Automotive radiator cooling fan contoller

[tyrebyter]

For those of you who are interested, I added some new bio info to my profile.

 

[MrAl]

Hi again,

Oh that makes sense now.

A latch is a very very simple device. First of all it's a logic device so it gets inputs of either 0v or +10v (for example) and outputs the same. For this basic latch there are two inputs, "S" and "R", and one output "Q". The S input is the "Set" input, and the R input is the "Reset" input. The Q output goes to either 0v or +10v depending on what inputs were sent to S and R in the past. This is sometimes called a "SR Latch", or "RS latch" (simply because of the two inputs).

So if we put +10v on S and 0v on R, the output Q goes to +10v and stays there.
But if we put +10v on R and 0v on S, the output Q goes to 0v and stays there.

Note the output Q 'stays there' which simply means it does not change again even if the +10v input goes back to 0v.
So if we hold R input low (0v) and put a high on S (+10v) the output Q goes high, and then if we make S go low (0v) the output Q stays high.
If we hold S input low and put a high on R, the output Q goes low and then if we make R go low again the output Q stays low.

So the output Q only changes when we make one or the other input go high. It never changes when we make one go low.

There is one disallowed input state and that is to make BOTH inputs S and R high at the same time. The output may be unpredictable so we avoid that input state. It may be defined for some physical devices, but in theory we dont want to do that and we dont have to do that for this app anyway.

We can get around having to use a latch by making our own latch out of an unused section of the comparator such as the LM339. It's a little tricky but not too complicated. Actual latches that work at levels of +14v are available in CMOS technology.

 

[KeepItSimpleStupid]

I believe KISS is mentioning 40khz as a good pulse frequency for an electric motor PWM controller. Under 20khz (upper limit of human hearing) apparently causes the motor to make an annoying sound.

That was the suggested frequency for pancake motors for PWM of pancake motors. It was chosen based on magnetics.

 

[alec_t]

add a transistor which would be powered from 12 V and can adequately deliver enough current to energize a relay at Q1-out, is this how the circuit would look ?

A sensitive relay could be driven directly by the comparator, whereas a transistor stage such as Q1 would be needed for a beefier one. What will you use the relay for?

 

[3v0]

I don't see how those targets can be met. Maintaining temperature within 0.1 degree,

Why would you need to maintain the temp within .1 degree? I expect the engine would be quite happy with 10 degrees.

 

[MrAl]

Hello again,

I assumed a MOSFET would be the main drive device eventually, but whatever he wants to use i guess. Alternately bipolars are very reliable devices.

As to the accuracy being 0.1 degree, it's hard to say if that is possible or not because there are so many factors that can vary and we dont have a true model of the system. It depends on the delays in the system and what we are willing to measure. If there is a delay in the system that is large enough (and most likely there is) then we'd have to measure more quantities in more places. That is how a super accurate system is designed because it is not possible to predict certain outcomes without measuring more variables in the system, and because of delays it is necessary to activate the control based on what the output WILL be in the future, not always what it IS in the present. That's because disturbances come in many forms for a system like this including how hard the human presses on the gas pedal (engine power means more internal heating), how fast the car is moving (natural air flow means more cooling), ambient temperature, etc. It would be nuts to measure all those things. So some variation has to be accepted, and it generally works ok for everything except nuclear reactors.

 

[tyrebyter]

It suddenly occured to me that I never supplied some very pertinant information. Perhaps this schematic of what is already in place (except for the solid state voltage regulator) will be useful. The stock bi-metal voltage "stablizer" that the car has now, provides an average of 10 volts to the voltage divider comprised of the temperature gauge and the temperature sending unit.



I am not trying to control the engine temperature with this fan. The engine's thermostat does that extremely well as long as the car is moving over 5-10 MPH. The cooling system has been modified. It has been tested and can keep the engine at normal operating temperature (180F) on a 100 degree day at or near top speed for over an hour.

