Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

March 28, 2024, 10:35:27 AM

Login with username, password and session length

Recent Posts

 

Fan control, indicating, proportional

Started by Roly, May 05, 2012, 02:44:21 PM

Previous topic - Next topic

Roly

Quote from: joecool85 on May 03, 2012, 11:52:55 AM
Quote from: Roly on May 02, 2012, 11:27:30 AM
Here's my indicating proportional "thermofan" design that has been used in several amps including a re-creation of an Acoustic 360 by J.C.Maillet;

http://www.ozvalveamps.org/techsite/thermofan/thermofan.htm

VERY cool!  If you don't mind, I'd love to have you start a thread here about this and include schematics, diagrams, pictures, whatever you can.

Ho-kay.


Design starts with a problem, and in this case the problem was a small home-brew PA that had been dropped on me to fix because it "has a couple of problems" and "doesn't work very well".  These would soon turn out to be major understatements.

The mixer portion needed a serious scrub up, mainly replacing 741's with TL071's, but the power supply and power amps were such a misbegotten mess that the only option was to totally rebuild them in a different case.  Since it was a "love job", to keep costs down the metal case from an old gutted VHS VCR was used.

The original unit didn't have a heatsink worthy of the name, about a square foot of 1/16" ali flashing screwed down to the wooden case floor, so again the workshop junk heap provided a length of pre-loved real heatsink.  But as the rebuild was coming together it became obvious that the heatsink would have to go inside the VCR case - somewhat less than ideal.

This in turn meant that the heatsink would need to be blown by a fan (again, dragged out of the junk pile), but the amount of noise it made led to a long ponder.  What was really required was some way of controlling the fan so that it provided cooling in proportion to the need.

A search of the InterWeb turned up any number of "bang-bang" or on-off fan controllers ranging from highly dubious to elegant overkill, but no proportional controls.

I eventually concluded that I would have to roll my own, and decided on using silicon as temperature sensors combined with a linear LM3914 "Dot/Bar Display Driver" (aka Line-Of-Light driver) as the heart of a proportional fan controller.  In bar mode the accumulating current drawn through the LED's could come via a fan, giving proportional current control.


The Controller




I chose this IC because its input side is floating and can be referenced over a wide range of zero and span settings; and because the LED driver outputs are programmable current sinks that can also be easily set anywhere between 2mA/step and 30mA/step.

This current mode output is an important property since it makes the circuit quite insensitive to supply voltage variation, and means the indicating LED's can be left out or doubled up (or more) in series, and that a second fan can be used in series on a supply up to 25 volts without any circuit changes.

The LM3941 has a stabilised reference voltage output of 1.250V available which is used to energise the zero point and span point settings, and in this case provide a stable bias voltage for the temperature sensor.

The zero and span are set by two twenty-turn trim pots in series, and while ordinary trim pots will work their setting is quite sensitive and tends to be a bit fiddly.  These are tweeked pretty arbitrarily so that only the first LED or two come on when idle at a moderate ambient temperature, and that the last LED comes on when the heatsink is getting too hot.

In this IC the reference output also doubles as a setting for the output current sinks which each sink ten times the current flowing out of the reference output.  Here the output current was set to about 20mA/step which relates to full fan current of 120mA a bit over half scale.

This current is set firstly by the value of the zero and span pots themselves - at 500 ohms each they present 1k ohm to the reference, drawing 1.25mA and therefore setting the output current to ten times this current; secondly the temperature sensor draws roughly another 1mA, so the overall setting is around 20-25mA/step.





The Sensor

One of the properties of a silicon diode is that it has a temperature co-efficient of about -20mV -2.1mV per degree C, that is the "on" voltage falls by 20mV -2.1mV for every degree C it gets hotter, and it's pretty linear from 0 to 100 degrees C.  Since the base-emitter junction of any transistor is also a diode, and since some transistors come in packages with handy screw holes, and since we don't need to know accurately what the temperature actually is, a specific temperature sensor IC isn't required.

I used a couple of BD139's because I wanted to monitor the temperature of both the power transformer and the heatsink, and they were to hand.  The mounting hole made them much easier to attach than ordinary power diodes (such as the EM410, 1N4001, etc) which are another option.  These two were connected in parallel so the hottest one rules, and there is no reason why this couldn't be extended to monitor several points if desired.

It is also possible to stack diodes or transistors in series to get a larger voltage swing with temperature if tighter control is needed, 2 = -4.2mV/C, 3 = -6.4mV/C, etc, but just for amp cooling applications I have only used a single diode drop and found that sufficient.

