Automotive radiator cooling fan contoller

[KeepItSimpleStupid]

I will be putting in what I have learned are called "bypass capacitors". One for each opamp. If memory serves me correctly (my head is swimming in lots of new tidbits of information), they will be on the inputs to ground and each about 0.01mfd. I'm still not sure how/what they do exactly but I understand it is specifically to prevent oscillation.

Bypass caps are caps that are placed on the power supply rails to ground close to the IC.

Bypassing is an art and sometimes multiple types of capacitors are used in parallel. The first choice is usually ceramic.

The automobile environment is NASTY. For an interesting read, see: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CDoQFjAA&url=https://www.littelfuse.com/~/media/Electronics_Technical/Application_Notes/Varistors/Littelfuse_Suppression_of_Transients_in_an_Automotive_Environment_Application_Note.pdf&ei=cZ6LUqXoHJazsAS5_4HoAg&usg=AFQjCNFtLnWTgd2nP7kFajjBuwe-dCcHBQ&bvm=bv.56643336,d.cWc

At a minimum, I would suggest an 18 V TVS diode (Unidirectional (reverse biased) or Bidirectional) and a 400 PRV 1n540x series diode reversed biased across the power in 12 VDC.

 

[MrAl]

Hi,

Thanks for posting that link, very informative i think.

All:
It's very hard to overstate the importance of careful design for something that is going to have to work in the automobile. The RF noise is quite extreme as well, and of course noise on the 12v power line.
The idea is to think of EVERY node in the circuit as if it were a small antenna capable of picking up RF energy spikes from the ignition wires, which with the spark plugs actually makes up a spark gap transmitter. Using that same technology under more controlled circumstances, it would be possible to transmit information wirelessly. Thus that makes for a noisy environment just because of that alone.
To mitigate the effects of the constant transmission, it's good practice to keep all impedances in the circuit as low as possible. That's one of the reasons i suggested resistor values of 100k or less, and the lower the better. Of course they cant be too low or else the other components will not be able to function properly, so the idea is to go as low as possible without hampering the other operations of the circuit. For example, we'd like to tack a 100 ohm resistor to the output of a comparator for a pullup, but that's too low because the comparator output sink current limit may be only 10ma. With 10ma max and a 12v system, the pullup lowest possible value is therefore 1200 ohms, but that's with the car engine not running. With the engine running the voltage can get up to 14v, so the lowest possible value is 1400 ohms, and 1.5k is probably ok too as long as there is nothing else doing any pullup function on the output of the comparator.

What i could do if you (tryebyter) prefer is show a circuit that is known to work in the exact fashion you are looking for, and then we could talk about it. At least then you would have a reference design to go by so you know exactly what to shoot for. The circuit i am proposing is not for an automotive application however, but the topology is similar.

 

[MrAl]

Hello again,

Here are some simplified simulations of the fan controller. The waveforms shown in the diagrams show the variation in temperature of the water at the sensor.
The temperature appears to the left of the graphs, but it is scaled by a factor of about 20, so multiply those values by 20. So 9.0 comes out to 180 degrees.

There are four diagrams, showing:
1. No Delay, No Hysteresis
2. Long Delay
3. Short Delay
4. Controlled Hysteresis

#1 is to show what a linear system or fast PWM system would do.
#2 shows what we see with a single comparator and a long delay between radiator and sensor unit
#3 shows what we see what a single comparator and a shorter delay
#4 shows what we see when we use one or two comparators with built in hysteresis that we can adjust

#1 Controls the temperature pretty well, but that's probable impossible, so we have to use one of the others.
#2 Shows the temperature going up and down quite wildly, but the important point is coming.
#3 Shows the temperature going up and down much less than #2, but the important point about these two (2 and 3) is that the upper and lower temperatures are NOT controlled by us, they are controlled only by the delay, and we have no control over the delay in a real automotive system.
#4 Shows the temperature going up and down quite a bit too, but here it is almost entirely controlled by us in how we adjust the upper and lower setting.

Note that #2 and #4 look almost the same, but the difference is we dont know what the delay will be so we can not really predict that waveform in #2, it might be turn out that it is really looks more like #3.
Note also that the delay varies according to engine speed, so the upper and lower temperature will vary with engine speed. With #4 however the upper and lower temperatures will still be close to the same no matter what the engine speed is.

 

[tyrebyter]

MrAl,

You are 100 % correct in assuming I am learning as I go. My electronics deign/building experience has so far been very limited. Years ago, I desigend and built a small adjustable power supply with a bunch of diodes/voltage multiplier, transformer and adjustable voltage regulator, a potentiometer and a
few other components. I was putting myself through a self-taught "Radio Shack - School of Electronics" whereas I bought several of the paperback books written by Forest M. Mims III and a few dozen electronic components. I also bought several other books including "Encyclopedia of Electronic Circuits Volume 1" (GRAF) and some other electronic projects books.

