Cheap and small - how to measure airflow

[paulca]

What would be the easiest and cheapest way to get "indicative" airflow readings?

The application would be in fan monitoring in a PC case.  So I'm not looking for highly accurate, calibrated measurements, just indicative "none, slow, medium, high" airflow.

As the air speeds might be quiet slow I can probably exclude spring vanes on a pot.  This probably excludes pitot sensors also. 

As space is confined I can probably exclude spinning cups annometer style.

That really leaves mini fan/dynamo style things, hotwire sensors or ultra-sonics?

I almost excluded pitot sensors, but would a basic barometric air pressure sensor (or two) not be able to give an indicative reading based on the difference between the sensor in the tube versus the sensor outside of the tube?  barometric air pressure sensors with 1/10th millibar resolution are cheap as chips.

[Rerouter]

feed some power into a ptc so its above nominal ambient for your region in still air (e.g. 50C) as the airflow increases, the current to maintain that temp will increase, giving you something you can measure, this covers both fan based airflow, and externally forced airflow, (mass airflow sensor)

otherwise, as its fan monitoring, you cant just watch the tacho feedback signal? a stalled fan doesnt provide a pulse.

[paulca]

feed some power into a ptc so its above nominal ambient for your region in still air (e.g. 50C) as the airflow increases, the current to maintain that temp will increase, giving you something you can measure, this covers both fan based airflow, and externally forced airflow, (mass airflow sensor)

otherwise, as its fan monitoring, you cant just watch the tacho feedback signal? a stalled fan doesnt provide a pulse.

Thanks.  The thing that puts me off the hot-wire technique is that a PC case is inherently dynamic in ambient temperature.  So I would have to monitor the ambient temp and raise the hot wire above that temperature as the case heats up.  Sounds like a nightmare to get any calibration at all.

As to monitoring fan tacts, that tells you the fans are turning, but doesn't show you the effective airflow at a particular point in the case.  Fans working in concert for example, push, pull, intake versus exhaust etc.

To be honest I'm thinking sticking a feather in the case and putting an LED up lighter on it might be better!  Or handing a bit of paper in the case might be more functional.

[jpanhalt]

If you smoke, try a cigarette.  If not, try a match or cigarette lighter.   Flames are quite good at detecting and grossly quantifying airflow.   Hard to beat the price too.

[Rerouter]

The hot wire approach just needs you to pick your maximum temp, set it to run at that in still air, and off you go, because a PTC is already a temperature sensor.

Otherwise you may have some luck with optical flow sensors that they sell for hobby drones.

[Gyro]

These days, those cheap little glass encapsulated thermistors give a better swing than hot wire. Put one in the airflow, one shielded, connect them in a bridge configuration and the size of the offset gives you the speed of the airflow. You need to have enough current flowing through the bridge to cause them to self-heat a little.

[paulca]

Thanks.  I think I'll try the dual PTC method.

The plan is to make some dynamic LED lighting for PCs that, critically, does not require facing the sheer mess of interfaces that motherboards present to people.  Add-on modules are no better either.  The software available for these is very pretty and next to bloody useless, only works in Windows, etc. etc.

So small SMD modules running 0.5m WS2812 strip LEDs off a 12V molex or SATA lead that don't require I2C or proprietary interfaces to function.

[mycroft]

Out-of-box idea: a microphone. The sound level should be proportional to fan speed and air flow.

[NiHaoMike]

I have seen a Nidec case fan that could sense airflow restriction using some DSP trickery and report that to the server it's in using a fifth wire.

I remember about small incandescent bulbs, with the glass removed, make surprisingly good airflow sensors. Basically a DIY version of the airflow sensors often used in car engines. (Which could possibly be another option!)

[paulca]

So if I was using the twin PTCs in a bridge config what would be the easiest way to shield one of them?

For a prototype I expect the device will be on proto-board wrapped in heatshrink, would the reference PTC being under the heat shrink be enough?