This voltage is input to the LM3914 on its "Sig In" line.  Because the collector and base of the temperature sensing transistor are held at the reference voltage (1.250V) the emitter voltage rises from 1.25-0.65 = 0.6 to 1.25-0.6 = 0.65 volts over the temperature span.  The Zero or RLO input is set for around 600mV and the Span or RHI input set for around 650mV.


The Display

The LED display is actually optional, 'tho I would at least include it inside for diagnostics even if not mounted on the front panel.  There can also be more than one display if the LED's for each step are wired in series.  I have used a range of LED's from dull green. bright green, through amber, brighter orange, to red and very bright red.  At typical on-stage distances it is difficult to see how many LED's are alight, so the change in colour and brightness give a good indication at a distance of how hot things are running.


The Fan

The fans used have been the modern electronically commutated type, nominally 12 volt at 120mA, but there is no reason why a brush-type fan could not be used.  These particular fans were recovered from dead PC power supplies scrounged free from local computer repair shops (and 'tho normally filthy, full of still useful bits).

The fan voltage is the main limitation to the supply voltage for the controller since the LM3914 itself will work up to 25V, so two 12 volt fans can be used in series if desired.  This circuit should also be capable of running two fans in parallel although this hasn't actually been tried, and it is worth noting that experiments indicate a rule of thumb that fans provide 90% of their cooling effect in the first 10% of their rev range -  a little draft has a big effect, and the law of diminishing returns sets in early.

In the event that electronic noise from the fan operation gets into the audio circuitry an electrolytic cap of suitable rating, say 100uF/25V, can be placed directly across the fan.  Similarly with the control circuit itself, although in practice the switching between segments is so slow that no "zippering" noise has been evident (as it sometimes is in VU meter applications of this chip).


Supply Voltage

Naturally if 120mA at 12 or 24 volts DC is available then supply is easy, but in practice it has been necessary to use a three pin regulator and supply side dropping resistor from the main amp positive supply rail.  The exact arrangement depends on what is available in each situation.


Construction

Construction is quite non-critical and has been normally done on strip or dab board, with the LM3914 and trim pots mounted on one board, ribbon cable to the LED's mounted on another board on the front panel, and the temperature sensors on flying leads and suitably insulated with heat-shrink.  A fairly important point is that the zero and span setting trim pots are accessible with the circuit mounted in place.

Here is one possible layout that I've used;



{errata - LED shown in wrong direction}

Operation

In operation the fan generally doesn't start rotating until things have warmed up a bit and three or four LED's are alight.  Once started however the fan will continue to run from very slow to full speed depending on the temperature.  In use it was found that something pretty gross had to be done to get the top LED's to light, such as driving the amp hard with a towel blocking the air paths.


Extras

A possibility that hasn't been tried is to use the highest indication as a thermal cutout or limiter as last-ditch amp protection.  This could be done by including an opto-coupler in series with the hottest LED to limit or mute the audio drive, or to operate a relay as a supply cutout.


Acoustic 360

When applied to the re-creation of the Acoustic 360 bass amp by JC Maillet it was desired to run three fans and this required the modification of the output circuit so that they were run via current mirrors.



In summary, this simple circuit has proven very effective in providing fan assisted cooling in solid state amplifiers, while minimising fan noise problems.




Add: 120729
Thermofan test in Acostic 361+ clone build
http://www.youtube.com/watch?v=ptxIJlRUrNw
If you say theory and practice don't agree you haven't applied enough theory.

joecool85

Very cool!  Pun intended!  Thanks for sharing, I think I might use this on one of my next projects...

I have stickied this topic so in the future folks can quickly find it without searching.
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

joecool85

Does the LM3914 get really hot when running the fan full tilt?  That seems like a lot of heat to dissipate.
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

joecool85

Quote from: Roly on May 05, 2012, 02:44:21 PM
...
One of the properties of a silicon diode is that it has a temperature co-efficient of about -20mV per degree C...

It is also possible to stack diodes or transistors in series to get a larger voltage swing with temperature if tighter control is needed, 2 = -40mV/C, 3 = -60mV/C, etc, but just for amp cooling applications I have only used a single diode drop and found that sufficient.

