anemometer

Mulitcomp pro used to be Tenma. At least it's designed in Taiwan.

Cheers

After a couple hundred pseudorandom button pushes I managed to get it to show velocity in FPM. Maybe I'll never have to push those buttons again.

We have 8 pcb's in a box. The two fans are fire hoses of air, so some boards get a lot and slot 3 gets almost none. Air is perverse.

The new idea is to make a big rectangular plastic baffle that hangs down from the top cover, just in front of the boards. We'd punch some hole patterns in that, low air transmission near the fans and very open where we need it. We need the hot-wire probe to sneak in between the boards, through 8 small holes in the top cover, and see how we're doing as we tune the baffle.

The idea is a wiffle ball. Fail.

Quite cheap. Guess they are using calibration techniques to get it to have only 1% accuracy. I mean, for a hot wire anemometer...

The idea is controlled air flow impedance into each slot. Why would that fail?

The thing about air flow is that a lot of people make intuition-based guesses, with mumbled keywords, and are usually wrong. We are tuning this by experimant, taping over holes in the baffle until we get fairly balanced air velocity across the boards.

Similar hot-wire gadgets cost from less than $200 to close to $2000. The Flukes and Omegas are over $1000.

The little propeller things start below $50.

20% accuracy would suit us. We are trying to tune a system to get approximate equal air flow through 8 channels, so we really don't need accuracy.

I made my own hot-wire anemometer once... broke the glass off a small light bulb as the sensor. We calibrated it using a bicycle with a speedometer. The multicomp has a ceramic thick-film sensor, probably a heater on one side and a thermistor on the other. Not literally a hot wire.

fredag den 19. august 2022 kl. 16.55.35 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

eight RTDs and eight voltmeters? in series with enough current to get some self heating

Yes, you could make a good flowmeter with just one RTD. I guess we could have an air flow sensor on every production board, but that's kinda silly. We will have temperature sensors on every board and servo fan speeds.

Our test box has 8 small holes in the top cover, so we can poke in the anemometer probe.

I was thinking of a modulated heater with a temp sensor downstream. Measure air transport delay.

I was questioning the direction of air flow in some of the slots between our PC boards. It's not inconceivable that some slots could have reverse flow, and a hot-wire flowmeter won't show that. A little strip torn off a Kleenex shows the sign. They were all the same way, front to back.

Maybe the propeller meter things indicate direction. Or you can at least look at the spinner. They need spirals, like on a jet engine.

fredag den 19. august 2022 kl. 18.14.41 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

I was just imagining using eight sensors for testing, instantly tells you if the flows are the same no need for absolutes

A flight instructor realized that I was using differential braking to steer a AA5B Tiger down the runway on takeoff. Don't do that, use the rudder, he told me.

Your baffle appears to use the same logic as brakes on takeoff.

Look at the wing flaps of the Douglas dive bomber.

Brakes? Takeoff?

Logic even?

They were trying to solve a very different problem.

What's interesting is that airframe designers have the smartest people and the best fluid dynamics software and the biggest supercomputers available. And still use wind tunnels.

Iterating the baffle holes with tape takes time. Poking one probe into

8 holes in the top cover is really incremental.

So far, varying fan speed doesn't seem to affect the relative distribution of air in the 8 slots.

fredag den 19. august 2022 kl. 20.45.47 UTC+2 skrev John Larkin:

formula one has for years had limit on how much CFD and windtunnel time they can use they can mix CFD and windtunnel time with 1 TeraFLOP = 1 hour windtunnel

I did an experiment once where we had a problem with the orientation of a product. So burned power into a resistor on the PCB, surrounded by 4 NTC resistors. We did in fact detect orientation. in 2 planes. But then, what about the third plane?

<snip>

There was a DIL chip for use as an inclinometer which had a miniature heater and two thermistors plus signal processing, and used the differential temperature due to air convection to provide tilt. It was a long time ago, before MEMS.

There were also electrolytic (liquid) inclinometers.

I designed a level sensor system for the C5A Boresight Alignment Kit. It used a Talyvel inclinometer, basically a pendulum and LVDT.

Electrolevels could be very accurate and reliable. Long ago, I installed many to monitor slope stability on an embankment and movement on a flyover in Wales, all cabled back to a hut which would get visited less and less frequently. I expect they're still there after forty years or more.

They obviously work, because the embankment hasn't slipped yet :-)

Omega is garbage these days. They struggled and failed to find a replacement for one of discontinued temp controllers. They clearly had no idea what they used to make, or what they make now.

Went with Novus. Weird pre-sales support (they're in brazil), but it was legit and they have real distributors in the US. Will bump to the top of the list for anthing temp control. The omron stuff is just overly complex.

Do you care about temperatures or the airflow numbers? It's sort of shocking that the cooling is poor/uneven in that chassis illustration.

We want air flow over the eight boards to cool parts. Some will have local heat sinks here and there. I was hoping for 200 LFPM on the worst board with the fans running at some tolerable noise level.

Any board that thinks it's getting too hot can request that the fan speeds be increased.

As noted, the two fans are fire hoses of air. Slot 3, about in the center, gets about a tenth the flow of slot 1, which gets a full fan blast.

Blowing air out the front panel (it currently intakes) has upsides, probably better distribution between boards. But it has downsides too.