Ice Detector Design

[ericgibbs]

It appears that both comparators need to trip at the exact same point for proper operation of the circuit. But, in practice, they will not due to differences in their offset voltage.

To resolve this you could use U2 as just an inverter by connecting the negative input of U2 to the output of U3 and adding a resistive divider to set the threshold of the positive input of U2 at about 0.3V.

Or am I misunderstanding the circuit operation?

hi carl,
The circuit is designed to give the user a wide range and easy setting of the High and Low thresholds.

IIRC you use LTspice, attached the asc file, try it out.

BTW: its a useful circuit for auto battery chargers.

 

[crutschow]

hi carl,
The circuit is designed to give the user a wide range and easy setting of the High and Low thresholds.

IIRC you use LTspice, attached the asc file, try it out.

BTW: its a useful circuit for auto battery chargers.

I wasn't questioning the operation of the circuit in simulation. That will work fine since the offset of both the comparators is the same. But with an actual device the two comparators will likely have slightly different offsets which means they will switch at slight different points. Thus if U3 switches before U2 there will be no hysteresis for U3 and it could oscillate.

That was my concern. That's why I suggested the circuit change to maintain the hysteresis and ease of threshold adjustment that your circuit has without this possible oscillation problem.

 

[Boncuk]

Hi ADWSystems,

here is another approach to measure icing on a heat exchanger.

Icing may occur before temperature is actually down to <0C and depends on atmospheric conditions. Ice might even build up at +5C with the right amount of humidity and airspeed through the heat exchanger.

I developed an "anti-icing" system for a car manufacturing plant (Audi, Ingolstadt/Germany).

The principle is very easy to understand: Intake air- and exhaust air pressure are almost equal (minus the resistance of the heat exchanger).

I used a differential pressure sensor with both (negative and positive pressure ports connected to PU hoses and fixed into position) and placed the measuring points directly at the intake and exhaust end of the heat exchanger.

During normal operation (no icing) the pressures are almost equal. As soon as ice builds up the pressure difference increases. The output voltage of a differential pressure sensor with amplifier is in the Volt-range (no necessity to deal with milli-Volts and amplification factors).

Using a 10mb differential pressure sensor you can set the amplification of the INA11 (instrumentation amplifier) for an output of 5V at half the span of the sensor. Sensor used is a Motorola MPX2010DP.

If icing is sensed you should disable cooling and use room air to melt the ice. If normal conditions are regained cooling could automatically be re-initiated.

Here are the schematic and PCB layout of a differential pressure sensor. The positive port must be connected to the pressure side of the heat exchanger while the negative port must be connected to the outlet.

Board size is 53.6575X47.4662mm (2.1125X1.8750").

The circuit is relatively cheap to build.

For comparison: A professional differential pressure transducer (Furness Controls - UK) is about €1,2000.

Boncuk

 

[ericgibbs]

I wasn't questioning the operation of the circuit in simulation. That will work fine since the offset of both the comparators is the same. But with an actual device the two comparators will likely have slightly different offsets which means they will switch at slight different points. Thus if U3 switches before U2 there will be no hysteresis for U3 and it could oscillate.

That was my concern. That's why I suggested the circuit change to maintain the hysteresis and ease of threshold adjustment that your circuit has without this possible oscillation problem.

The circuit as previously built showed no sign of instability, I believe that you are looking for problems that do not exist.

All the User has to do is SET the Low threshold and then SET the High threshold, while setting the High threshold, the Low remains at its set value.

If you would care to post your version of a working solution, I would be pleased to see it.?

 

[ADWSystems]

Thank you for the post, Boncuk. I actually got the idea to detect the ice from an HVAC air exchanger at a facility I worked at. Unfortunately the dehumidifier is not sealed well enough to permit differential pressure measurements. Additionally, there is dehumdifier is all on or all off. I can't turn off the cooling alone. As it is, I had to install an extra fan to help warm up and melt the ice when it does currently shut off. The extra fan actually helps the air flow now that I have installed a filter. I think the head of the LM34 placed x mm from the cooling coil will be sufficient. When the ice builds out that far the LM34 will detect the temperature drop.