[SiliconWizard]

I was thinking of a small fan used as a sensor (which shouldn't be too difficult). There are two drawbacks though: one is that it will be somewhat directional (something you may or may not want depending on what you're trying to achieve), the other is that it will have a threshold effect (enough airflow to make it start spinning) with some hysteresis (will require more airflow to make it start spinning than to maintain its rotation) that could be a problem, again depends on your requirements.

[max_torque]

why do you need to monitor the airflow?

if the fan has failed, your device gets hot. At some point, it turns off to protect itself?  Just monitor it's temperature surely?

[BravoV]

As its for PC, most modern PC fans have the sensor 2 pulse/rotation line, why not just tap into that cable as its open collector ? Cmiiw

[C]

If you look, current generation 4 wire PC fans make controlling fan speed and feedback of fan speed easy. These signals lets you check for fan problems.

Here is a apt sheet for PSOC
http://www.cypress.com/applications/fan-controller

Now knowing commanded fan speed and actual fan speed you can temp control a space by adding a temp sensor.

As you increase fan speed the output air temp sensor should show a decrease. This change is air flow.

C

[paulca]

If you look, current generation 4 wire PC fans make controlling fan speed and feedback of fan speed easy. These signals lets you check for fan problems.

As its for PC, most modern PC fans have the sensor 2 pulse/rotation line, why not just tap into that cable as its open collector ? Cmiiw

Have you tried to interface with the motherboard to get that data in a reliable way?  I expect not, because it depends on a whole raft of things such as whether you are in UEFI or and OS and which one?  "I just run the tool that came with the motherboard.", these tools are gimmicks and barely work.

The access and control protocols are a mess with no two manufacturers implementing the same things.  Add on fan controllers are worse as they are usually proprietary drivers with expensive hardware and dodgy pretty but useless front ends and...

Few currently work reliably on (b)leeding edge motherboards.  Including the one that came with mine!

The net is... I want to create a device that does not rely on interfacing directly with the motherboard or a fan controller.

why do you need to monitor the airflow?
if the fan has failed, your device gets hot. At some point, it turns off to protect itself?  Just monitor it's temperature surely?

Thermal management.   When you have a device which has multiple large (several hundred watt) heat sources, multiple different temperature sensors to monitor, multiple fans to control, knowing enough air is flowing at the right times and in the right direction is very important.

Thermal protection is not graceful or even reliable and is always undesirable.  Quite often "thermal protection" is just the system crashing in an uncontrolled fashion which is undesirable.  Sometimes it's just a controlled "HALT" instruction, still not desirable.  It can even happen that all reported temperatures are within tolerance and 30 minutes later at those temps it locks up due to heat soak in other areas cause by high ambient temperatures inside the case.

You can just set all the fans to 100% and sit listening to a small hurricane under the desk filling itself full of dust 4 hours a day or you can make the fans dynamic.  Silent while browsing the web and hurricane when over clocked in a modern AAA rated game.

I'm not (yet) interested in producing a fan controller, I "am" using one of those annoying desktop systemtray apps, while it urks me.

The idea was to make some useful and completely decoupled RGB lighting gizmos that provide instrumentation inside the case.  (Most gaming PCs these days have glass side panels and RGB lighting). 

Ambient Temperature - Easy enough to give a ball park between a low and high level
Ambient noise level - Easy enough to give a ball park between a low and high level
Air flow / Cooling effect / Wind chill factor - Not so easy, so I started asking.

This would form a small (as I can make it) controller which takes 12V from a fan rail and shows several pretty LED bar graphs which can be mounted somewhere in the case to be both pretty and functional.

[westfw]

Have you tried to interface with the motherboard to get that data in a reliable way?

You don’t need to interface to the mb, just pass power and passively tap the “speed” signal.

[BravoV]

What would be the easiest and cheapest way to get "indicative" airflow readings?

The application would be in fan monitoring in a PC case.

