Hi,
I have observed (both first hand and anecdotally) that this is a very common problem with MDD G4s under OS 9, and that people have come up with solutions like replacing the fans with quieter versions (probably not a terrible idea *anyway*), downgrading from a dual to a single CPU card, and, the most obvious, running Mac OS X exclusively (and using Classic if running Mac OS X =< 10.4.11) and using the CHUD tools.
With all these workarounds, I've noticed that nobody has even considered an attempt at duplicating OS X's control mechanism in an OS 9 environment.
It seems like, at the very least, an OpenFirmware hack could be devised that would at least allow the fans to run more slowly, but a more flexible approach would be to somehow write an extension and/or control panel that would exert the same kind of control over the fans as OS X's AppleFan.kext, which would probably be ideal. I've wondered why Apple itself wouldn't have done something similar, but it was probably due to the fact that the MDD was released when OS X was taking over (and OS 9 was on its way out), so it's quite understandable. Likewise, it's understandable that nobody else would've cared as much since the Quicksilver, the MDD's immediate predecessor, was perfectly adequate under OS 9, so there was little reason to bother trying to run it flawlessly on an MDD, despite the fact that the MDD was an OS 9 bootable machine (except the FW800, of course).
So, is anyone willing to take on this project? I'd like to do it myself, but I don't know enough yet 
c
I've been thinking about this post throughout the day and I just wanted to throw in my two cents.
I don't know enough about the hardware to make a firm statement, but my assumption would be the fans operate at a specific speed for a reason. Noise reduction would almost certainly, in my opinion, be most safely be achieved by replacing the fans with newer, quieter counterparts rather than try and throttle the current fans and end up with an overheating machine.
That said, I like the idea of working on a new MacOS 9 app. I've done some reading and it looks like a common workflow from the time would have been something like MPW (C) + ResEdit?
I am actually very interested in working on something like a hardware monitor (CPU temp, RAM usage, network activity, etc.) Just something static that would be a neat widget.
I am going to look into getting a copy of MPW and ResEdit working on my Sawtooth. If anyone has a better workflow or can speak from experience with MacOS 9 software development, I'd be glad to hear their thoughts.
I don't know anything about doing this at the OS level, but thinking about the hardware level gives me a couple of ideas, the most interesting of which would broaden the project to include USB capable, rather than just the MDD.
I've been watching these Digital Thermal Probe or Sensors as an alternative to thermal imaging for quantifying Compact Mac convection cooling nod results. I think they might work out well for this application.
Place the sensors in critical airflow paths. Examples off the top of my head:
_____outflow side of the heat sink airstream
_____inflow side of PSU
_____inside the PSU
Using a modern CPU widget(Pi, Arduino/whatever, I was looking at using my NuBus DIO card for my project) to report sampled temps to the Mac via USB and control the DC->DC Power supply widgets voltage powering the fans using output data from the Mac over same seems like the way to go to me.
_____place the sensors
_____jumper wire the fan's power from the DIO processor thingamajig.
_____digital (or analog) controlled rheostats(?)
The controller app would be a background app. As a safety feature, the microcontroller widget supplies full voltage until the app is up and running or if the app fails to respond over USB when polled in the case of a bombed system.
Dunno, haven't had enough coffe to really think it out, just thought I'd offer up this notion for a general purpose multi-Mac hack.
IIRC, this approach would put the hack methodology squarely within your knowledge/experience, CC.
jt
Looking at my 2003 1.25GHz MDD, there's no sensor lines on any of the fans. I suspect then that PSU fan speed is controlled by thermistors on the PSU itself. On the 120mm case fan there's a third pin on the motherboard connector, but only two wires go to the fan.
If OS X runs quieter on the MDD, it could be that it's more efficient and places less demand on the PSU than OS 9.
jt, I like your microcontroller suggestion.
An arduino with a temperature sensor or two could control fans and report information via serial->usb, and possibly receive commands as well to control several fans.
Initial uploading and updating of code on the Arduino would have to be done in OS X, but once running it would be OS independent.
I'm kind of full on projects at the moment, but I might try to at least get OS 9 talking to an arduino just for giggles.
OS X has the ability to put a PowerPC processor into nap mode, using CHUD Tools 3.5.2 as an example. Under OS 9. there is no CHUD Tools, so no nap mode for the processor. Therefore, a G4 MDD will run full-throttle, fan-wise, whereas, under OS X, it will spin down a bit when the processor(s) is(are) put into nap mode.
-J
I don't know too much about the cooling systems for relatively current computers, graphics cards and such, but it seems like they only need full throttle fans when the system is under heavy load.
A G4 like my QS'02 idling the hours away while I work in Illustrator and GraphicConverter under OS9 puts out some heat, but it's not the blast furnace it might be running some resource intensive game or other at stupid frame rates/resolutions.
Am I right in assuming that the fans run at the same speed in either mode?
If so, the DIO sampled temps at critical airflow points ought to be able to be used to tweak the fan speeds, keeping the temperature ranges at the sampled points at optimal levels while reducing airflow/sound appropriately for CPU loading under any load scenario.
I like the idea of going all the way back to the first PCI(USB) architecture machines in hardware, but that's me. I usually take crap like this all the way over the top.
I can't program micro-controllers, so I'd probably just take a whack at doing it like I would have in college . . .
. . . hack a combination analog (read mechanical) thermostat/rheostat kluge and be done with it. :ebc:
is there anyway i can stop the fans from powering up + down on my mdd 1.25ghz? its driving me freaking crazy i hate hearing the fans speed up + slow down again..
as long as the fans stay at the same speed i get used to it and it doesnt bother me, but its the audible change that i hate..
theres no CHUD tools for mac os 9.. what can i do about this? the only thing i can think of is to replace the power to the fans with a steady voltage, drawing from a 4pin molex straight to the fans inside the power supply.. rather then having them connected via the original power source that gets its voltages changed......
worst ever design by apple!!!!!!!
im at the pt of just turning it off and never using it again, favoring my quicksilver 933mhz which doesnt have any noise at all by comparison