This is fundamentally completely wrong
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
This is fundamentally completely wrong
Y'know it's not really proper to be undervolting a fan to control its speed. The results are irreguler, unstable, and generally unpredictable if you want things like the tachometer to work for very low voltages.
Current fan speed limiting is really a blunt process, sort of like buying a fast car and then removing the rubber tires, and driving it on the rims to slow it down. Sure, this works, but it's sloppy, it introduces instability, and it can make operation unpredictable. Plus (as with the 7v mod) it can be damaging to the system and cause grief for the operator.
Fan motors are basically DC-steppers with an onboard stepper control circuit. What needs to be done is for the stepper coils to be broken away from the controller, so that you can use a proper controller to set the fan speed and do away with hacking it with voltage limiting.
We need fans that are nothing more than coils, magnets, and the rotor, with a 4 (or so) wire cable coming off the fan hub, direct from the drive coils. This ultra-dumb fan is then plugged into an EXTERNAL motor controller.
With an external controller, you could set the coils in the rotor to spin at whatever speed you want, without the voltage trickery. You want 1000 RPM? Fine, dial it up and the controller will deliver full 12v pulses to spin the rotor at 1000 RPM, or whatever.
Hell, put a tach in the rotor and feed that off on the wires too, and the external controller can now sense rotation.
Really, this undervolting is a sloppy process. It works, but it's like using a golf club to pound nails. We need a more elegant solution, though it's probably not going to happen until fan makers take notice of the special needs of their silent computing customers, and are willing to design controllerless fan rotors.
-Scalar
Current fan speed limiting is really a blunt process, sort of like buying a fast car and then removing the rubber tires, and driving it on the rims to slow it down. Sure, this works, but it's sloppy, it introduces instability, and it can make operation unpredictable. Plus (as with the 7v mod) it can be damaging to the system and cause grief for the operator.
Fan motors are basically DC-steppers with an onboard stepper control circuit. What needs to be done is for the stepper coils to be broken away from the controller, so that you can use a proper controller to set the fan speed and do away with hacking it with voltage limiting.
We need fans that are nothing more than coils, magnets, and the rotor, with a 4 (or so) wire cable coming off the fan hub, direct from the drive coils. This ultra-dumb fan is then plugged into an EXTERNAL motor controller.
With an external controller, you could set the coils in the rotor to spin at whatever speed you want, without the voltage trickery. You want 1000 RPM? Fine, dial it up and the controller will deliver full 12v pulses to spin the rotor at 1000 RPM, or whatever.
Hell, put a tach in the rotor and feed that off on the wires too, and the external controller can now sense rotation.
Really, this undervolting is a sloppy process. It works, but it's like using a golf club to pound nails. We need a more elegant solution, though it's probably not going to happen until fan makers take notice of the special needs of their silent computing customers, and are willing to design controllerless fan rotors.
-Scalar
Hi scalar,
Long time listener, first-time caller.
You bring up some excellent points in your post. Your absolutely right that undervolting is not the optimal method for decreasing the speed of a fan. However, there are a few things to keep in mind.
Panaflo fans, as an example, are rated for operation at 7V. So therefore, application of a 7V supply should not harm them in any way.
Also, most DC fans are not really designed around stepper motors. Stepper motors adjust the voltage of the four poles to acheive motion in steps. These steps would not get any slower if the voltage was to decrease unless the timing part of the circuit was to be affected by the source voltage. I suspect that DC fans are based around a shunt coil to maintain the speed.
All that being said, I still agree with you that there is a better way to control the speed of a DC fan. The best solution is something a lot of the more expensive controllers have built in. That is pulse width modulation (PWM). This solution rapidly turns the 12V supply off and on. Any fan could theoretically acheive any speed you want with this scheme.
I did a quick search on google for it:
http://www.analog.com/library/analogDia ... _speed.pdf
Best regards,
AJ
Long time listener, first-time caller.
You bring up some excellent points in your post. Your absolutely right that undervolting is not the optimal method for decreasing the speed of a fan. However, there are a few things to keep in mind.
Panaflo fans, as an example, are rated for operation at 7V. So therefore, application of a 7V supply should not harm them in any way.
Also, most DC fans are not really designed around stepper motors. Stepper motors adjust the voltage of the four poles to acheive motion in steps. These steps would not get any slower if the voltage was to decrease unless the timing part of the circuit was to be affected by the source voltage. I suspect that DC fans are based around a shunt coil to maintain the speed.
All that being said, I still agree with you that there is a better way to control the speed of a DC fan. The best solution is something a lot of the more expensive controllers have built in. That is pulse width modulation (PWM). This solution rapidly turns the 12V supply off and on. Any fan could theoretically acheive any speed you want with this scheme.
