Adventure with a Cheap Amazon Temperature Controller

The post below actually started as an Amazon review for this product. A few months back in late-2018, I was contacted by the seller to get free product in exchange for reviews. Most of their stuff looked kind of like run-of-the-mill cheap junk, but one thing was a halfway decent looking temperature controller board that piqued my interest for a couple of projects I had in mind at the time. Unfortunately it took me a good many months to get around to any of them.

Product image of the temperature controller from amazon
This is the controller I received to test.

The Project

This is its own whole blog post (underway), but I recently bought a hot tub and built a semi-DIY aftermarket heater for it. A marine 12V mains-pressure water pump pushes water through a propane-fired camp shower to heat the tub much more rapidly than its anemic built-in electric heater. The propane burner ignites and extinguishes itself automatically with water flow, but if the pump is left on, there’s no temperature control and the water will heat up indefinitely since the water is recirculated. The idea was to see how well this temperature controller works by using it to close that loop.

Review unit "installed" in the hot tub system for testing.
Installed in-line with the 12V heater pump on my semi-DIY hot tub.

First Impressions

The user interface is actually pretty excellent, and generally self-explanatory. The features are also great. It has a segment LCD which is super crisp and legible from all angles, thanks to inverted drive and a good bright LED backlight.

  • Setting the control temperature, and whether it’s intended to control heating or cooling, is straightforward. The button interface is generally pretty obvious.
  • 0.1 °C increment is definitely adequate.
  • There are some advanced features I didn’t realize at first.
  • The temperature probe could be a bit longer, and since it’s connectorized, making it longer is just slightly more difficult than I’d probably like to do neatly.

Features

This thing is actually kind of packed.

  • It’s only normally-open, but it’ll do heating or cooling control, which makes that not a problem.
  • It allows for calibration of the NTC thermistor. It’s a little confusing how, since 3 digits of a 7-segment display isn’t a lot to get descriptive. But the manual does make it easy enough.
  • You can set the hysteresis of bang-bang control in .1° increments. I think 2 °C is default, but I set it to 1° since the large body of water doesn’t change temperature quickly.
  • There’s actually a UART interface for controlling the device – all settings seem to be available, as well as enable/disable, manual overrides, temperature read-back, and so on. I can’t imagine why you’d use a controller like this if you were already going to the trouble of writing a microcontroller firmware, but I suppose for the hobbyist market they’re targeting, it gets you a thermistor, relay, and control loop without any real hardware design. And you could easily use it to control the process from something like a computer or a raspberry pi which might be useful.

Misgivings and Reservations

The thing I like least about this controller is that it pretends to be usable with wall voltage (220V, even!) despite shipping without so much as a case, much less board-level consideration for mains.

Amazon product image showing 220V connection.
Do NOT do this. Everything about this is bad.

It’s a bit of a tired internet trope to warn against handling household mains voltage in DIY projects, but a few very specific design deficiencies come to mind in this particular case that are worth exploring for educational reasons:

  • It doesn’t have a case. This is sort of self-evident, but yeah. Wire mains to this thing, and every brush by and button press becomes Russian roulette.
  • The low voltage and high voltage terminals are too near each other. This might also be pretty evident. There are more specific reasons below, but what most jumps out at me is that it’d be REALLY easy to hook up some heavy stranded wire like lamp cord, only to accidentally short live to neutral or live to system ground with a stray strand that didn’t make it in the relatively small screw terminals. This is enough of a problem at low-voltage DC that mains should terrify you here.
  • Inadequate isolation. The board has nothing like high voltage isolation slots or even keepouts to separate the (potentially) HV switched side from the LV poke-your-fingers-at-it side. Two parameters are important for high voltage isolation: Clearance and Creepage. Clearance (the through-air, as-the-bird-flies distance between traces/pins of differing voltage) is pretty low on this board, with relay pins and terminals right next to other pins. Creepage (the distance an ant would have to crawl over solid material) is exactly the same, since there are no slots. Of course, this is a particular problem for a board that has no case and which looks like it wasn’t cleaned post-assembly. Creepage is a problem because contamination can build up on a board and provide a path from pin to pin.

This board simply isn’t designed for operation with high voltages, even if electrically it won’t immediately burst into flame. And for reasons below, I’d bet it’s immediately dangerous switching the 15A you’d expect to see from the kind of heating or cooling appliance you’d probably control with it.

Using It

So obviously I got this set up on the hot tub and it seems to do the trick.

The controller was plugged in and turned on around 5PM, and seems to have gotten through one off-on cycle.

Unfortunately, I couldn’t gather much data before the controller failed totally around 9PM.

Failure Analysis

The pump loading the controller runs at something like 60W, so about 5A from a 12V supply. Well within spec of the 20A relay. Unfortunately, I don’t think the thermal design of the board will allow for anything greater than very light loading.

Let me preface this by saying the board was sitting on a hot tub, with people in it, and failed around the time that one of them says she moved it to make sure it didn’t fall in while putting the cover back on. So it’s possible that she dripped some water on it and that’s what killed it, but it wasn’t wet at all, there’s no evidence it was ever wet, and I don’t think that’s what did it.

The controller powers itself with a switching converter from the screw terminals, so that in cases like mine, the same 12V switched to the load can also power the controller. Here’s what mine looks like, wired with barrel jacks.

Controller on bench with barrel jacks
And my assistant out of focus below the desk

Using a switching converter makes this device much more useful since, in theory, it should allow a pretty wide supply range at which you don’t need a second supply, if your load is DC anyway. The problem seems to be that the switching chip is located right next to the relay, and the relay coil, despite being rated for 20A, gets QUITE hot. I’m not sure if this is contact resistance at 5A load, or coil resistance at 12V input voltage, but the sum-total is that that area of the board gets toasty. While the device was working, I measured the plastic case of the relay at 120 °F – so the contacts and nearby copper were probably quite a bit hotter.

After failure, the controller as-wired put my bench supply in current limiting mode clamped to 2V, so something failed short. There’s a big pockmark on the 8051 microcontroller on the back side, so that’s probably where power is shorted to ground. Thermal imagery shows that, despite not doing anything at all, the 8051 gets pretty hot even with input supply limited at 200mA.

Thermal image of the main 8051 MCU at 96F
Pretty toasty for an MCU that’s not doing anything.

My theory is this: because of the thermal load around the relay/switching converter area, the switcher heated up past its rating, with the main switch finally failing closed. Because of the boost converter topology, this shorted my 12V external supply onto the internal system supply rail, which instantly blew up the 8051. In fact, it’s possible that the dead MCU is the only failure on board. If the switching converter shuts down with its main switch closed due to an over-temp condition, it’s entirely plausible that it still works, as long as the short circuit on its output is removed.

Thoughts About the Seller

I’m not quite sure how I feel about the seller. “Lucy” who contacted me specifically asked, towards the end of our interaction, that I not mention receiving the product for free, which obviously seems unethical and probably against Amazon’s review rules (and also against what seems to be the status quo on Amazon of mentioning this kind of thing freely). Also, when I contacted her again with these details, the email address was deleted. So that’s not confidence inspiring.

Next Steps

If I get around to it:

  1. Upload more pictures of the failed 8051
  2. Unsolder the 8051 and see if the switching converter still works, now that it’s cooled off.

Be First to Comment

Leave a Reply

Your email address will not be published.