The tube seems a little short to me. The ratio of length to diameter seems smaller than Mizuno's. I suggest you make it considerably longer and then -- as I said -- crush it down at a point about one-third of the length from the blower. To form a Venturi. That should mix the air. It may not be exactly what Mizuno has, but the point is to make the flow rate the same at every point in the orifice. Right? Pretty much the same. Who cares how that is done, as long as it works.

By the way, you can confirm that Mizuno's flow is well mixed. You can confirm that his calorimeter read a single value everywhere across the orifice, and it continued at that value over time. You can do this by looking at the data for low power calibrations, and then running the equations backwards. That is to say:

Assume the average for input power, inlet and outlet temperatures are correct. Use the STP value for specific heat of air (it hardly changes anyway). Assume there is no heat lost from the walls, because very little is lost below ~50 W.

EXAMPLE:

For a 30 W calibration, the averages are: 29.69 W input power, 1.75 K Delta T, heat capacity 1.005 kJ/kg*K (close enough for government work: https://www.ohio.edu/mechanica…tables/air/air_Cp_Cv.html)

Energy kJ/s = Weight kg * specific heat kJ/kg*K * temp K

So, Weight kg = temp K (Energy kJ/s * Specific heat kJ/kg*K)

1.75 K / (0.02969 kJ/s * 1.005 kJ/kg*K) = 0.01688 kg/s of air

Based on the blower power, and assuming the flow rate is uniform, and calibrating blower power to the anemometer and cross checking . . . Mizuno computes the average weight of air per second is 0.01679 kg/s. Which is very close to 0.01688 kg/s.

It is a little lower because, in fact, there are losses from the calorimeter walls. Taking one thing and another, and adjusting specific heat to the actual temperature, Mizuno computes the output captured in air is 29.04 W, a little lower than 29.69 W input.

Anyway, we can work back from the weight of air 0.01688 kg/s to derive the blower flow rate in m/s. Assume the orifice really is 66 mm in diameter. THH will not accept that until we measure it to the nearest nanometer, but let's just assume it is 66 mm. Run the numbers and you get a flow rate of ~4.1, which, lo and behold, is what Mizuno recorded. Okay, let us add some meaningless digits of precision to satisfy THH: Mizuno's average is 4.12171222. (Actually, no matter how many digits I provide, he will say he needs more precision; 10,000 digits would not suffice.)

I did this for 10 W and 50 W as well. I have calibration data for 80 W, 100 W and so on, but it does not work as well, because losses from the walls become significant. We can account for these losses by deriving a function, as shown in Fig. 3. I have two functions: one to compute percent losses for input power, and another for the temperature of the reactor surface (which is what Fig. 3 shows). However, these functions depend on the flow rate being uniform and correctly measured, so you cannot confirm the flow rate with the function.