Combustion Predictive Thermometer Reviewed And Rated

Combustion’s Predictive Thermometer is a new entry into the wireless temperature probe product category. These products typically use Bluetooth to convey temperature data from the probe’s sensors to a receiver or smartphone app. There are no wires connecting the probe, so the user doesn’t need to fret about routing cables. However, the internal electronics must be protected from high cooker temps, and this necessitates using the food as a heat sink.

In the typical wireless probe, two sensors measure the temperatures at the probe tip and the handle. The probe tip measures the food temperature and the handle measures the ambient temperature inside the cooker. Placement of the probe is critical to obtaining the core food temperature, and it can be difficult to measure the right spot and simultaneously protect the probe electronics. A poor guess might result in food cooked to the wrong temperature or a fried probe.

To address this problem, Combustion’s probe has eight sensors arrayed along its length. The internal circuitry looks at the measured temps and chooses the lowest value – the core temp of the food – and the food’s surface temperature. The sensor in the handle captures the ambient temperature. The dynamic selection of the sensor values removes the restriction of placing the probe in exactly the right spot, eliminating a possible source of measurement error. All three temperatures – the core temp, the surface temp, and the ambient temp are readable either on the repeater or in the app.

Another source of frustration with all WiFi devices is connectivity. Bluetooth is often criticized for its short range. This is especially true if the probe is in food being cooked in a closed space. The metal exterior of a cooker acts as a shield, preventing the signal from getting from the probe to the receiver. Combustion solves this problem by using a receiver/transmitter (a repeater in radio terminology) that can receive the weak signal from the probe and re-transmit it with more power. This boost allows a much greater distance between the cooker and the phone app. The use of the repeater is optional; if you’re close enough to the cooker, the probe can communicate directly with the phone.

Temperature accuracy is of paramount importance if one is to avoid under- or over-cooking. Undercooked food is potentially unsafe to eat, and nobody wants to overcook an expensive roast. Since we’re talking about food temps, we really only care for temps in the range of 100° to 210°F (38° to 99°C). The table below shows how well the Combustion probe measures the food temp:

Actual: Measured
100°F: 100.0
110°F: 110.0
120°F: 120.1
130°F: 130.1
140°F: 140.0
150°F: 150.0
160°F: 160.2
170°F: 170.1
180°F: 180.3
190°F: 190.3
200°F: 200.3
210°F: 210.2

Throughout the food temperature range, the probe was within 0.3°F, an excellent result. It should be noted that when first turned on, the probe acts as an instant-read thermometer and reads in whole degrees only.

The ambient temperature measurement is less critical than the food temp. In the predictive tests below, I used an oven temperature of 325°F (163°C) as measured by a Fireboard probe of known accuracy. The Combustion probe’s ambient sensor typically reads within 10-15°F of the Fireboard, better than most of its competitors. The maximum safe temp for the ambient probe is 575°F (300°C). While most ovens/cookers won’t get anywhere near this temperature, a steak on a grill could cause flare-ups that could damage the probe. In all fairness, such temperature flare-ups would likely damage most cabled temp probes as well.

One minute after the repeater loses the signal from the probe, the temperature display will show all dashes. If the app is reading the probe directly, without the repeater, the display will gray out. Unfortunately, there is no audible alert when either loss of signal occurs.

One frequent complaint about wireless temp probes is their diameter. Most devices I’ve tested are 0.25” (6mm). The Combustion probe is 0.19” (4.8mm), a 25 percent reduction. The useful length (minus the handle) of the probe is 3.8” (10cm), and the minimum insertion distance is about 2” (5cm). The repeater and the app can monitor four probes simultaneously by cycling through them.

Predictive Cooking: We all want to know when our food will be ready, and we’d like to know as soon as possible. One of the features of the Combustion wireless thermometer is its ability to predict how much longer the cook will take. Ideally, we’d like to know as soon as the food is placed in the cooker, but that isn’t realistic. To make a prediction, the probe must collect several data points that permit fitting the temperature changes to a known curve, and then calculating how much time will be required to reach the desired temperature. Sounds easy enough, right? Nope! Food temperatures won’t follow a simple exponential curve. Other factors creep in to complicate the calculation, like turning the food, evaporative cooling (aka the “stall”), changes to the ambient temperature (flare-ups), carry-over cooking, and other intangible factors.

The Test Procedure: I simulate roasting a piece of meat by using a rolled-up sock that has been nearly saturated with water. I shaped it to resemble a typical steak of 1.5” (38mm) thickness. The dimensions are adequate to insert the probe beyond the minimum depth. I insert the probe into the middle of the sock, in parallel with a conventional cabled probe for reference. I positioned a conventional probe, held in place with a grill clip, adjacent to the wireless probe’s handle as an ambient temp reference. I place the sock and probes into a Breville oven and adjust the temp to 325°F (163°C) as reported by the reference ambient probe. I then monitored the temps and times as reported by the probe under test.

Consider the following scenarios: Cook to 130°F (54°C): This is pretty straightforward because there won’t be any stall or big changes in the internal moisture content during the cook. I started the “cook” by placing the apparatus into a pre-heated (325°F or 163°C) oven at 1351 (1:51 p.m.) and starting a stopwatch. The phone app displayed a percentage (of what??) for about nine minutes when the percent was replaced by a time to finish. The first prediction was displayed at 1400 (2 p.m.) for 13 minutes, with the cook prediction at 1413 (or 2:13 p.m.). Eleven minutes later, at 1411 (2:11 p.m.) the time to completion showed six minutes, or a completed time of 1417, an increase of four minutes from the first prediction. At 1420, the set temp of 130°F (54°) was reached. The actual time to do the cook was 29 minutes, or seven minutes longer than the first prediction. The closer the temp was to the target, the closer the estimate was to the actual time required.

Cook to 160°F (71°C): This scenario is more complicated because the evaporative losses nudged the sock into a classic stall where the temperature rise stops. This confuses the prediction algorithm, resulting in ever-longer predictions of done time. We started at 1422 (2:22 p.m.) with the sock at 49°F (9°C). After 11 minutes (1433, 2:33 p.m.) the predicted done time was computed to be 1453 (2:53 p.m.). At 1453, the time the initial prediction said would be the done time, the time remaining was still 20 minutes, or 1513 (3:13 p.m.). By 1510 (3:10 p.m.) the predicted done time was 1554 (3:54 p.m.), an increase of over an hour beyond the first prediction. The time predictions kept getting longer as the sock “stalled.”

Unless there is a flaw in my test procedure, I think that the prospect of an accurate prediction early in a cook is still an elusive goal.

The build quality of this thermometer is high. It is easy to use. I’ve been told to expect additional features in future firmware updates. Plots of temp vs. time would be useful, as would audible annunciations of communication loss. The countdown timer in the repeater doesn’t sound in the app.

Both the probe and the repeater have internal rechargeable batteries that are topped off using a supplied USB cable. There is no USB wall wart, however.

I prefer to see a printable user’s manual rather than going to the company’s website to ferret out the features. Call me old-fashioned, but I like to have a piece of paper with instructions that I can highlight. A QR code doesn’t cut it. (Maybe I’m asking too much of a product that is in the early development stages?)

How to rate this product? At this point in time, the prediction algorithm is still immature. The feature set is comparatively sparse but expected to grow. Temp accuracy is very good. It’s comparatively pricey, but the quality is high. I think we might be looking at a Platinum rating when the product is more fully developed. For now, I give Combustion a solid Gold.