 

[ADWSystems]

What's wrong with the simple hysteresis setup diagrammed in the LM393 app note? I don't think I need something that elaborate. I still need to boost the output signal, 10mV/deg F is not enough room to stick all this. It's been a long time since I've built a single transistor amp. I know you need to stay in the linear range but I can't remember where that is. Using a 2N2222A as an example, where is the linear range of a single transistor? The 2N2222A has a gain of 50 but I only need 5, how do I turn down the gain?

 

[ericgibbs]

What's wrong with the simple hysteresis setup diagrammed in the LM393 app note? I don't think I need something that elaborate. I still need to boost the output signal, 10mV/deg F is not enough room to stick all this. It's been a long time since I've built a single transistor amp. I know you need to stay in the linear range but I can't remember where that is. Using a 2N2222A as an example, where is the linear range of a single transistor? The 2N2222A has a gain of 50 but I only need 5, how do I turn down the gain?

hi ADW,
Not wishing to sound picky.
But you did ask for suggestions for you project, yet you are dismissing them out of hand, because you consider them to elaborate.??

A 10mV signal is easily detectable.
Why would you want to build a transistor amp when simple OPA's or Comparators are so cheap and simple.?
A 2N2222 transistor has a gain of approx 200, not 50.

If you think the d/s LM393 comparator circuit will do the job , why dont you build it and see.

 

[ADWSystems]

I'm not dismissing anything. I'm taking all suggestions and evaluating them based on, but not limited to, parts at hand and equipment available. I've recently lost access to a o-scope so I need a design that can be debugged with a voltmeter (or two). My original question was actually a feasibility question on the mechanics of the design, not on the circuit to do it. But, again, I welcome the discussion and the suggestions. I've learned a lot of the "other ways to skin the cat" form the memebers of the board here.

I was thinking because I am operating at, essentially, DC I don't have any frequency or AC aspects to worry about. A simple transistor amp may be easiest. And I have them on hand. I don't have any rail-to-rail opamps on hand. Plus I'm trying to fresh my memory on transistor amps.

I'm fairly sure the single LM393 circuit will work but I'm worried about setting the hystersis in the mV range with a voltmeter, if I don't pre-amp the temperature sensor output.

P.S. what does d/s stand for?

 

[ericgibbs]

hi,
Ok, no problem.

d/s is datasheet.

Lets assume that you amplify the 80mV range change to say a 5V range, how are you going to use the amplified signal.?

 

[alec_t]

If you want greater sensitivity you could use an NTC thermistor instead of the LM34. For example, a typical 10k (at 25C) thermistor has a resistance which varies (albeit non-linearly) from ~ 33k to ~20k over a temperature range of 0C to 10C. [Figures taken from **broken link removed** ]

 

[crutschow]

The circuit as previously built showed no sign of instability, I believe that you are looking for problems that do not exist.

All the User has to do is SET the Low threshold and then SET the High threshold, while setting the High threshold, the Low remains at its set value.

If you would care to post your version of a working solution, I would be pleased to see it.?

As an engineer I tend to look for subtle problems (sometimes perhaps when they do not exist).

But I think my concern is valid, even if you did build one circuit that worked without problem. A different chip with a wider offset between comparators could, I believe, exhibit oscillations.

Anyway, attached is your circuit with my modification. I didn't have the model for the LM393 so I used the LM339.


View attachment LM339WideHystIce3.asc

 

[alec_t]

There is a single-comparator option which provides a just-as-easy setting feature and wide hysteresis. Simply replace the positive feedback resistor with a 1 Meg trimmer. U4 is then used to set the lower threshold and the 1 Meg trimmer sets the hysteresis amount (and hence indirectly the upper threshold). Because the trimmer value is much greater than R12 it has little effect on the lower threshold setting.