As its for PC, most modern PC fans have the sensor 2 pulse/rotation line, why not just tap into that cable as its open collector ? Cmiiw

Have you tried to interface with the motherboard to get that data in a reliable way?  I expect not, because it depends on a whole raft of things such as whether you are in UEFI or and OS and which one?  "I just run the tool that came with the motherboard.", these tools are gimmicks and barely work.

The access and control protocols are a mess with no two manufacturers implementing the same things.  Add on fan controllers are worse as they are usually proprietary drivers with expensive hardware and dodgy pretty but useless front ends and...

Few currently work reliably on (b)leeding edge motherboards.  Including the one that came with mine!

The net is... I want to create a device that does not rely on interfacing directly with the motherboard or a fan controller.

With above useless ramblings, again, just google the rotation output signal pin/wire from the fan, popular named tacho line, and again ... "TAP" into that to get the "indicative" airflow reading as you originally requested.

No need to touch mobo, os, bios, uefi, fan controller at mobo, fan controller and etc.

Hope this helps.

[BravoV]

Have you tried to interface with the motherboard to get that data in a reliable way?

You don’t need to interface to the mb, just pass power and passively tap the “speed” signal.

Don't think the OP gets it.

[paulca]

Have you tried to interface with the motherboard to get that data in a reliable way?

You don’t need to interface to the mb, just pass power and passively tap the “speed” signal.

Don't think the OP gets it.

No I do.  The speed of a single fan does not tell you air flow.  It doesn't matter if you have a dozen fans at full RPM you can end up with zero air flow if they are mounted wrong.

Not all fans are spinning at the same time, the same speed or for the same reason.

In particular I have 3 fans dedicated to case flow and 2 with a case air flow as a side effect only.

The CPU has two fans on the radiator which only respond to the CPU temperature.  As a side effect they intake (very slightly warmed) air into the case.  When the system is cold (mobo temp < 32*C) and the video card is cold (under 45*C) these are the only fans running.  They run nearly silent at 600rpm constantly, but spin up on a curve until 100% 1400rpm and noisy when the CPU reaches 70*C.

I have 1 general case flow fan which starts at motherboard temperature of 32*C and reaches full rpm at 40*C.  It's also an intake.

I have 2 fans dedicated to purging the video card heat out of the case starting up when the video card reaches 45*C and reach full rpm when the video card heatsink reaches 60*C.  One intake below the video card, one exhaust above it.  This is the only exhaust fan.

The large inbalance of intake to exhaust is by design to maintain positive case pressure resulting in air only ever exiting via the various vents and grills and thus not inhaling dust.  The intake fans have filter screens which are removable and cleanable.

There is also the water pump tact, but in my case this is running at 100% constantly as it's a silent maglev type impeller anyway.

All this gizmo is there to tell you is that the internal case is being cooled and a rough gauge as to how aggressively. 

It's 20% functional, 50% a gimmick and 30% a pretty RGB light for the case.

I suppose to implement your suggestion I could add dupont headers for half a dozen fans and show half a dozen read outs, but... not all fans have the same rpm range, so it would need calibrating (though it probably will anyway).

In a poorly designed case an airflow sensor should reveal that when the fans intended to cool the case spin up there might not be too much airflow.

Obviously the important one is the temperature as various sensors are located in various places, the closest to a general ambient internal temp is the motherboard temp, but that is I believe that is located quite close to the CPU and RAM.  Thus an actual ambient temperature read out with the sensor in free air near the top of the case would be very useful.

All of the above is moot if you are in an OS without access to the temp sensors, fan tacts or PWM controls, such as MacOS or Linux.  Then you are solely relying on the BIOS fan controls and your ears/eyes to know things are being cooled correctly.

EDIT:  If you think the fan set up is overkill, with the out-of-the-box setup, the system looked fine when loaded, regarding the spot temps of the various heat sources such as CPU and GPU, but a temp probe on the exhuast revealed temps of 47*C and a hard lock up was only about 45 minutes into playing an intensive game.  With the setup I have now exhaust temps don't exceed 36*C and all spot temps are 5-10* lower.  It still virtually silent when idle.

