5
\$\begingroup\$

I have a server room with an air conditioning unit. The unit is supposed to run 24x7, but ND tends to get 100 °F in the summer and -40 °F in the winter. This is a little harsh on the equipment and we have had failures. I have a temperature sensor, but I want a little more warning.

I am trying to cook up some solution that monitors the air flow and sends an e-mail when the air stops. I have looked at various anemometer (wind speed indicators), but they all seem to log data and not provide updates, plus they all seem to be rather large and not a good thing to have spinning in a server room. I want to hang the sensor from the ceiling in front of the air conditioner.

\$\endgroup\$
  • 1
    \$\begingroup\$ Are you here to make your own circuit or are you wanting to buy one off the shelf? \$\endgroup\$ – AbstractDissonance Jan 25 '11 at 23:04
  • \$\begingroup\$ I'm cool with either. I would need some guidance on what parts to buy either way. I just need to make sure I cover both the detection and reporting issues. I actually would prefer to build something. \$\endgroup\$ – mccoma Jan 26 '11 at 0:26
  • 1
    \$\begingroup\$ Sending an email is a very different problem from detecting airflow or temperature. The sensors can be monitored with a $0.25 microcontroller, but email requires a much, much, much more complex system. Do you have a computer that could run a program to continuously monitor a serial, USB, or Ethernet connection and do the email sending from there? Serial is cheap and easy, USB is in the middle, and Ethernet is the hardest and most expensive. (That's probably a silly question in a server room, but I don't know). \$\endgroup\$ – Kevin Vermeer Jan 26 '11 at 14:12
  • \$\begingroup\$ yeah, I've got an OpenBSD box and an OS X 10.6 box that could do the heavy lifting and poll a sensor. The OpenBSD box has serial and USB. The OS X has USB. \$\endgroup\$ – mccoma Jan 26 '11 at 15:59

13 Answers 13

7
\$\begingroup\$

For a very simple sensor you could hang two parallel strips of thin (very flexible) foil in front of the air vent. When air is flowing they'd end up pushed together, giving you (essentially) a switch closure you could read.

You'd still need something to notice and give a yell, like the Arduino + ethernet shield mentioned elsewhere, but this is about as simple a sensor as you could get.

\$\endgroup\$
  • 1
    \$\begingroup\$ I was going to suggest the same thing. They won't make a continuous connection, but if air is moving, they'll connect intermittently, so you'd be looking for them to stop making contact for a long period of time or make continuous contact (stuck together). Neither of these states should last for long if air is moving. Use several pairs of strips for redundancy, connected to digital inputs of a micro. If all the inputs remain unchanged for a few minutes, the air is still. \$\endgroup\$ – endolith Feb 15 '11 at 21:59
4
\$\begingroup\$

Use a hot-wire anemometer, they can be very small.

\$\endgroup\$
  • \$\begingroup\$ They are small, but I do worry about having a powered element hanging from the roof. \$\endgroup\$ – mccoma Jan 26 '11 at 0:30
  • \$\begingroup\$ Why are you worried about it? \$\endgroup\$ – Leon Heller Jan 26 '11 at 2:56
  • 1
    \$\begingroup\$ I've heard of using a diode junction (actually an abused transistor) as an anemometer; jam some current through it to heat it up, then measure the voltage which will vary with temperature. The accuracy pales in comparision to a hot-wire anemometer, but if you're looking for a binary flow/no-flow should work. \$\endgroup\$ – Nick T Jan 26 '11 at 3:48
  • \$\begingroup\$ I've built hot-wire anemometers using transistors. I used sot23-6 devices (dual transistors) but you could use two bigger ones with a little less accuracy. The idea is to use one as a temperature sensor and the other \$\endgroup\$ – akohlsmith Jan 26 '11 at 16:17
  • \$\begingroup\$ I've built hot-wire anemometers using diode-connected transistors. I used sot23-6 devices (dual transistors) but you could use two bigger ones with a little less accuracy. The idea is to use one as a temperature sensor and the other as a heater. You have a DAC or PWM to vary a current to one of the transistors and a simple PI loop. The amount of current to keep the transistor at temp is proportional to the airflow over the pair. You will need to calibrate for air density (humidity) but I'm guessing in your application this doesn't change much. \$\endgroup\$ – akohlsmith Jan 26 '11 at 16:23
4
\$\begingroup\$

