"Standard supplied Udoo Bolt fans" are at least two different COFAN fans, and they are both 5V fans, not 12V. And yes, we had tried, stock fans will rotate at top speed, even when Bolt is in shutdown state :| And yes, some of us had tried 12V fans against Bolt's 5V (supplied either by Arduino or CN7), and depending on particular fan some would work fine at their lowest (and quietest) RPM. Whether this approach is enough to cool fully loaded Bolt for indefinite time is matter of opinion, and probably also depends on a particular fan
I have been doing x265 encoding where I run the CPU at close to 100% for hours on end and the temperature still does not get over about 70-75C so it should not be throttling. With the stock (noisy) fan the temperatures are about 10C lower on the same workload when it is running continuously.
Another update: I flashed the latest bios and now unfortunately there is a very annoying noise that wasn't there before running the bios my Bolt was delivered with. The noise comes from the coils of my pwm fan. This might be some sort of harmonics, since the noise is somewhat tonal at certain rpm when the rpm changes. I guess the signal frequency has been reduced and produces harmonics in the audible range now. Also, there is a new option for "Generic PWM Fan" doesn't make any sense to me whatsoever... What exactly do you mean by generic pwm vs 4-pin fan? Need help
And I forgot to mention, there was a serious increase in boot time, even delayed key response in bios after changing values in the external fan temperature settings. The bootime and behaviour returned back to normal after I detached the RTC battery. There should be a ticket filed, although this happened while running the old bios, but the issue persisted after the update so I have it permantly disconnected now. Just mentioning so bugs are more easy to be found and eliminated...
Sorry, not intending to spam here but I reset the bios to default settings now, and then set my desired values for the "4-wire fan", not using the "generic pwm" only once. The new noise that appeared after the bios update is now gone and the fan is almost inaudible again at 23% fan speed. Seems to me like there is some bug connected to changing and storing some values in the bios settings.
The CN7 PWM signal is designed to switch the supply voltage to the fan. It has capacitors on it so it wont work as PWM for a signal pin and explains why @Zgembo sees a saw tooth. To make a decent square wave it would need a fairly low value pull down resistor, I would try 220R but I am not a home until next week. There is no way "green dot" stock COFAN would ever be able to control fan speed, as the PWM signal settings in BIOS does not affect voltage in any way PWM doesn't affect voltage unless it is post filtered, that is the point of it. It normally does affect fan speed though, even on fans with no PWM input. The COFAN seems odd in this respect but when I am home next week I will do some experimenting.
Well, to me it seams that there is significant difference between what it is designed for and what it ended up to be. According to the https://www.udoo.org/docs-bolt/Hardware_References/Pinout_Headers.html, part stating: "For example, this is the circuit used to drive the 5V fan installed above the heatsink". If this diagram is really to be trusted on what lies between FANOUT1 (raw Embedded Controller 3.3V PWM signal) and CN7 pin2, what is your opinion or insight into the situation @nophead ? The power CMOSes seems fine to me in their roles of switching to 100% cycle when CPU is hot (APU_PROCHOT, Q66A), shifting the signal from FANOUT1 3.3V to CN7:2 5V (Q66B), and adding some current into the picture (Q65), 'cause COFAN is power supplied by nothing than that... But aren't these C887 & C889 (and maybe C886 too) capacitors to blame that CN7:2 signals are so much distorted at higher frequencies? What I am saying, and what my measurements totally agrees to, is that below ~1500Hz CN7:2 signal is totally PWM compliant, between 1.52kHz and 9kHz it is an ugly square-sawtooth hybrid (heavily depends on duty cycle on higher freq.) and at above ~9kHz it stops being a signal at all, just flat 5V. It seems to me that CN7:2 signal is designed to nothing other than that non-green dot COFAN (F-6010M05BII), which is one super obscure fan (try googling 'GR276 hall effect fan driver' that controls it and you'll see why). I see this as a miss-design, it effectively screws up the ability to RPM-control 99% of most other "normal" CPU fans out there. Maybe it wouldn't end up as a big issue if COFAN fans were any better, but as most of the Bolt owners already know they are pretty much nothing but a noisy pieces of crap, totally inappropriate for a calibre of device that Bolt is (both quality and price-wise). Now, I wonder what would be an effective way to DIY-fix the situation with CN7:2, making it PWM capable at 25kHz? It doesn't seem to me that a simple low pull-down resistor would be all it takes to remedy this?! Would it be possible, at least in theory, to fix signal by removing or replacing some or all of capacitors I mentioned?
The circuit is obviously designed to drive a 3 pin fan with PWM. The capacitors will then have little effect because the current taken by the fan motor will discharge them quickly. Why it doesn't work with the COFAN I have no idea until I get home and make some tests. It would work with most 3 pin fans I have come across because when you turn the fan off it slows down, it doesn't stop instantly, so when you turn it on again it starts to speed up. If you do that fast enough the rotor inertia averages out the speed. You may hear some buzzing though. I recently designed a circuit to control the speed of a tiny 17mm 5V fan. I used a PWM output on an RPI0 and a transistor wired as a Miller integrator, so it turned the PWM into a steady voltage, so doesn't add any noise. The 5V fan works down to about 2V. Still quite noisy at any speed though, the problem with small fans that have to spin very fast to move any air. When driving a 4 pin fan the PWM pin is a logic signal and will not draw enough current to discharge the capacitors quickly. Adding a 220R pulldown will give a time constant of about 44us with 200nF so will be able to PWM up to around 20kHz.