I did a quick search on google for it:
http://www.analog.com/library/analogDia ... _speed.pdf
Best regards,
AJ
-
- Posts: 13
- Joined: Sun Sep 15, 2002 6:32 pm
- Location: Madison, WI
Scalar was very likely refering to the common "12v and 5v make 7v" mod using standard Molex connectors; with this mod, the concern is not for the fan, but rather for the PSU which is then sinking current into your 5v line.AJ913 wrote:Hi scalar,
Panaflo fans, as an example, are rated for operation at 7V. So therefore, application of a 7V supply should not harm them in any way.
Will,
Thanks for pointing that out. I reread scalar's post and I see what he meant when he said "system". That brings up another issue: has anyone blown their power supply by implementing the mod?
Mike,
Although the pulses inherrently introduce noise into the fan, I believe there are fixes for that by decreasing the slew rate of the pulse edges (i.e. how fast the pulse edge takes to go from one state to the other).
When I have some time, I'll look into designing one just to see if it can work.
AJ
Thanks for pointing that out. I reread scalar's post and I see what he meant when he said "system". That brings up another issue: has anyone blown their power supply by implementing the mod?
Mike,
Although the pulses inherrently introduce noise into the fan, I believe there are fixes for that by decreasing the slew rate of the pulse edges (i.e. how fast the pulse edge takes to go from one state to the other).
When I have some time, I'll look into designing one just to see if it can work.
AJ
-
- Posts: 539
- Joined: Fri Apr 04, 2003 6:35 am
- Location: Cambridgeshire, England
Yup.AJ913 wrote:That brings up another issue: has anyone blown their power supply by implementing the mod?
Well not me personally, but I'm helping a buddy at work who brought in his system tonight for me to look at.
I gave him a couple of 92mm NMB fans that were on special at Directron (real screamers!) and told him to run them at 5V (move the supply pin from yellow to red). He did that but found that his system didn't have enough airflow and that the fans clicked (hey, they were cheap!). So he read online (not here!) about the notorious 7V mod and tried that (without asking me...dumbass!). Upon power up, his PSU heaved up it's inards and let out the "magic blue smoke". It also took out his new AthlonXP 2400, mobo, RAM, hard drive (smoked three surface-mount resistors on controller daughtercard). Luckily, his vid-card, modem, and dvd/cdrw survived but the rest were 'crispy-critters'.
Verdict: 80% system meltdown!
I agree with scalar. DC fan design hasn't changed dramatically for years. With the exception of TMD (too loud). But the TMD design looks promising. Hopefully, they can implement PWM controllers into the motor design.
With an external controller you could shape the drive waveform to be quiet. Square-wave drive too loud? Fine, instead drive the coils with pulses that gently slope up from zero, or perhaps a sinewave drive.
Also, there's no need to supply full power once the fan is running at speed. So supply full 100mA pulses to get the rotor up to speed, but once it gets there, switch to a soft, smooth 20mA sinewave to maintain it there quietly.
An external drive design would allow you the flexibility to do this, with signal control far beyond what is available now.
-Scalar
Also, there's no need to supply full power once the fan is running at speed. So supply full 100mA pulses to get the rotor up to speed, but once it gets there, switch to a soft, smooth 20mA sinewave to maintain it there quietly.
An external drive design would allow you the flexibility to do this, with signal control far beyond what is available now.
-Scalar
I wonder if something like this could be achieved using a currently available fan with a tachometer wire. Feed suitably shaped pulses into the power lines, synchronised with the tachometer signal. Tricky, but it might work.
Ultimately, I think the best solution would have the control circuitry in the fan. The fan would have three wires: 12V and 0V for power, and a control line which tells the fan how fast it should go. Possibly a fourth wire for an alarm signal or tach signal.
Ultimately, I think the best solution would have the control circuitry in the fan. The fan would have three wires: 12V and 0V for power, and a control line which tells the fan how fast it should go. Possibly a fourth wire for an alarm signal or tach signal.
This is pretty common with industrial fan control systems and there are several application notes out there on how to implement this. This reduces but doesn't eliminate noise by any means. The vantec stuff uses a simple version of this.scalar wrote:With an external controller you could shape the drive waveform to be quiet. Square-wave drive too loud? Fine, instead drive the coils with pulses that gently slope up from zero, or perhaps a sinewave drive.
Don't forget that as the voltage is reduced the tachometer signal disappears. This makes it very difficult to measure RPM, especially as the RPM gets very low.scalar wrote:An external drive design would allow you the flexibility to do this, with signal control far beyond what is available now.
Hmm... what about using the AC power directly from the wall outlet with a transformer and few diodes to half-rectify the voltage? That would be like 50+% idle PWM, but the duty cycle would be sinewave (and less than 50%). Most fans can tolerate 7-14V as operating voltage, so one would still have something to adjust. 50/60 Hz might be too high though.scalar wrote:With an external controller you could shape the drive waveform to be quiet. Square-wave drive too loud? Fine, instead drive the coils with pulses that gently slope up from zero, or perhaps a sinewave drive.
