Electronics Fan sensor Questions

Ahaaaaaaa im back and you thought u were all well rid of me

For now ive just got a quick question about fan sensors but if i get a bit further with this project (and im about a quater of the way there already) then ill post more about it here

In fans with tachometer output there is an ic which im assuming handles the sensor output. Can you power the fan seperately from this ic so you can still use PWM to control a fan and use the tachometer output to get its speed

 

i think the RPM sensor is a normal hall sensor, so you could jusr run some tiny wires directly to it... just remember to cut the original PCB tracks going to it..

i'm not sure about this though

 

A honest comment would get me banned.

Short answer is no, as if the IC supply is not continuous, it will miss pulses in the PWM spaces. Trying to make the ic supply continuous with smoothing cap or another feed? Depends. I haven't dissected any fans with speed o/p, but my Panaflo L1, a small hammer reveals, has a hall sensor and a 16-pin ic (NBC3801, find me a datasheet & I'll have your babies), motor coils fed H-bridge style from 2 ic o/ps, so smoothing the ic supply won't work easy, it's also the motor supply. Some are even simpler, just a 4-terminal hall switch to produce the rotating magnetic field, bit like this. To add a speed o/p only needs an extra transistor buffer as shown by Spiv here.

Best bet IMHO is to integrate the PWM back to a DC level. You lose some advantages of PWM (runs at low mean voltage) but keep the efficiency, as if that matters on a 2W fan or two.

 

right PWM seems a waste of time (if its gonna be converted back anyway) so over to plan B

Ive just had a look at http://www.bit-tech.net/article/32/ and if i understand it correctly you can use any value of resistor for the pot and R1 providing that the pots max value is 10 x r1.

I havent used variable voltage regulators before as i havent needed to so sorry if thats a load of rubbish.

If what i said is correct tho i could use a digital pot to control the voltage out (rapid only sell a 100k one) and controll it from a microcontroller (im playing with p89c66X chips atm at uni) and in theory control a pretty much unlimited number of fans from a 3 pin serial interface . . . which is nice.

do you think this would work? my electronics is a bit rusty as im not doing anything more complex than resistor networks at the moment at uni (its really depressing).

 

Ive been pndering about the PWM and RPM sense line problem a little myself.

Heres a thought: Instead of applying a 12V amplitude PWM signal to the fan, why not apply a 7V amplitude PWM signal ontop of a constant 5V DC voltage?

I havent gotten really far into how to do this yet. Take for example the output of a PWM generator. This would drive a transistor who's ground reference is the 5V line. The transistor would supply a 5V to 12V swing.

Once that is figured out, the RPM sense line would constantly have at least 5VDC to power the halleffect sensor, thus eliminating the problem of it not having supply current to power the output.

Problems: Most fans just have an NPN transistor for grounding the sense line. But some fans use internal resistors for pull-ups (pulling up the fan supply voltage) and some fans use internal zener diodes (so as to not allow any more than the zener voltage to be output on the sense line). If a fan uses the zener, then the 5V-12V PWM signal probably wont work. On the other hand, not many fans use this internal setup.

Thoughts?

 

you could put a cap in series between the motherboard and fan sense line. IIRC this allows ac signals to pass and DC gets blocked so the pulses from the fan should go sailing through

the only problem is the speed will change cos a 50% duty cycle isnt 50 %@ 12 50% @0 its 50@12 and 50@5 so more like 75% duty cycle

 

Ratio for 12v output is about 9:1, though with a nasty 317T 8:1 will do as you'll only get about 10.5v max. from the 12v ingoing. And top recommended size for the low-value resistor (carrying the reference voltage between o/p and adj pins) is 1k.

LDO regs are better, you'll get 11.9v out and they behave better with higher-value resistors, though a digi-pot may still need a buffer, 100k is a lot & I don't think they can handle much current.

Thinking about it, I'd design
(a) Digipot giving 0-12v output from 12v input, feeding
(b) High-gain NPN/PNP emitter-follower to feed the fan.

Easy-peasy, 11.3v max.

Look at this Winbond system which combines a pwm system with fiddles to keep the speed signal. (Then tell me what PWM frequency the fan sees ).

 

i think this is how the winbond chiup does it.
the chip used in the pwm article on the bit site uses a freq of 30 hz.

i dont understand a lot of the winbond datasheet but it mentions PWM frequencies of 480 odd hz.

if the pwm speed is fast enough you wont miss the pulse from the pwm sensor it would look something like this on a scope

Code:

normal sensor signal
        -----        -----        -----
--------     --------     --------     --------

sensor signal with PWM
        - - -        - - -        - - -
--------     --------     --------     --------

so all u need to do is make really sure you only read each pulse once

 

It can be done that way , but I'm going to try the Micrel integrator, should work on any pwm signal. I've already got a spare opamp on my LM324 circuit.

 

heh was about to post a question about emitter followers but thought id google it 1st. . . google turned up cpemmas page

/me feels silly

cheers again cpemma

 

Don't bother with op-amp integrators, unless you buy a £4 rail-to-rail they'll only give about 10v

The fixed resistor is only so a full turn on the pot covers about 5.5v-11.3v output. With a digipot the control input is what you make it so doesn't really need one AFAICS. Pot across 12v, wiper pin to NPN base. Go for really high-gain transistors to keep the wiper pin current low.

 

errrrr can u give me an idea of what to change the values of the resistors to to work with a 100k digital pot as i havent really used transistors before appart from in college where i ended up with blisters on fingers and a circuit that didnt work

<edit>also can you tell me how to work this out for future reference pls?

 

Easiest way is to work backwards. If the fan draws 200mA at 12v then worst-case NPN base current (=pot wiper current) is 0.2/h_fe1h_fe2

So if you used a BC108C (h_fe=600) for the NPN and a BC328 (h_fe=100) for the PNP, total gain is 60k and you're only drawing 0.2/60k=3.3uA max from the digipot. 500mA of fans would draw 8.3uA.

The pot will already have 12/100k=120uA through it so the extra is negligible.

A simpler way is to use a single NPN Darlington, the ZTX605 has gain of 5k and can dissipate 1W, roughly 330mA-worth of fan.

Are we talking DS1267? I've only had a quick glance but it says max wiper current is 1mA, so that's OK. Adding the op-amp buffer they show you can use some meaty power transistors and run a few amps fan loading.

 

cheers again

dont know why im making this fan controller really tho as i only have 3 low noise fans in my pc :-/ oh well ill give it to someone when im done or something still its an interesting project (thats what the lcd code i put up is part of)

and yeah its the one from rapid made by dallas think thats the one

<edit> by op amp buffer do u mean voltage follower?

 

Yes, same idea. I was reading that stuff about wiper resistance up to 1k, if you got near the 1mA limit you could lose 1V worst case. An emitter-follower is like a voltage follower, what goes in comes out but you lose at least 0.7v anyway, & without the opamp you'd lose a fair bit more if you had a big load. The buffer also protects the expensive chip against noobs.

edit: I keep forgetting op-amp voltage followers can only swing up to about Vs-2