Cool, thanks for your insight, I see that pull-down may need a rather small resistance for time-sensitive stuff... I will actually try inserting ~200R into the connector and see if the signal gets any better, just if I find such small resistor lying around.
Finally decided to mount a Noctua NF-A6x25 5V PWM on the outside of the case while leaving the stock fan inside. The rubber plugs that hold the fan don't line up perfectly with the slots on the case but it's close enough to make the lock into place. The external fan runs quiet without the internal fan unless the temperature rises 55 degree using these settings: For the Internal Fan Settings: AC0: [70] AC1: [55] Temperature Hysteresis: 4 Fan Duty Cycle (%) Above AC1: 80 Speed change duration: 3 For the External Fan Settings: AC0: [70] AC1: [40] Temperature Hysteresis: 4 Fan Duty Cycle (%) Above AC1: 50 Speed change duration: 3
Nice. I'm currently using ONLY internal fan, with settings, being DISABLED, 25%, 25%, 1. This gives almost non audible constant rotation and low temperature around 50C (idle) to 60C (maxload). p.s. don't use Ring https://www.eff.org/deeplinks/2020/01/ring-doorbell-app-packed-third-party-trackers
I was getting a nuisance buzzing sound with the internal fan at low speed. So opted with external mounting solution which cools the CPU to 39-43 degrees. Good to know about Ring!
Thanks @Zgembo! I'm running two Noctua NF-A4x10 5V (non-PWM) fans in parallel off CN7 in a very cramped case. I had assumed from documentation that CN7 would supply a smoothed DC V output and have been frustrated trying to get smooth fan control. I can get a single low speed at anything from 50%-97% duty cycle, slightly faster with annoying coil whine at 98% and 100% speed at 99-100%. At any duty cycle they suddenly speed up to 100% for about 1/2 second every few seconds. This explains it perfectly! I was thinking of putting an RC low-pass filter across pins 1,2. Do you think that would help?
@ajs It might, but I just never had time to check myself. If you happen to check it out yourself, please report here...
Sorry to revive this. Just got my Bolt and the noisy fan issues are just frustrating. Got a 3 pin 5v fan and to my dismay it looks like I'll need a 4pin 5v fan. I've tried to use the Noctua site to map the pins (https://noctua.at/en/what-pin-configuration-do-noctua-products-use). Am I correct in saying: 5V to 19 GND 17 PMW 11 RPM Speed ?12? (unsure about this one)
Heya, hopefully you don't still need this info given it's been nearly two months, but for anyone passing by.... 5v from the *Arduino* header (the upper one) from pin 19, 39 or 40. (Pick any one) GND from *Arduino* header also, pin 15, 17, 33 or 38 (Pick any one) PMW signal from *Embedded controller* header (the lower one) pin 11. RPM speed / fan tach from the *Embedded controller* header from pin 12. (So you had it all correct, just making sure to note that the pin numbers are from two distinct headers
Sorry for the late note: the stock fan I received (which has no Green Dot on the side of fan body) which came with the Kickstarter campaign, is a COFAN model F-6010M05BII. It has a 3pin (on a 1.25 pitch Molex connector): Black (Gnd), Red (fan motor voltage), and Yellow (tachometer fan speed output) and is a 5V 60mm low-profile fan. Its power input can take in voltage as low as 2.5V (consumes only 100mA) for low RPM and as high as 5V (consumes 225mA) for high RPM. Actually when I run it off the heatsink, it is fairly quiet. I guess the mechanical coupling and heatsink air restriction makes it louder. What type of fan motor is it? It is a very basic motor. You give it a higher supply voltage, it would spin faster. What the BOLT mother board does is to use several transistors: one logic level voltage shifter, one preamp, and one last-stage power MOSFET motor driver, to convert the PWM signal coming from the MEC1705 Embedded Controller, into variable motor power supply voltage. P.S. why the regular, integrated PWM fan won't work? It is due to itself has its own PWM-to-motor drive circuit inside. Thus it won't take the variable power input voltage well. P.S.2, use revised firmware to enable PWM, and also matching the "pwm duty cycle" to the "fan speed" profile.
Hi All, i just finished my Bolt Hardware-Setup, using a lot of inspirations from this thread and others in the forum: + replaced the not really working Green-Dot-Fan with a working one (which was kindly provided by Udoo, many thanks again!) + replaced the (non working) external power button with another 12mm one which includes a power LED + replaced the Wifi- and Bluetoth antennas with external case-mountable ones + added a secondary fan (Noctua 8cm 5V 4Pin-PWM) using the wiring +5V and GND from one of the Grove Connectors, Tach and PWM from the Embedded Controller I/O Header as already mentioned in this thread + added a simple blue 5V-LED for the drive-lights (from the Buttons/LED header), mounted it below the power bolt symbol of the metal case. Some confirmations from my side: - The Noctua fan indeed works, and very quiet even when active (unnoticable in my environment) - The replacement Cofan-Fan works too, and is much quieter than the gren dot one - BIOS settings and temprerature control for both fans work fine (current BIOS version, 1.08) I did not test much on the software side, but so far everything seems to work well: Installed Xubuntu 22.04 LTS, Boot on eMMC, Root/ on an Transcend Sata-SSD in the 2260 slot, some data on an Samsung NVME-SSD in the 2280 slot. Added an external Sata SSD i had lieing around just to check if it works (what ist does). Getting the audio to work took some time, but using an optical S/PDIF cable that is working too. So, just want to say i am happy with the Bolt: great little device, great community and superbe customer service from Udoo!