Also, there's no need to supply full power once the fan is running at speed. So supply full 100mA pulses to get the rotor up to speed, but once it gets there, switch to a soft, smooth 20mA sinewave to maintain it there quietly.
An external drive design would allow you the flexibility to do this, with signal control far beyond what is available now.
-Scalar
Cheers,
Jan
-
- Posts: 37
- Joined: Thu May 08, 2003 2:32 pm
- Location: Maine
Could be overthinking this
I suspect that most cheap DC fans are speed controlled by friction. Air reisistance and bearing/bushing friction probably limit the speed.
Therefore, lowering the voltage simply reduces the force opposing said friction and therefore reduces the speed.
Otherwise, reducing the voltage would increase the amperage and melt all the damn wires in the thingy.
Therefore, lowering the voltage simply reduces the force opposing said friction and therefore reduces the speed.
Otherwise, reducing the voltage would increase the amperage and melt all the damn wires in the thingy.
I don't think friction could be used to accomplish this as it would be difficult to repeat/maintain. The speed control system is probably like any other brushless DC motor. Motors use a opposing coil to ensure that, once the expected speed is acheived, the additional magnetic force required to increase the speed of the motor is cancelled out.
-
- Posts: 37
- Joined: Thu May 08, 2003 2:32 pm
- Location: Maine
Back EMF
True back EMF would be the greates source of friction, sorry I didn't remember to include that, however, having the back EMF in sync is a major part of the motor design in the first place otherwise the motor would be hopelessly ineffecient.
For a really good overview of DC motor design and control have a look at: http://www.angliac.com/st/data_from_st/7209.pdf
For a really good overview of DC motor design and control have a look at: http://www.angliac.com/st/data_from_st/7209.pdf
Worse, in my experience, there's a lot of fan stall, esp for our beloved panaflos, and fan startup is (again from my experience with the fanstorm) a probabilistic behavior, which is not something one wants.MikeC wrote:The other issue is that PWM creates its own problems with pulsing & other nasty noises in fans when slower speeds are desired -- at least this is my limited experience. I am told the fan and PWM circuit have to be designed for each other to avoid this problem.
I've never heard of the pairing of fan and PWM circuit before, but it makes some sense. Do you have some references regarding it?
Some notes...
o Only use a PWM solution where you can control the frequency
o PWM solutions can require all fans to be same model & size & type
The reason is that the optimum frequency for one fan, may be the
absolute worst frequency for another fan - and cause loud growling.
Remember fan designers go to extraordinary lengths:
o Panaflo use extremely quiet circuitry
---- the circuitry itself is a major factor in low-rpm fan noise
o Bearings are a critical factor
---- re thrust loading & face area
---- re off-power loading & kick-back reaction
Those 2 factors in particular can considerably vary how PWM sounds
from the paper-specification into the real-auditory-impact. Very often
it can be surprising and *very* frustrating to have a design fail.
Example:
o PAPST 4312, 120mm - smooth silence on PAPST PWM
o PAPST 6122NM, 174mm - utterly atrocious on PAPST PWM
o Panaflo 120x38mm - sound like rotweillers on the PAPST PWM
The real benefit of PWM is this:
o Logic guarantees fan start
---- eg, runs fans at 100% voltage for 2-10secs
o PWM creates an /average/ voltage well below fan start voltage
---- some PWM controllers get down to == 3.5V
Also, PWM circuitry can add in temperature alarms & other services.
We still have an old problem:
o Fan failure is not easily detected since the detection circuits need 12V
o Yet both PWM & voltage variation affect that monitoring voltage feed
Yes, the solution is to remote control the fan re speed control.
This is done by some NMB & Nidec & Panaflo fans - using a grey wire.
However, it's usually not a great solution - just "lower or scream" noise.
Sadly much equipment doesn't use quiet fans, but also does not use
fans in a manner where low overall product noise can be achieved.
So it comes back to a Gestalt based solution:
o A silent product is the sum of the components within it
o In particular the sum of the noisiest vs quietest components
When you do get close to silence, the problem changes:
o Qualitative vs quantitative measures of silence
o Some prefer the bearing noise of Panaflo over PAPST
o Others prefer the whitenoise of higher-rpm NMB fans
PWM is a nice solution:
o Can control a very large wattage of fans - 25-65W is a doddle
o Can do so *very* efficiently - re power & energy wasted
---- there is minimal thermal heating by a PWM controller
o Can tailor rpm/noise to the thermal environment suffered
However that latter point poses one of the biggest problems for PWM
solutions - they are a System. Where you site the thermocouple or
thermistor will directly affect the rpm of the fans connected to it. That
can result in an over-heated or over-noisy enclosure with a lot of fiddling
to achieve a good balance. NMB stick the thermistor on the fan to solve
this, PAPST use a remote thermistor instead for hours of sticky-tape-fun.
