Maker Pro
Maker Pro

Airflow detector

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
Our hackerspace is adding some smarts to the laser cutters to prevent stupid and expensive mistakes.

Another guy is doing the logic (in hardware -- for resilience).

My job is the airflow sensor.

After discussing a few options (including hot wire sensors and the like, I have sorta settled on a pair of thermistors in thermally connected enclosures, one of which will be in the air path.

The circuit is as follows. (It's very conceptual at present -- I'm not really sure what voltages I have available).

Does anyone want to cast their beady eye over it and make comments?

The delay between activation/loss of air and reporting can easily be a couple of seconds, but I'm aiming on making it sub-second, but slow enough that mains interference and other noise will be filtered out as "high frequency".

I'll also let you read the summary I sent to our members:

As promised, while at the buck's night last night I put more thought into the airflow sensor. But not much...

This morning I discovered I have only 2 platinum resistance thermistors, so the thought of burning on out (to determine the max current carrying capacity) and then using two of them was not going to be an option.

Instead I pulled out some 470 ohm NTC thermistors and determined that at 50mA (where the resistance falls to around 140 ohms) they are very sensitive to any air movement (and surprisingly fast to respond). The power dissipated is around 350mW per thermistor at equilibrium (falling from 1W at cold startup).

One possible problem is that the air coming from the compressor will be at a temperature different from ambient. In some circumstances this might be a problem. One option is to mount each thermistor in the airflow, but having one given a highly restricted airflow. Another option is to mount both inside metal enclosures that are thermally connected, one open to the airflow, the other closed.

My original idea was to set these up as a voltage divider driven by an AC source to give a variable amplitude AC output. I'm now considering each thermistor having its own constant current source and comparing the two outputs to generate a signal.

In the proposed arrangement any common mode noise will be pretty well cancelled out, so shielded twisted pair cable (similar to audio cable) should be ample to reduce noise. Our interest in low frequency signals (determined principally by our requirement for a particular reaction time) means we can further attenuate signals above (say) 1Hz without any real degradation.

What are the voltages easily available inside the laser cutter? Is there a 12V rail?

Attached are three images.

LaserAir-1 is the circuit to generate a pair of filtered and buffered voltages, one from the reference thermistor, the other from the thermistor in the airflow.

LaserAir-2 is the error outputs. I can detect either input pulled too close to the upper or lower rail (failure of current source or thermistor) and the reference thermistor returning a voltage lower than the airflow sensor (probably misalignment or thermistor failure). These can be combined into a single active low output.

LaserAir-3 contains the circuit for reporting airflow.

It is likely that any or all of the conditions (error and/or air) could be reported on startup. If the controller has suitable capability, it should wait until a few seconds after the power is applied, then, if there is no error, turn the air on and off to validate the operation of the solenoids, availability of air, and the ability of the circuit to detect airflow.

I'm going to pass this conceptual circuit pass a few people for comments before I start thinking about component values. Knowing the power supply rails available and the required specs for the outputs (these can't drive more than a couple of mA) will be of great assistance.


And here are the conceptual schematics:

LaserAir-1.png

LaserAir-2.png

LaserAir-3.png

Any other thoughts? Perhaps power supply out of tolerance (e.g. power disconnected, low input voltage, etc?)
 

Alec_t

Jul 7, 2015
3,587
Joined
Jul 7, 2015
Messages
3,587
I'm sure that's workable, but I'm surprised you abandoned the original idea of using the two thermistors as a potential divider? That divider plus a trimpot could make a DC-energised bridge. A single opamp could detect bridge unbalance.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
One of the issues is that if I set them up as a divider, if i don't use a constant current source, the feedback effect of being one thermistor being cooled also cools the other. If I use a constant current source, each thermistor reaches equilibrium with about 7 volts across it. This is fine for a single thermistor, but two in series may not potent a constant current source (unless at a very low current where the desired heading would be minimal) to supply the regulated current from a 12v source.

as it is in my current circuit the resistors can have the same current, but remain isolated. In effect it's like having them on different arms of a bridge, comparing the voltage across the bridge.

the circuit here died only use a single comparator to detect airflow, the others are to detect fault conditions. I guess you could argue I'm using 2 op-amps tho buffer the voltage as opposed to one, but a stiff voltage reference would probably add to the package count. Also, two plasma in a package and 4 comparators in a package is hardly exotic.

But thanks for your comment, it has forced me to think again about the circuit. I will try to thermistors in series to see how they behave in practice.

If it's better than I think, I'll go that way. Also, I'm pretty sure I have some 100Ω terrified somewhere...
 

Alec_t

Jul 7, 2015
3,587
Joined
Jul 7, 2015
Messages
3,587
Point taken about the series connection.
So how about this bridge arrangement?
AirflowSensor.PNG
Being ratiometric it's insensitive to supply voltage variation (simulated here by a 1V sine on top of 12VDC), so no constant current source needed.
Current through each thermistor is ~45mA for the component values shown.
 

duke37

Jan 9, 2011
5,364
Joined
Jan 9, 2011
Messages
5,364
A simple bridge,supplied with DC, with two equal resistors in one leg and a thermistor and variable resistor in the other leg will balance when the resistor equals the thermistor resistance. This can feed a comparator as #4.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
Yeah, that circuit is pretty much what I'm using. It's operating in an electrically noisy environment (hence the filtering of the signal), and I require some hysteresis so there's no airflow setting where it switches on and off randomly.

Since I'm relying on self-heating of the NTC thermistors, a constant current drive has the advantage of removing any chance of thermal runaway. It also allows me to run three wires from the board to the sensor (with a simple reconfiguration that was omitted for clarity), and does not require 50mA (ish) running through a pot.

OK, that circuit would also allow me to run 3 wires from the sensor to the board, and a small change to have a pair of resistors and a lower value pot would minimise both dissipation in the pot and the chance of smoke release if anyone decided to wind the pot to either extent of its travel. I will also take a look at that. The time to reach equilibrium might be slower, but that may not be a huge issue. The circuit has less negative feedback, so will be more sensitive. I'll experiment with that too -- thanks.

The rest of the complexity is to allow for detection of failed sensors, sensors unplugged, sensors connected backwards(!), etc.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
I've tried out the pair of thermistors in series.

If I drive them from 12V the current is only 10mA and there is no significant response to blowing on one of them.

If I drive them from a current limited (50mA) 30V rail, the thermistors settle down to equilibrium (with about 13.5V across them) in maybe 5 or 6 seconds. They respond to air being blown on one of them, but it takes 3 or 4 seconds to get back to equilibrium.

I need to check the time to equilibrium for a single thermistor driven from a current limited 12V rail, but I'm sure it was less. If it is, the circuit with a thermistor and a resistor in series across 12V will also likely be slow to respond.

But thanks for the suggestions. It's really useful to try other options before I commit to anything.
 

Alec_t

Jul 7, 2015
3,587
Joined
Jul 7, 2015
Messages
3,587
Here's a mod which reduces the pot current to ~8mA and adds hysteresis to tolerate about 20mV of noise on the opamp inputs.
AirflowSensor2.PNG
I hadn't considered thermal runaway. If you used PTC instead of NTC thermistors you could avoid it.
 
Top