I built several of the projects outlined in these books and then started designing some circuits of my own. I have been able to make some repairs on various pieces of electronic equipment and owe it largely to what I learned so far. I have trouble with transistors however. Back when I was going through the electronic books and trying to learn this stuff, I kept burning out transistors.

I never was able to succesfully design a circuit from scratch, using transistors. Something about the bias voltage, I think. I was able to use transistors and make curcuits by following schemeatics of projects in these books. Opamps never seemed to give me quite the same amount of trouble.

This project is for a real-life situation to be put into a car, if I can get it sorted out to my satisfaction. I will not be putting anything into the car until I feel it will provide 100% satisfactory service and in all real-life situations. I am not in a rush to finish the project but I need to get it done as fast as I am able to learn this new-to-me material.

I will definitely be putting this circuit through some bench-testing before it ever gets near the car. I own a Protek brand 20mHz CRT oscilloscope which hopefully will help me when the time comes to start soldering. I've used it enough to set the duty cycle on electronically controlled carburetor idle jets (square wave pwm) and when testing other automotive and electronic devices to know some of it's capabilities and shortcomings. It's certainly no Textronix and the 20mHz ceiling may make it useless for the PWM part of the circuit.

I WANT R8 there to deliberately MAKE that 1/4 of the LM324 to run in the linear range. The lo-speed will be driving a PWM to operate the fan at low speeds/low temperature, if I'm successful.

I removed R3 and R4 from the other 1/4 of the LM 324 which I WANT to MAKE run in the digital (on/off) range. It will be operating the fan hi-speed "on" and "off". It still works exactly the same, without the resistors ! Thanks for that tip.

KISS, Back to you later.

I'm having no luck getting my simulation software to provide good graphs for a square wave generator circuit. That is why I have not been on here for several days.

 

[alec_t]

Can you post the .asc file of the square-wave gen?

 

[MrAl]

Hi,

Are you using LT Spice now? That would be good because we can all then follow your work piece by piece.

Your 20MHz scope should be more than enough for this relatively low bandwidth project. We could also give you suggestions for testing various things in the circuit too if you like such as comparator output oscillation.

 

[tyrebyter]

alec-t,

I am still not able to do much with LT Spice yet. I am attempting to learn how to copy library models from the Yahoo group so I can use them. In particular the LM324/NS opamp. Apparently, I am not the only one having trouble with this. Maybe I need to find suitable Linear Technology supplied replacements ?

For the record I have LTspice IV version 4.20b which I downloaded and istalled Nov. 08 2013. As a test, I downloaded a 2-transistor square wave generator .asc file. I cannot get the simultation to work properly with LTspice. It goes runs a sequence of events that show on the bottom of the window but does not show me any curve when I run the simulation. All I end up after a short wait, is a black screen with a time scale/axis along the bottom. Obviously I need to figure out how to palce a Y-axis for voltage.

MrAl,

I'm working on it. Please be patient. I have only a certain amount of time every day to devote to this project and I guess I am a slow learner - especially with LTspice apparently. I'll be looking at the linear Technology website next for some tutoring.

 

[MrAl]

Hi,

With LT Spice sometimes you have to click on a node in the circuit to get a waveform to show up in the black screen with the axises. Try that next.
I'll try not to rush you.

 

[alec_t]

If you have a black window, LTS isn't psychic so can't know what to plot in it until you click something in the schematic. Clicking on a component will plot the current through that component (note the change in the cursor as you move it over the component). Clicking on a link between components will plot the voltage on that link.
Press F1 if you want help on how to use LTS. The help is pretty comprehensive.

 

[tyrebyter]

GOOD NEWS !

Thanks MrAL and alec-t for pointing out how to plot graphs. I am finally getting something besides frustration after opening an .asc file with LTspice. I can now look at a square wave output after running a simulation with LT !!! Thanks to messages I read in the Yahoo user's group archives, I even managed to open a file that includes the elusive model of an LM324/NS op-amp which helps in my still-a-little-shaky understanding of how to import third party models.

Please bear with me as it might still be a couple of days before I can post my latest schematic as an .asc file.

Switching from the simulation software I've been using to LTspice is looking like it will be easier than I thought it would be, at first. The two software packages use opposite philosophies. LTspice wants you to name the "whatever" first and then apply the verb - as explained near the beginning of the help file. Witht he Simetrix simulation software I had gotten used to, I was selecting what action I would take and then naming the item to take action on. I didn't quite grasp the entire extent of this philosophical difference and you can imagine it had profound effects on my learning curve.

More catching up when I come back on here to post an .asc file.

 

[tyrebyter]

I'm nearing completion of the .asc file representing the circuit I have, so far. The circuit that seems to work with the other simulation software. I'm still bogged down in utilizing non-LT componenets. I found what to do to get the LM324 op-amp but am now grappling with the BC107 npn bjt. I'll be back with another progress update when I have made some ... progress, that is.

 

[alec_t]

Personally I wouldn't be picky over which transistor model I used, provided it's somewhere in the ball-park. A sim is not real life; just a guide.