[BravoV]

What would be the easiest and cheapest way to get "indicative" airflow readings?

The application would be in fan monitoring in a PC case.  So I'm not looking for highly accurate, calibrated measurements, just indicative "none, slow, medium, high" airflow.

I guess the word "indicative" means differently in my brain as English is not my native.

Well, good luck on your quest.

[paulca]

My gut feeling was originally some form of mechanical device.  Like a friction-less potentiometer with a very light spring on it.  Or a bit of plastic tape with some way to electronically measure the angle it is being deflected to.

[german77]

Air flow generates pressure, you can measure pressure the same way you mesure weight. With a strain gauge with little electronics is needed to get it to work. The only calibration you need is to measure no airflow vs full airflow. With very low time response and not worrying about temperature.

For the dual PTC may not work to measure air flow. The electronics it's actually the same as the strain gauge. They are unidirectional cheaper and easier to install in a small case. with a really big downside, the flowing air will be warmer than the ambient and will be almost imposible to know if you have low air flow or it's just warmer.

Thinking on your case. It seems easier to make a small PTC 2D array and instead of measuring air flow you measure hot spots. Then you can increase fan speeds according to the temperature of those hot spots. This way you can detect if there are places were there is no air flow as the temperature increases in that place. The electronics for this is just an voltage divider. No need for calibration as PTC are already characterized and precise values aren't needed. 

[Teledog]

What about a cheap <$20 anemometer from fleabay?
Shows airflow & temp, hack it as you may..for the price..

[CatalinaWOW]

if you don't have airflow you get heat, so a simple thermistor should tell you what you need to know.  Place it at the location you are concerned with.  The varying ambient temperature could be dealt with using existing temp sensors on video card and mother board.  That information could be used several different ways depending on the details of what you are after.

[JohnnyMalaria]

Perhaps some ideas here? http://www.hobbyprojects.com/S/Air_Flow.html

[Jwillis]

 Just a thought
Why not use a hall effect wheel from a mouse attached to a fan .Measure the signals as the fan turns the mouse wheel through a circuit of some sort .Maybe one of you really bright guys could come up with something . 

[max_torque]

Thermal management.   When you have a device which has multiple large (several hundred watt) heat sources, multiple different temperature sensors to monitor, multiple fans to control, knowing enough air is flowing at the right times and in the right direction is very important.

ok, so use several temp sensors? The important word in THERMAL MANAGEMENT is the word thermal, ie relating to heat!  Close the loop on the temp error (measured vs target) apply a gain to that error and then the result to your fan speed controller and bingo, a perfect system!  Everything else falls out in the wash!  And if your output hits the end stops (ie max) then you know your system is incapable of moving enough heat, due to component failure or un-specification etc.

You'd be crazy to start worrying about measuring airflow when it is completely irrelevant.......  (and hence no one does)

[paulca]

Thermal management.   When you have a device which has multiple large (several hundred watt) heat sources, multiple different temperature sensors to monitor, multiple fans to control, knowing enough air is flowing at the right times and in the right direction is very important.

ok, so use several temp sensors? The important word in THERMAL MANAGEMENT is the word thermal, ie relating to heat!  Close the loop on the temp error (measured vs target) apply a gain to that error and then the result to your fan speed controller and bingo, a perfect system!  Everything else falls out in the wash!  And if your output hits the end stops (ie max) then you know your system is incapable of moving enough heat, due to component failure or un-specification etc.

You'd be crazy to start worrying about measuring airflow when it is completely irrelevant.......  (and hence no one does)

That much - fan control based on temperature - I already have.  Slightly more complex using software curves for temp to rpm, either BIOS or Windows software.