Perhaps a pressure meter would do the trick? Use a tube and align it so that the airflow is going into the tube. Use the pressure meter at the other end of the tube. When the airflow stops, the pressure should drop. Sparkfun sells a barometric pressure sensor that, "Under ideal conditions, this sensor can detect the pressure difference within a 9cm column of air." http://www.sparkfun.com/products/8161

For a mechanical solution, you could use a flex sensitive resistor strip (http://www.trossenrobotics.com/c/robot-force-sensor-fsr.aspx). Hang it so that the airflow will force the strip to bend. You'll have to be careful that the strip doesn't oscillate, but that would be very easy to interface to a microcontroller, or a comparator to turn on a LED or a buzzer.

\$\endgroup\$
  • 1
    \$\begingroup\$ I was thinking of a strip, but I was reading about the fatigue and wondering if being constantly bent would be a problem. As to the reporting, I actually need to have something that can tie into a computer (serial, usb) or send e-mail itself. I am generally not here when it all goes bad (2 and 3 in the morning last time). \$\endgroup\$ – mccoma Jan 26 '11 at 0:31
  • \$\begingroup\$ @mccoma Why would something like that break when you were around, that would just be too convenient. \$\endgroup\$ – Kellenjb Jan 26 '11 at 5:56
  • \$\begingroup\$ @mccoma good point about fatigue. Maybe having it oscillate would then be a good idea. You'll just need to measure the oscillation and determine if/when it stops to detect an airflow problem. \$\endgroup\$ – gallamine Jan 26 '11 at 15:27
4
\$\begingroup\$

For something this critical, I'd seriously consider buying off the shelf.

But if you want to roll your own the simplest solution is two temperature sensors. One in the airflow and one in a "dead" air spot. With air flowing, measure the two sensors and apply the difference to one as an offset (static_offset below) so the difference == 0. This is your calibration step. Due to self heating the sensors should work as a hot wire anemometer and when the air flow stops, the temp difference will go up and that will be your trigger.

Basically you are constantly reading (T1 - T2 + static_offset). When this calculation exceeds a threshold you identify, send the alarm.

\$\endgroup\$
1
\$\begingroup\$

If you want to monitor just one fan, maybe a Hall sensor on the fan for the rpm.

\$\endgroup\$
  • 2
    \$\begingroup\$ The unit I need to monitor is a Sanyo KS1872 which is enclosed so I can't attach anything to the actual fans. (If I total missed on what you meant, sorry) \$\endgroup\$ – mccoma Jan 26 '11 at 23:37
1
\$\begingroup\$

How about buying a standard wind meter and hooking it up to one of your free computers. http://www.weathershop.com/wind_.htm

Or you can build one @ http://www.instructables.com/id/Digital-Anemometer-wind-meter/

\$\endgroup\$
1
\$\begingroup\$

A temp sensor in front of the outlet might be better than an airflow sensor, since as monpjc mentioned, the fan could continue to run even if the unit is not cooling. The temp there will generally be cooler than the ambient temp in the room, so you could probably find a threshold that would give an early indication that the unit is not cooling before the room temp rises noticeably. If the airflow does stop, the temp in the airstream area will quickly rise to the general room temp, so you are covered both ways.

\$\endgroup\$
1
\$\begingroup\$

I design equipment for data centres, so get where you're coming from here.

Gallamine is on the right track, you want a pressure sensor but a differential sensor like the ones you can get from RS or Farnell. Some are just switches and some have an I2C interface or voltage output that you could read. You can then look at the outlet air press compared to some place out the air flow.

I would also use temperature, as I guess there could be times the unit will switch off the fan when down to low temperatures and is not required - depends on design, but normally run 24/7.