 

[ericgibbs]

hi Carl,
Thanks for posting your version of the circuit, if you are more comfortable with your version I do not have a problem.

Whatever the OP decides to do, [ if he uses either version] I hope he will post some feedback on his results.

 

[ADWSystems]

I'm going to use the amplified signal as the input to the comparator. At the given values the critical points will be 320, 370, and 650mV. the difference between freezing and room isn't a big deal to tune with a meter (~300mV). But I'm thinking trying to tune the difference between 320 and 370 might be a bit touchy. I'm thinking a 5x gain would give 1.6, 1.85, and 3.25V and be easier to tune.

I guess you could say I would like something more sensitive. You could also say I'd like something with less range.


-------------------------------------------------------------------------------------------------------------------------
Cole's law:
Thinly sliced cabbage.

 

[ericgibbs]

I'm going to use the amplified signal as the input to the comparator. At the given values the critical points will be 320, 370, and 650mV. the difference between freezing and room isn't a big deal to tune with a meter (~300mV). But I'm thinking trying to tune the difference between 320 and 370 might be a bit touchy. I'm thinking a 5x gain would give 1.6, 1.85, and 3.25V and be easier to tune.

I guess you could say I would like something more sensitive. You could also say I'd like something with less range.


-------------------------------------------------------------------------------------------------------------------------
Cole's law:
Thinly sliced cabbage.

It is possible to use half of a LM393 as an amplifier and the other half as a comparator. [the LM393 is a dual device, in a 8pin DIP]

 

[crutschow]

It is possible to use half of a LM393 as an amplifier and the other half as a comparator. [the LM393 is a dual device, in a 8pin DIP]

A comparator needs compensation to avoid oscillating when used as a linear amplifier with feedback. Such a circuit is shown at the top left of page 6 here. The 15k resistor and 0.5µF cap on the LM393 output provide the required compensation.

 

[ericgibbs]

A comparator needs compensation to avoid oscillating when used as a linear amplifier with feedback. Such a circuit is shown at the top left of page 6 here. The 15k resistor and 0.5µF cap on the LM393 output provide the required compensation.

hi Carl,
I am aware of that circuit and have used it in the past, thats why I suggested something like it to the OP.

EDIT:
This is an example for a previous project.

 

[ADWSystems]

Cool solution. But question (as I have not tried to use a comparator as an opamp or opamp as a comparator): Would it be easier to configure an opamp as a comparator? I ask because I would be afraid to attempt to build such a circuit with a high risk of oscillating and without the aid of an oscilloscope.

 

[ericgibbs]

Cool solution. But question (as I have not tried to use a comparator as an opamp or opamp as a comparator): Would it be easier to configure an opamp as a comparator? I ask because I would be afraid to attempt to build such a circuit with a high risk of oscillating and without the aid of an oscilloscope.

Who has said a compensated LM393 OPA/Comp combination would oscillate .??

When you make up your mind which solution you would like to use, let us know.

 

[ADWSystems]

A comparator needs compensation to avoid oscillating when used as a linear amplifier with feedback. Such a circuit is shown at the top left of page 6 here. The 15k resistor and 0.5µF cap on the LM393 output provide the required compensation.

He said "to avoid oscillating". I read that as "it is probable, that if not done correctly, it will oscillate". If it does or is oscillating, I will have no way of knowing.

I will decide on a circuit when you guys stop posting new/other options.

Any comments on the lastest point at hand, as how to configure an opamp as a comparator? Would it be less likely to oscillate?

In looking for a rail-to-rail op amp, I found the LM358. The input common mode range is specified as 0 to V+-1.5 with V-=0 and V+=30.
1) I would not think that the specification of the -1.5V is going to vary much with V+. Am I correct on this?

2) As previously stated my input range is going to be 320mV to about 750mV, at 5x Gain the output would be 1.6V to 3.750V. Looking at the output voltage range, at V+=5V, I may have a problem reaching the 3.75V, correct?