The original idea behind this device was to do something useful with RGB lighting instead of it just looking pretty and useless.  Without having to use the clunky Windows based software APIs to the motherboard.  I would design them as a bit of fun and possibly pass on the schematics or even excess boards to other PC enthusiasts as hobby projects.  Potentially moving up to kits.  Effectively combining "maker space" electronics and PC modding.

So I selected a few things that might be useful that could be displayed with an RGB strip.  Spot temperature, air flow, noise level.  I envisioned that while gaming with headphones on, you could simply look at the PC, see the temp lighting turning amber, but the airflow and noise bar graphs increasing and feel a little reassured.  Still a gimmick though.  That said I still want an ambient lighting strip that changes based on actual case air temp, so I'm starting there.

It is however seeming that airflow might not be the best choice.

[IDEngineer]

Reading through this, I'm a fan (no pun intended) of using an IC pressure transducer. SPI/I2C interfaced parts are under $2 at Qty25, a bit more for analog voltage out. If you have an MCU with SPI/I2C you can get 16-24 bits and make decisions in the digital domain; otherwise a comparator can convert the analog voltage to a go/no-go signal, or run the analog voltage into an MCU's A/D. Either approach gives you the "none, slow, medium, high" thresholds mentioned in the TLP.

http://www.te.com/commerce/DocumentDelivery/DDEController?Action=srchrtrv&DocNm=MS5637-02BA03&DocType=Data+Sheet&DocLang=English is $1.73/25's at DigiKey, over 17K pieces in stock so you won't get orphaned. Point its hole in the direction of measured airflow and you're done. 3x3x0.9mm is about as small as any solution imaginable. No moving parts. Ultra low power consumption. Four wire interface.

There have been several clever suggestions in this thread, some of which would be very fun to play with, but for a 1) quick-to-implement solution, 2) compatible with the specified "confined space", 3) very likely to work with few unknowns, 4) at minimal cost, I think the pressure transducer wins.

[aandrew]

I have some interest in this.

Steven Woodward has an interesting circuit that has unfortunately been eaten by the many EDN acquisitions. It provides a linearized pulse width which is trivial to read with any microcontroller and with good resolution.

The base URL for Design Idea 1927 is https://www.edn.com/edn-03-14-96-self-heated-transistor-digitizes-airflow/, but the link to the schematic is missing. The original circuit, also from Mr. Woodward, is EDN Design Idea 1841, from the March 14, 1996 EDN. I spent a good two weeks, including contacting Mr. Woodward himself, trying to find a clearer copy of the schematic and ask him some questions about shrinking the design to use tiny SC70-6 or smaller transistors (with a reduction in drive current) but did not get much of a response, and of course EDN is a completely dead end. Infuriating.

I've recreated the schematic and made a couple boards for testing. It works reasonably well and is *cheap*. I would be absolutely thrilled if some of the more experienced analog people here could help me understand how the circuitry around Q2, Q4 and U1B linearizes the output, and how I might scale the current and temp differential between Q3 and Q5 down to work with very tiny physical sense transistors, targeting airflows in the 80-120LPM (2-5 CFM) range, in 15-35 degree Celsius air.

[DenzilPenberthy]

Im working on a project now using these flow sensors. Very easy to use and have an I2C interface to read out the flow rate.

https://www.mouser.co.uk/ProductDetail/Sensirion/SFM3300-D

They have other ones in the range which measure higher volume:
https://www.mouser.co.uk/new/sensirion/sensirion-sfm3019-digital-mass-flow-meters/

[kripton2035]

@aandrew:
Elektor did something similar with less and more modern components here :
https://www.elektormagazine.com/articles/simple-wind-speed-meter-using-3-transistors-and-an-opamp

[spostma]

nice to see someone who remembers and loves the jewels that Steve Woodward published in EDN.

Every circuit Steve designed is simple and elegant, and solves analog problems by utilizing
the law of physics that we usually rather ignore.

I learnt a lot from his work; he is a true master of analog electronics!