Overclock.net › Forums › Cooling › Other Cooling Discussions › Corsair H100 / H80 / H60 noise, grinding pump - *fix* + official Corsair response
New Posts  All Forums:Forum Nav:

Corsair H100 / H80 / H60 noise, grinding pump - *fix* + official Corsair response - Page 16

post #151 of 1734
First up I'd like to say thanks to wlw wl for all the hard work and diligence on this thread, its really rare to find someone that's willing to go the extra distance like this, especially for no personal gain.

I recently bought a Corsair H100 cooler, and all was great for about half a day, then it started the rattling hard drive-ish noises. Its slowly been getting louder. So I checked around online to see what fixes if any where to be had. Very quickly I stumbled upon wlw wl's diode based voltage dropper and got excited. However, installing the diode jumper I made, only makes the sound faster. While looking at the CPU fan RPM monitor the pump does indeed drop from 2050-2200 down to 1880-1960ish in RPM so the diode does its intended job. But the noise gets worse. I'm curious what I should attempt next. Leaning the case to the side makes the noise nearly vanish, however I don't have the room to lay the case on its side in my current location. I've tried tapping the hoses and the water block/pump itself doesn't seem to help. I have two Noctua NF-P12 fans mounted to the radiator instead of the corsair brand fans. I never actually tried the corsair brand fans. Could that make a difference? By the way my batch number is 11529403.
post #152 of 1734
First off, The 1N4001 fix has worked for me. I had a huge problem with the rattling of the pump. It would come and go (mostly always there). Rattling the hoses seemed to help but it would eventually return. A quick stop at "The Source" (Canadian Radio Shack) and a few minutes later the sound is gone!

Larhleolynx, I am also running Noctua fans but I'm using NF-F12 fans. They are nicely silent and still give excellent air flow with good pressure. The P12's shouldn't be an issue but the NF-F12 fans are made for radiator applications.
post #153 of 1734
Thread Starter 
Larhleolynx - can you be bothered to take the H100 out of the case?

greymonk - another soul saved biggrin.gif
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
post #154 of 1734
I can do that, I don't have any spare thermal paste though. Would take about 3-7 days to get some delivered.
post #155 of 1734
Thread Starter 
It might be a good idea to take it out, place the radiator above the pump vertically so the hoses go from the pump straight up to the rad, turn on the unit and shake it a bit and let it run for some time - that should move any leftover air from the pump to the rad and that might help. Surely it won't hurt...

The pump is very weak and it might not be able to move the air on its own, if there is any air inside that is.
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
post #156 of 1734
Thread Starter 
I thought it would be nice to have everything in one place, so here follow my thoughts and ideas about what I found inside H100, this applies to H80 and H60 too, the latter excluding the fan controller part, of course.

This is a collections of various posts that I wrote.
About H60 incorrectly reporting its speed as 4000 RPM (Click to show)
Quote:
It's not a bug, it's a feature biggrin.gif

H60 has the RPM signal coming straight from the motor controller, H80 and H100 has that signal routed through the fan controller's PCB (not present in the H60), so there might be some additional filtering present in the H80/H100's controller IC (which validates my previous statement of the signal on H60 being noisy or something like that).

I've personally seen H100's speed reported as 4000, so that just happens to all of them, but they all run at the 2000-ish RPM, end of story.
Again about why H60 reports its speed incorrectly as 4000 while the H100 an H80 don't - boring details (Click to show)
Quote:
The description of the FG output pin is rather clear - "rotation count detection function can be implemented using this output, which corresponds to the phase switching". Maybe this isn't crystal clear at first glance, but when you look at the block diagram of the LB11961 driver, it's clearly shown that the FG output is just an amplified signal from the Hall sensor.

The stator has 4 poles and two phases, just like this one: http://en.wikipedia.org/wiki/File:Poles.jpg
This means that the magnet on the rotor has 4 poles ("main" poles, there might be another 8 poles superimposed, to eliminate 0 torque positions, these don't matter), each taking 90 degrees of the circle, SNSN for full circle. Now let's assume that the Hall sensor goes high (H) when it's next to N, and low (L) when it's next to S (or the opposite, doesn't matter). Then, a full rotation of the rotor means that the Hall sensor "sees" SNSN, so its output states are LHLH per one rotation.

So there you have it, there are two "ticks" (depends on the sensing circuit which transition - rising edge [L->H] or falling edge [H->L] - counts as a "tick") per one full rotation of the impeller, therefore the frequency on the FG output is double the rotation count. An that is why they wrote that a function should be implemented for this, in our case to divide it by two. However the raw signal is passed to the CPU fan header, so you see 4000 RPM where it should be 2000 RPM.

I don't think I've said that the signal isn't noisy, or that it is, I don't know that. You'd have to hook it up to an oscilloscope, but my guess is that the signal isn't symmetrical - either one of two falling or rising edges is less pronounced, or every other H is lower then the one before or something like that.
If so, one sensing circuit will trigger on (count) both "ticks", while another might only trigger on only one of every two edges, effectively dividing the frequency by two.
About the pump's motor and its driver (Click to show)
Quote:
Truth is that the motor used in the Hydro pumps is just like the one you can find in a PC fan, a bipolar BLDC.
As we both now, the motor is driven by a LB11961 single-phase bipolar motor driver (designed to drive PC fans, BTW). This driver has built-in speed sensing and variable speed control based on a thermistor (more on that later). Speed detection can be done by either Hall sensors or back EMF sensing. PC fans use Hall sensors. EMF sensing does not require additional components, but is harder to handle and the engine has to be started "blindly". As far as I can tell, Hydro units use Hall sensor(s) too.