 

[MrAl]

Hi,

I have to agree that picking the transistor for this app probably isnt that critical. In some apps it is but for this one it probably wont matter too much. Just pick one with a similar CE voltage rating or higher and collector current rating or a little higher. Also watch the gain.

For that BC transistor it looks like a 2N2222 would do for now.

 

[tyrebyter]

It works ! In simulation anyway. This is what I have for now after working with LTspice. Probe the two outputs "PWM driver voltage" and "Hi-speed fan switch". R1 and R2 are my attempt at getting the two outputs somewhere near where I predict they will need to be.

I am beginning to get more of an idea why capacitors need to be included as the voltage hovers close to making things happen, sudden voltage changes/surges could create enough oscillation in the circuit to cause it to "stutter" the hi-speed relay - as KISS predicted. For now I am still going to try to get the PWM circuit included and will include that when I get there.

 

[tyrebyter]

I'd still like to have a better understanding of how to utilize non-LT components. I'll be looking into that more as I work on the PWM part of this project.

 

[alec_t]

I'd still like to have a better understanding of how to utilize non-LT components.

You seem to have mastered the basics, including making use of the 'include' command .
I realise you're still tweaking things, but why not combine R2,R5; likewise R1,R6? And add a bit of positive feedback to U1?

 

[MrAl]

Hello again,

Well i am happy to hear you got the simulation working. That can tell you a lot about how this works.

Since you got that far then you may want to experiment with a little hysteresis in one or both comparators. The attached drawing shows the connection for one comparator and the other one would be done the same way.
The resistor RS can be maybe 1k, and RF can be say 100k. Since 1k is 1 percent of 100k, the percent hysteresis is about 1 percent. That would mean that if the sensor voltage at 200 degrees was 2 volts and the voltage at 300 degrees was 3 volts (perfectly scaled) then the percent temperature hysteresis would be 1 percent, so roughly 200 degrees to 202 degrees, or maybe 199 degrees to 201 degrees. But it's usually not that simple so 1 percent hysteresis does not translate to exactly 1 percent in temperature but could be more or less depending on the sensor. For example, if the voltage is 5v for 200 degrees and 6v for 210 degrees, 1 percent of the difference voltage (6v minus 5v is 1 v difference) is 0.01 volt, and the difference temperature is 210 minus 200 equals 10 degrees, and one percent of that is 0.1 degree, so you see in that case it would be much less then 1 percent temperature wise. If 5v was 200 degrees and 6v was 600 degrees, then the difference temperature is 400 degrees and 1 percent of that is 4 degrees.
That's not exactly how it works but that gives you an idea of how it does work, and some experimentation is a good idea just to see what you actually end up with, unless of course you want to calculate everything more exactly. To do that we would need to see the full curve of the sensor vs temperature, or at least three relevant measurement points.

In the drawing you see RS is simply in series with the non inverting terminal, and RF is the feedback from output to non inverting terminal.

 

[tyrebyter]

You seem to have mastered the basics, including making use of the 'include' command .
I realise you're still tweaking things, but why not combine R2,R5; likewise R1,R6? And add a bit of positive feedback to U1?

Alec-t, Thanks for hanging in there with me. R5 and R6 comprise a 20k potentiometer. I could not find any symbol or Spice model for a variable resistor/pot.

Meanwhile I'm wading itno PWM starting with square wave generators. I have sucessfully simulated several circuits to produce square waves with the 324 op-amp. The 40kHz target seems too high for these. According to LTspice the wave turns into a triangle nearing 10kHz and above. I'm assuming the slew rate of the 324 is not high enough at the full 10 Volts which I'm running in the simulation.

I have an incomplete table somebody posted on the web of resistance/temperature for a similar sending unit. I have been working under the assumption that it is at least in the ballpark. It lists 158F@170Ohms, 176F@142Ohms, 194F@106Ohms, 212F@79Ohms.

If you look at the Simetrix graph I posted a while back it shows the red curved line as ohms with respect to voltage after running a DC sweep of the sending unit resistance. It seems that LTspice does not allow for sweeping a resistor value. I have not been able to figure out how to, if it does.

If you go back and look at #48 in this thread you'll see the red curve I just mentioned. R2 in that schematic is the sending unit and the gauge is R1. The two (R1 and R2) make a voltage divider that provides a varying voltage with respect to the changing coolant temperature. I arbitrarily chose 0-1k Ohms to sweep the resistor value figuring it would cover the range I am interested in.

 

[tyrebyter]

Hello again,

The attached drawing shows the connection for one comparator and the other one would be done the same way.

MrAl,

For some reason, I don't see the drawing.

 

[alec_t]

I could not find any symbol or Spice model for a variable resistor/pot.

The Yahoo group has one.
You're right about the LM324. That, and many other opamps, are too slow for PWM at, say, 20kHz (which you may need if you don't want to hear your fan whistling).