Microchip makes the explorer 16 kit which you can get an Ethernet plug in module that allows you network access. They also have a full TCP/IP stack you can load onto it and this will allow you to send emails - I know cause we have done it. Would also give you a web interface so you can go look at pressure / temperature remotely to have the feel good factor its still working.

Other options include an Arduino and an Ethernet shield that again will give you networkable access - seen people sending tweets as alarms.

Others are right - not a quick simple project but possible. Finding an off the shelf unit in a one off will be hard and you may well have to give on your requirements.

\$\endgroup\$
  • \$\begingroup\$ You might also check out the XPort or WiPort from Lantronix. They do serial<->ethernet and also have GPIO built in. \$\endgroup\$ – gallamine Jan 26 '11 at 15:28
0
\$\begingroup\$

Surely the core need is to monitor room temp, which you say you are already doing, so isn't trying to monitor airflow just overkill and creating unnecessary complexity? At the end of the day: room = cool enough = happy you?

When I looked into automating the monitoring of the server room in which I worked, I ended up with a USB temp sensor off ebay for very little indeed and we just plugged it into one of the PCs already in the room. There were plenty of apps on the internet to read the sensor and dump the temp to a file and I had a simple batch file to read this and send an email if the temp rose above a set value.

My only extravagance was to look at a USB IR Tx/Rx module that I planned to program to learn and send the aircon on/off commands as the room was a semi-office area and could be adequately cooled by opening some windows when the room was occupied, but once in a while the occupants would forget to turn on the aircon again when they closed up and left for the night. Ultimately, I left the job before putting this part of the plan into action.

\$\endgroup\$
  • 1
    \$\begingroup\$ I think what this person is asking for is a much quicker alert. If your just measuring temperature, then the room takes time to get to the alert threshold level. If your measuring airflow, then you'll get the alert as soon as the aircon cuts out. \$\endgroup\$ – BG100 Jan 27 '11 at 22:04
  • 1
    \$\begingroup\$ I have a room temp sensor. I am looking for something to specifically monitor the airflow because it would give me more time and allow me to page additional people / groups who deal specifically with the air conditioner. \$\endgroup\$ – mccoma Jan 30 '11 at 18:55
0
\$\begingroup\$

There is a cheap alternative that you can put in every point you like. It's a USB Temperature Sensor from Magiant.it that sends email and/or SMS when temperature is above the thresold and could also shutdown the pc/server.

\$\endgroup\$
0
\$\begingroup\$

Omega sells a hot-wire anenometer w/ the appropriate range for under $700. It's small. It is a logger, but it has a serial interface and PC-side programs available (for logging, but if its a simple serial interface, you can write your own supervisory program on whatever platform you want to.

You could, of course, build your own hot-wire unit, or use a thermistor in self heating mode. I find this tricky.

Alternatively, you might consider measuring the current drawn by your AC unit.

I really like the flex sensor idea, and you could possibly extend the concept to include any type of moving vane anenometer -- so long as you can measure the angle of deflection, you're good to go (but you'll be running fairly uncalibrated).

\$\endgroup\$
0
\$\begingroup\$

I'm suprised noone has mentioned a pitot tube, which is used to calculate the inertial contribution of a moving column of air. Simply put, a pitot tube is arranged so one pressure sensor faces into the wind, and one pressure sensor faces perpendicular to the wind.

That said, I'm not sure the numbers will work out for reasonable sensitivity. Someone needs to do the math.

Another method, used to measure pulmonary function, is measuring the pressure on two sides of a mesh screen, which creates a slight resistance to flow. Ohms law tells us that (P2-P1)=(flow)*(resistance to flow).

Of course, the latter involves more of a mechanical change to the airflow system.

\$\endgroup\$
0
\$\begingroup\$

The military units I worked with had huge cooling fans built into each rack. In the airflow there was a vane on a switch placed perpendicular to the airflow. If air was flowing, the wind blew the vane some which actuated the switch.

An easier solution, in my opinion, would be to replace the switch with a low-torque potentiometer and monitor the resistance via a voltage divider. You'll be easily able to determine whether or not the A/C unit is blowing with this setup...

\$\endgroup\$

protected by Olin Lathrop Jan 29 '13 at 13:31

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.