As a system, what we need is better enclosure design:
o acoustic baffling
o ducting of the airflow to hot objects
o ducted removal of heated airflow from hot objects
o intelligent fan use & selection
I'll leave those sufferers with Apple G4 Cube's 120mm 150cfm 57dB(A)
to add their own comments about how good thermal components (big
passive heatsink) can be destroyed by poor selection of active components (fans).
Look for "PWM" solutions in the hobbykit & other suppliers.
For more advanced solutions, see www.control-res.com for an
example of how commercial controllers approach the problem.
Lastly, PWM can show up bearing & quality differences quite markedly.
--
Dorothy Bradbury
http://homepage.ntlworld.com/dorothy.br ... anaflo.htm
Panaflo L1A 5.99ukp with Free 1st-Class Insured delivery
o Only use a PWM solution where you can control the frequency
o PWM solutions can require all fans to be same model & size & type
The reason is that the optimum frequency for one fan, may be the
absolute worst frequency for another fan - and cause loud growling.
Remember fan designers go to extraordinary lengths:
o Panaflo use extremely quiet circuitry
---- the circuitry itself is a major factor in low-rpm fan noise
o Bearings are a critical factor
---- re thrust loading & face area
---- re off-power loading & kick-back reaction
Those 2 factors in particular can considerably vary how PWM sounds
from the paper-specification into the real-auditory-impact. Very often
it can be surprising and *very* frustrating to have a design fail.
Example:
o PAPST 4312, 120mm - smooth silence on PAPST PWM
o PAPST 6122NM, 174mm - utterly atrocious on PAPST PWM
o Panaflo 120x38mm - sound like rotweillers on the PAPST PWM
The real benefit of PWM is this:
o Logic guarantees fan start
---- eg, runs fans at 100% voltage for 2-10secs
o PWM creates an /average/ voltage well below fan start voltage
---- some PWM controllers get down to == 3.5V
Also, PWM circuitry can add in temperature alarms & other services.
We still have an old problem:
o Fan failure is not easily detected since the detection circuits need 12V
o Yet both PWM & voltage variation affect that monitoring voltage feed
Yes, the solution is to remote control the fan re speed control.
This is done by some NMB & Nidec & Panaflo fans - using a grey wire.
However, it's usually not a great solution - just "lower or scream" noise.
Sadly much equipment doesn't use quiet fans, but also does not use
fans in a manner where low overall product noise can be achieved.
So it comes back to a Gestalt based solution:
o A silent product is the sum of the components within it
o In particular the sum of the noisiest vs quietest components
When you do get close to silence, the problem changes:
o Qualitative vs quantitative measures of silence
o Some prefer the bearing noise of Panaflo over PAPST
o Others prefer the whitenoise of higher-rpm NMB fans
PWM is a nice solution:
o Can control a very large wattage of fans - 25-65W is a doddle
o Can do so *very* efficiently - re power & energy wasted
---- there is minimal thermal heating by a PWM controller
o Can tailor rpm/noise to the thermal environment suffered
However that latter point poses one of the biggest problems for PWM
solutions - they are a System. Where you site the thermocouple or
thermistor will directly affect the rpm of the fans connected to it. That
can result in an over-heated or over-noisy enclosure with a lot of fiddling
to achieve a good balance. NMB stick the thermistor on the fan to solve
this, PAPST use a remote thermistor instead for hours of sticky-tape-fun.
As a system, what we need is better enclosure design:
o acoustic baffling
o ducting of the airflow to hot objects
o ducted removal of heated airflow from hot objects
o intelligent fan use & selection
I'll leave those sufferers with Apple G4 Cube's 120mm 150cfm 57dB(A)
to add their own comments about how good thermal components (big
passive heatsink) can be destroyed by poor selection of active components (fans).
Look for "PWM" solutions in the hobbykit & other suppliers.
For more advanced solutions, see www.control-res.com for an
example of how commercial controllers approach the problem.
Lastly, PWM can show up bearing & quality differences quite markedly.
--
Dorothy Bradbury
http://homepage.ntlworld.com/dorothy.br ... anaflo.htm
Panaflo L1A 5.99ukp with Free 1st-Class Insured delivery
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
Nothing written that I've seen. It came up in conversation with a fan control engineer. But this only makes sense in some kind of completely integrated system. EDIT -- as "Dorothy" spells out so well aboveI've never heard of the pairing of fan and PWM circuit before, but it makes some sense. Do you have some references regarding it?