This voltage is input to the LM3914 on its "Sig In" line.  Because the collector and base of the temperature sensing transistor are held at the reference voltage (1.250V) the emitter voltage rises from 1.25-0.65 = 0.6 to 1.25-0.6 = 0.65 volts over the temperature span.  The Zero or RLO input is set for around 600mV and the Span or RHI input set for around 650mV.

Wouldn't this 50mv swing mean only a change of 2.5 degrees C?  That's not much differential.
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

J M Fahey

Even worse.
The silicon diode voltage drop vs. temperature (called Temperature Coefficient) is 2 (two) millivolts per Deg, Centigrade. (ºC).
But an Op Amp can *easily* multiply this by 100 and/or a Comparator circuit can easily be made to switch with a 1 or 2 mV difference, so your fan can start at, say, 80ºC and stop at 79ºC.
So in practice it's not that bad.
Some of what I said is "built in" the LM3914, so a couple ºC can be enough for it to go from 1 to all Leds ON, if you wish.

Roly

Whoops!  {Up too late.  xP }  Yes, the tempco of a silicon junction is -2.1mV per degree C, NOT 20mV as I stated above, however the mid point of the zero and span pots is 0.625V and the RHI and RLO are about 20mV (10 degrees) above and below, giving a range of about 20 degrees, depending on how you set it up.

Yes the LM3914 gets warm but it's rated for over a watt (1.36W strictly), and when it's at its worst case is when the cooling fan is going hardest.

50mV/-2mV/deg = 25 degree span.

It is also quite possible to re-arrange this sensing input to use a transistor set up for a DC gain of (say) ten and use the transistor to amplify its own change of Vbe by taking the signal off at the collector when the tempco will then be 20mV/degC, (or by whatever you set the gain to).  This may need some capacitive bypassing to stop it amplifying stray AC noise as well.

The main problem with using a sensor with a small tempco is that any stray noise turns up as "chatter" at the control output and is normally met by using a Schmitt trigger with some backlash, but this tends to widen the control range.  In the application above there is some chatter apparent as the control moves between steps, but it is minor and hasn't been a problem.

Really the main problem with this circuit is organising a suitable supply where the only available voltages are the output stage supply rails, perhaps 35 volts, which is a bit too high and has to be limited by a three-pin regulator.  Where I have applied this I have used a power resistor between the supply rail and the input of the three-pin regulator to drop the bulk of the supply voltage before it gets to the regulator input (a so-called "economising" resistor).

In the case of a LM7812 12 volt regulator the minimum input voltage needs to be about 4 volts above the output, so the resistor value is chosen to drop the available supply to 12+4=16 volts at the maximum fan current, 120mA.  For a 35 volt rail this is 35-16 = 19 volts, and at 120mA requires R=E/I = 19/0.12 = 158.33 or 150 ohms at P=E*I = 19*0.12 = 2.28 watts, say a 5 watt resistor.
If you say theory and practice don't agree you haven't applied enough theory.

J M Fahey

Cool and accurate Math, plus *practical* suggestions, which makes this post even more useful.
Thanks  :dbtu:

joecool85

Very good answers.  I'm thinking about building one of these to control a PC van in a vent that goes from my laundry room to the living room to help move airflow from the wood stove.  Probably something like 22C it turns on at level 1, and level 10 at 27C.  The whole circuit will be turned on/off by a thermostat in the laundry room.  If it is below 18C it will run, otherwise it's off.  That way it is only circulating when the laundry room needs the heat AND when the living room is hot from the stove running.
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

Roly

Well there's an application I never thought of.  :cheesy:  Just keep in mind that when using a brushless fan that they seem to need a bit of current to get moving initially, but once running they will go down real slow before stalling.

A couple of things to note from the LM3914 data sheet are that operation in bar mode may start to get a bit erratic below 200mV RHI to RLO and while this circuit goes well under that no problems have been encountered, however closing the span even more may encounter non-linearity.

The other point is that having low values for the span and zero pots gives a significant improvement in the internal temperature coefficient of the resistive divider, and this will be come progressively more important as the working span is reduced.

Be interesting to hear how it goes.
If you say theory and practice don't agree you haven't applied enough theory.

joecool85

Interesting about wanting at least 200mv differential for bar graph mode.  I suppose I could use 10 diodes (they're cheap enough) in series, or I could make a small opamp circuit to boost it. 

Even better, I could set it so that level 1 is 20C and level 10 is 60C (really this would be more ideal as I could take a temperature near the stove as opposed to air temp of the room).  This would give me 40C differential, x 2.1mv per degree C = 84mv swing.  I could put 2-3 in series and have plenty.

