Hi. I did numerous experiments on this in April and May 2002. You might want to search for my posts (macklin01) in that time range. Let me share a few findings:
1) The fan RPM signal depends on there being a common ground with the motherboard. (Basically, it oscillates between an open ground and a higher voltage.) So, if you use a 7V mod, where you have changed the fan's ground to 5V, the motherboard won't understand the signal. So in a word, the fan's RPM signal won't work with the 7V mod.
2) At the standard voltage, you can still pass the RPM signal to the motherboard. Simply extend the yellow RPM sense wire from the fan to the appropriate motherboard pin.
3) The RPM signal DOES work so long as the ground connection remains unmodified. This means that it still works if you vary the voltage on the positive side of the fan. For example, you can use a rheostat or NPN-transistor w/ rheostat control, or voltage regulator to vary the positive voltage coming into the fan's positive terminal, and as long as the negative terminal is still connected to the ground, the RPM signal will work.
4) Preliminary testing with PWM (pulse-width modulation) methods indicate that even in the "off" portion of the duty cycle, the RPM wire continues to transmit a signal for a short time, but that signal may well be spurious and provide false readings. On the other hand, the earlier points suggest that if you have a PWM scheme that varies the +V voltage between +12V and a lower, but still positive voltage, you might get an accurate RPM reading.
Using these and other results, I constructed a fanbus design that:
1) Passes the fan RPM signal on to the motherboard.
2) Allows the motherboard to turn the fans off when teh computer enters standby.
3) Toggles between full speed and variable speed.
4) Doesn't require heatsinks for voltage regulators, rheostats, or anything of the like.
5) Provides a safeguard "minimal allowable voltage" to ensure that fans don't accidentally get slowed so far that they stop rotating.
Schematic:
Some data on measured fan speed vs. voltage across thefan:
A screenshot of the fans reporting RPM data while under the variable speed setting:
Some more notes:
1) Instead of using the standard connections for the fans that you often see, where the fans molex connections are cut off and the wires are screwed into the PCB, I recommend buying 3-pin fan extension cables and usign their molexes to form the connection from your fanbus to the motherboard and from your fanbus to the fans. That way, you can easily swap fans in and out, and your design is as compliant with existing standard as possible.
2) You can solder 2 NPN transistors in parallel to obtain higher current capacities and reduce the load on each component.
3) I would recommend a smaller variable resistor than 100 kOhm. Perhaps 1/10 that.
4) You can also use a zener diode (5.1V, for example), in combination with regular diodes (usually .5V - .7V drop) to get a minimal voltage you want. I think I use a 5.1V zener diode in series with a single regular diode where the schematic shows a single zener diode. Testing is necessary for attaining the proper configuration.
I hope this is helpful for you. If you have additional questions, please post here. Have fun, and good luck!! -- Paul