It's only when the car is stopped, that I am concerned with.

I want to be able to set the temperature when the electric radiator fan turns on by turning a knob/potentiometer. Ultimately I'd like to be able to adjust when the fan turns off with another knob/potentiometer to within as little as little as 0.1F degree lower than what the turn-on temperature is set to.

 

[tyrebyter]

I tried putting the schematics Mr Al and Alec T posted but I cannot get my simulator to run on them. I think I'm not using the proper voltage sources for each one.

Mr Al, I plan on incorporating MOSFET(s) for driving whtever I end up using on the output side - one or two windings or relay or PWM or some combination thereof. I could not find a schematic of a MOSFET switch. I did find a bipolar NPN switch schematic which is what I used. I am "borrowing" pieces of schematics I find online and putting them together to make what I am posting.

I'm still trying to wrap my head around "latching".

 

[tyrebyter]

Alec T, Depending on which scheme I finally end up using, I may have a relay to turn on the hi-speed function. If I use PWM across the board, I will not use a relay.

 

[tyrebyter]

3v0,

This engine has a barely streetable, radical camshaft profile and of course it displays the characterisic "lumpy" idle. In fact I feel lucky/very experienced/good-at-it to have been able to get the engine to idle at the stock speed of 750 RPM, +- 50 RPM. The primitive but very large (SU) carburetors barely have enough airflow through them to draw fuel at idle and the idle air/fuel mixture setting is VERY finicky. Actually 10 degrees does upset this particular engine's idle characteristics, considerably !!! Thus my quest for better fan control when the car is stopped.

 

[KeepItSimpleStupid]

Aside: Have you ever used the "Propane Enrichment" method to set the idle. It's super simple and works every time. Not sure how to adapt it to multi-carbs though.

 

[tyrebyter]

KISS, No, I have never used the propane injection method of idle adjustment on any Triumph. British Leyland never published figures regarding target RPM when propane is admitted into the airsteam entering the carburetor.

In case you are wondering how I do it and I realize this is not really pertaining to this thread. My apologies to everyone else who is keeping to the matter at hand. I use the "idle drop method" whereas you lift the air piston 1/32" and a piece of heater tubing to listen to the pitch of the air hissing past the throttle plate when tuning SU and Stromberg carburetors.

Many years ago, I won a fair chunk of change one day when a co-worker bet me it was impossible to set the mixtures properly nor within specification accoridng to the underhood sticker, on multiple carbs by listening to them. His claim was that an HC/CO sniffer was necessary to achieve the precision mixture adustment that the smogged cars we were working during that time, required. I told him not only could I do it but I upped the bet saying I would also have less HC in the exhaust as him. I was able to get the engine exhaust wihtin spec and to have lower HC than he did, and I won the bet. Unburned hydrocarbons were the major bug-a-boo in the semi- early days of smog controls circa 1978.

Besides that, this engine now sports THREE - count them - THREE SU carburetors each with with a 2" bore/throttle plates. The stock Stromberg carburetors only numbered as one pair. Both of them had only a 1 5/8" bore/throttle plates. I've never heard of anyone using the propane injection method of setting the mixtures on multiple carburetor setups. Maybe I missed something in the 45 years I've worked on Triumphs and MGs. Yes, since before I could legally drive, I was working on Triumphs and MGs.

Shouldn't this question be on another thread or another forum altogether, KISS ? Please understand, this engine idles PERFECTLY even when the radiator fan turns on. What I want is an electronic controller circuit that will make it shut off MUCH sooner than it does now and one that I can adjust turn-on and turn-off time if they are not going to be the same.

Thanks for asking though. Maybe it would be better if any other specific automotive suggestions - not electronic in nature - were sent via IM or email ?