Some ideas/questions.

First, please update your original post to show the -2.1mv per diode instead of -20mv.

Second, wouldn't it be pushing ~4.2w at full blow?  Most PC fans I see are 0.15 amps.  LEDs are normally 20ma each, so 200ma for 10.  0.15A + 0.2A = 0.35A x 12v = 4.2w.  If the chip is only good for 1.36w, how's this work without blowing it up?

Or is it that it would push a maximum amps of what goes through the LEDs, IE 0.2A x 12v = 2.4w (still over the chip's rating) and then if that is that case, wouldn't that mean the LEDs and fan split the voltage?  So the LEDs would get say 2.4v and the fan only 9.6v?
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

joecool85

#10
I may end up going another route with a basic transistor driven fan controller.  A little less simple to "dial in", but possibly better to my application.  I'll keep you posted regardless as the one I'm looking at can handle up to a 700ma load at 12v.

**edit**
I moved my circuit to: http://www.ssguitar.com/index.php?topic=2599.0
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

joecool85

Did some research and answered my own questions.  The LM3914 doesn't overheat because it is just passing current, not shunting voltage.  So it's good for a max of 300ma.  To get full voltage to the fan you would need to run more than 12v in though since the LEDs suck up ~2.5v or so.  Ideally you'd run 12v + LED voltage and then you'd get full power.

Not sure yet if I will build one of these or a new LM317 circuit I'm working on.
Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

Roly

I've corrected the tempco in my original post.

The suggested minimum span of 200mV relates specifically to clean switching of the LED elements as a bar graph.  In this application the indication is secondary to the fan control function and sometimes not even used, however in the ones I've built, and the current mirror version shown in J.C.Maillet's video, this isn't actually an issue and simply appears as a bit of glimmer as each LED comes on, rather than a neat clean switch; it's cosmetic rather than functional.

I mention in several places that the LED outputs of the LM3914 are programmable current sinks.  The current per step is set at ten times the current drawn from the reference output.

The power dissipation in the IC would be due to the total current multiplied by the drop across each output current controller, not the total current by the supply voltage - you are calculating the power in the load, not the IC.

Because the IC outputs are current sinks the maximum current is determined by the IC, not by the fan rating, and the IC is operating within datasheet specs.

In all the builds so far the IC is also in the area being cooled by the fan, so this tends to be self-regulating anyway.

If indication LED's are included then yes, they will drop somewhere in the region of 2-3 volts, however electrically commutated brushless fan drives do not have the same voltage/speed characteristic as a conventional DC motor which tend to have a  pretty linear voltage/speed relationship.

Brushless fans have a more "S"-curve relationship in that they don't even start turning until they get a few volts across them, then the speed ramps up fairly steeply against voltage and they are running at close to full speed by the time they get to about 10 volts, the remaining couple of volts making little difference in revs, and even less in cooling effect.

Experiments I conducted back in the days of mains powered 5-inch fans and 10-inch floppy drives showed that the rule of thumb is that a fan produces about 90% of it's cooling effect in the first 10% of its rev range (or call it 80/20 if you like), allowing considerable slowing and quietening with only slight loss of cooling.  In those days I simply added a small capacitive dropper in series with the fan to take it down to about 30% revs, and experienced no overheating as a result.

Most fans are way overkill because it only requires a slight draft to remove the hot air layer against the surface of a heatsink, and that's where most of the cooling effect derives.

In free air the temperature of the air layer at the surface of the heatsink may be somewhere around 20 to 50 degrees hotter than ambient to drive convection, but once you replace it with air from a fan it makes very little difference if the layer is now 5 degrees or 3 degrees hotter than ambient because the limiting factor is the transfer of heat from the heatsink to the air, not the removal of air so heated.

This is why a semiconductor will still cook on an undersized heatsink even in in a howling gale.  Thermal circuits are still circuits, and changing the ground resistance from 5 milliohms to 3 milliohms won't help if there is still 5 ohms of resistance elsewhere in the circuit.
If you say theory and practice don't agree you haven't applied enough theory.

joecool85

Life is what you make it.
Still rockin' the Dean Markley K-20X
thatraymond.com

Roly

You could be right.  That's JC Maillet's adaption which I haven't built, but since the current drawn by these pots from the ref output sets the output current, and JC had some LED's lighting up, 5k does look too large.  I've written to ask him.
If you say theory and practice don't agree you haven't applied enough theory.