. I was quite taken aback by this behaviour with these TM-902C meters.

 However, I've just checked the resistance of the cheap 1 metre K type probe these are typically supplied with, which proved to be a whopping 18.5 ohms, regardless of the test meter polarity (I'd been expecting to read a value somewhere around the 2 to 5 ohm mark). It only needs a 1uA current with one of these cheap thermocouples to add almost half a degree offset.

 As it happens, as far as I can make out, both seem to be generating a more modest 130nA bias current which equates to a 2.4uV offset with an 18.5 ohm K type TC which in turn equates to an offset of ~0.06K. Even with a 3 metre long version of these probes, the error will still be within 0.2K which, considering their specified accuracy, will be the least of your worries.

 For an added bonus datum point, that 5 metre length of high temperature 0.4mm gauge chromel/alumel wire thermocouple extension cable I'd converted into a thermocouple sensor reads 25 ohm (about a 0.08K offset error with a TM-902C).

[magic]

I'm not surprised. Chinese roulette

My unit had something like 1µA if not more. And I found that most of it came from an internal 1MΩ pullup resistor which is meant to drive the input over range when no probe is connected. Removal of the resistor decreases the current somewhat, causes the meter to display ~120°C with no probe and reduces readings even with the included 15Ω probe.

The thing about thermocouples is that the finer they are, the faster they react. One of mine is 70Ω and only 1m or maybe 1.5m long.