 

[3v0]

3v0,

This engine has a barely streetable, radical camshaft profile and of course it displays the characterisic "lumpy" idle. In fact I feel lucky/very experienced/good-at-it to have been able to get the engine to idle at the stock speed of 750 RPM, +- 50 RPM. The primitive but very large (SU) carburetors barely have enough airflow through them to draw fuel at idle and the idle air/fuel mixture setting is VERY finicky. Actually 10 degrees does upset this particular engine's idle characteristics, considerably !!! Thus my quest for better fan control when the car is stopped.

Have you tried rigging up some cold air intakes for the carbs?

 

[tyrebyter]

3v0,

I appreciate your helpful suggestions but I really don't see how a cold air intake will help me control when the electric radiator fan turns on and off.

I also don't see how a cold air intake will do anything at idle speeds when the car is stationary and airflow entering the carburetors is down to about 0.01 CFM or less. By the time air goes through any tube or duct, when it finally gets to the carbs, it won't be cold anymore. ;>)

Maybe you missed the post where I mentioned how the engine idles so I will reiterate. The engine idles 100% PERFECTLY, albeit with a noticible lope due to the camshaft profile. The engine develops more than double what a stock TR-6 engine is capable of. It took, many, many modifications to do so. This kind of power increase always brings a certain trade-off. In this case it is a finicky idle, which after much work, I was able to tame. As long as the engine temperature reamains constant within a very small range, the engine will idle for an indefinite amount of time. Presumably until it runs out of gasoline. Even when the car is stopped and the radiator fan is running. It's not a race car. It's driven on the street, regularly.

It has NO carburation issues other then the fact that they are a very primitive design and have absolutely no automatic temperature compensation ability. Add to this that the engine is breathing VERY poorly at such a low idle RPM due to the extreme nature of the camshaft. There is such a small "depression" across the carburetor bridge pieces that communication between the combustion chamber and the mainjet orifice is almost nil.

If the previous paragraph has lost you 3v0, maybe you should think more about how to help me with an electronic fan controller and stop worrying what mechanical modification you can advise me to try ... beyond what I've done after decades of high performance engine building experience plus the extensive research and testing I did ten years ago with this engine. ten years ago, back when aftermarket radiator cooling fans and fan contollers were not as popular as they are now and not in such a diverse assortment of offerings.

My only issue with the car - and the reason I am on this forum - is that the electric fan runs too long when the car is stationary. The aftermarket has better fans now but still there is no controller I can buy, that satisfies the needs of this particular TR6.

I thank you 3v0, for your concern that I may have overlooked something as far as the engine and carbs goes, but what I need is help with designing the circuit to control the fan.

 

[3v0]

Good luck.

 

[tyrebyter]

Thank you 3v0. I have learned to not rely on luck but I obviously need it.

MrAl, I believe I understand enough about the nature of digital latching components, now. At least enough to see why it's necessary to use a latch in order to keep the fan running as the temperature drops down below the turn-on point.

After considerable thought, I have changed my mind from my earlier postings and now I believe having an adjustable turn-off that is different than the turn-on may not be needed. I am planning to use a circuit without the independant turn-off feature to see if there are drawbacks.

In other words the fan will turn on at an adjustable setting, say 181 F. The coolant flowing past the sending unit will still be above 181 F, while the fan is running, making the radiator cooler. Once the radiator coolant enters the engine, and the engine cools enough, the coolant flowing past the sending unit will eventually get below 181 and the fan will turn off.

Is keeping the buffer comparator in an on state like this for the entire time the fan is running a bad idea ? Will using a latch in this scenario be necessary to prevent problems with the 339 oveheating/overworked ? Same for 358s if I use them instead ?

 

[alec_t]

I have changed my mind from my earlier postings and now I believe having an adjustable turn-off that is different than the turn-on may not be needed.

In that case a single comparator with an adjustable turn-on point and fixed hysteresis will do the job. Some hysteresis (electrical or thermal) is necessary whatever you decide to do, to prevent the fan chatttering on/off/on .....
The comparator with hysteresis is somewhat like a latch. Keeping it in one state or the other for prolonged periods won't bother it in the slightest.

 

[KeepItSimpleStupid]

The biggest enemies of OP amps/comparators are temperature and oscillations. Another concept with amplifiers is the Gain * BW or Gain-bandwidth product. This term is looked at as a constant. So, an OP amp with a BW of 1 MHz can only have a gain less than 1E6.
Another selection criteria sometimes is whether an amp is Unity Gain Stable. Sometimes, you have to select a different OP amp for stability.
So, what I'm saying is a high bandwidth part has more of a tendency to oscillate and therefore more attention to layout is important. Stray capacitance causes problems. If the system oscillates at a high frequency, that's a sure way to raise the temperature. The parts, so far, selected are not high frequency parts.

The latch does prevent, if I'm reading right, a short burst of high temperature from whatever, whether or not it's real or from noise on the system from causing the output device to switch on glitches.

So, as it was said earlier, the output device, whatever it is, is going to try to heat up. They have some inherent losses. MOSFETS essentially use voltage and very little current to turn them on. The result is low on resistance, typically milliohms or less. Bi-polar transistors always have an diode-drop to contend with and are turned on by current may be 20-400x less than the actual controlled current. There is a special topolgy caalled the Darlington Pair, that can increase the gain to the product of the gains of the individual transistors used.

 

[MrAl]

Hello again,

A digital delay will create a sort of built in hysteresis, but a purely linear delay is not usually a genuine delay unless it is specifically made to work that way. So whether or not the single set point model will work well depends on the nature of how flowing water transports heat from one place to the other.

The way i see it, the water itself acts as mostly an insulator because the convection current circulation time is much too long in a pipe of the length we would have to consider. So for transporting heat the water flow acts as a genuine delay. The actual delay period however may not be long enough to provide enough hysteresis. That's because it depends then on the flow rate, and i think the flow rate is pretty high in the radiator. Let me provide a quick rough example.

If the length of the pipe is 3 feet long and the diameter is such that the 3 foot long section can hold 1 gallon of water and the flow rate is 0.25 gallons per second, then the entire gallon of water inside the pipe gets replaced once every 4 seconds. That means if the water in one end of the pipe is heated it will take about 4 seconds for that heated water to reach the other end of the pipe. Likewise if the water is cooled it will take 4 seconds for the cooler water to reach the other end of the pipe.
It's a pseudo digital operation, so it's very close to a genuine delay which provides for real hysteresis. But what about the actual time period? It's quite short in this case, 4 seconds approximately, and that's in the ideal digital case.
So what happens if we stick a temperature sensor in the far end of the pipe and use that to turn a fan on and off to cool the close end of the pipe. Ideally, the sensor senses the temperature has risen, turns on the fan immediately, the fan cools the water, the cool water starts to flow, the cool water reaches the sensor after 4 seconds, the sensor senses the cool water and turns the fan off. So the fan turns off 4 seconds after it turned on.
Next, the sensor detects hotter water again once the cooler water makes it through the pipe. This could take longer because the water cooled in the radiator may be more than one gallon. If it is two gallons, then we can approximate that it has twice as long of a reserve of cool water, so it would take 12 seconds to reach the sensor.

So here we see the fan turn on, turn off after 4 seconds, then turn back on again after another 12 seconds. I dont think that is very good. On the other hand, if we allow a small temperature deviation between turn on and turn off (build in some electrical hysteresis) then we see a much longer turn on, turn off period. How much longer exactly would be the subject of an experiment, but it should allow for less turn on/turn off cycles over the operating time.

Note also the flow rate in an engine is probably much greater that the example. Perhaps you can find this information out too and that would help us to more accurately model the system.

 

[KeepItSimpleStupid]

Not sure what "brake power is": https://www.physicsforums.com/showthread.php?t=566125

Coolant Flow rate/Brake power graphs.