Now the interesting part - the driver can work in a variable speed control mode with a thermistor or at full speed always. Unfortunately the latter is the case with Hydro units, the pump goes as fast as it can, the higher the voltage, the faster - hence problems with grinding are more common on >12V PSUs and that's why the voltage dropper or a fan controller usually works.

The grinding. I think that actually that can be fixed easily by changing the PWM base frequency, which is adjusted by the C2 capacitor on the driver's PCB. Normally that's 100pF which sets the clock to 25kHz. Setting it lower isn't desirable due to the proximity of acoustic range, so we're talking about increasing it a bit, but that's a matter of trial to see if it stops resonating at say 30kHz. So changing one capacitor might be all it takes to cure these units. I don't have a sample to experiment on, which I would gladly do, so I can't confirm that in practice - but it would be the first thing to try, because it's so easy.

Another solution would be to use the feature built into the driver, which is low speed setting. It could be used to cheat a bit, set the so called "low speed" to say 1900 or 2000RPM and simulate the thermistor with a resistor, so that the driver never goes into the "high temperature - full speed" mode. I'd personally use a precision potentiometer in place of aforementioned resistor, which would allow for fine tuning of the pump's rotational speed with the driver in variable speed control mode by simulating changing coolant temperature.
Last but not least, I would add a 12V Zener diode on the input, to cap the power supply voltage to 12,00V. Cheap and effective solution. Perhaps in conjunction with the PWM frequency change, that would put all the problems to the rest.

The speed sensing that goes out to the CPU header comes right off the FG pin of the driver (they even marked the pad for the third wire "FG" on the PCB). This pin changes state each time a phase is switched, so it "ticks" twice per each rotation of the impeller with this motor and driver, that's why H60 reads in BIOS as 4000RPM while the actual speed is 2000RPM. Again, not a bug, it's how it's supposed to work. Like I've said before, on H80 and H100 this signal is routed through the fan controller's board IC, where a hardware divisor might be easily implemented to divide that by two, and probably is.
This should have been implemented on the BLDC driver PCB (which is the same in H60, H80 and H100) because that's basically two transistors (bistable multivibrator) or a little IC.
About the stupid voltage regulator IC of the fan controller's circuit (Click to show)
Quote:
I'll also add that on H80 and H100 the fan controller's circuit is being powered by a
National Semiconductor LP2992AILDX-3.3 Micropower 250 mA Low-Noise Ultra Low-Dropout 3,3V Voltage Regulator (package marking M13AC L014A B).

I believe this is the main cause of the controllers suddenly dying for no apparent reason (no lights on startup and ever again) because this regulator is meant for other applications (battery powered mobile appliances), in this circuit it has to work with relatively high input voltage (+12V) which means that it has to disperse moderate amounts of heat. The IC has its own thermal safety and there is a heat sink polygon on the PCB, but still the IC isn't well-suited for this application is in my opinion it's the weakest link of the design.
There is no point in using a low-dropout regulator when there is such a big difference between the input voltage and desired output voltage.
There is plenty to choose from, for example Texas Instruments UA78M33 in TO-252 or SOT-223 packages, both of which would fit this application much better.

Also I think that the fourth LED, the little human figure, should be connected directly to the voltage regulator's output and no to the controller IC as it is now. Then it would be clearly visible which part of the circuit died - if the control IC would break, only the 3 profile LEDS wouldn't work, but the fourth "fella" LED would still be on. Now, when the "fella" is connected to the control IC, it will malfunction in both the control IC fault and the voltage regulator fault scenarios, effectively being of no use at all as at least one profile LED is always on when the unit is powered and working, so the fourth LED serves none but a decorative purpose.
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
post #157 of 1734
Ok! I've went and placed an order to get some new thermal paste, and I'll try your suggestion once it arrives. I'm curious if anyone has any experience with Intel's Ivy Bridge CPUs and the H100 yet? I haven't tried running my 3570K with the stock Intel cooler, and I can't overclock the CPU because I'm using an older Gigabyte Z68 motherboard. I went straight with the Corsair H100, I'm mostly wondering if these are temperatures I should be seeing with a properly seated and functioning H100 cooler. (stock voltage, stock SpeedStep clocks)

Currently at idle the cores are as follows:
Core #0 29-30 C
Core #1 25-27 C
Core #2 36-38 C
Core #3 26-28 C
Package 34-35 C

Under Prime95:
Core #0 48-49 C
Core #1 52-53 C
Core #2 54-55 C
Core #3 45-46 C
Package 57-58 C
post #158 of 1734
wlw, did you post this on the Corsair thread? If not, you should, this is very informative and helps a lot.
post #159 of 1734
Thread Starter 
Larhleolynx - these temps are similar to what I saw with H100 on i7-2600k. There are some posts by people with Ivy over at Corsair forum.

bwebmasta - yes I took it from there, but not the big "H100 Pump noise" thread, but the "H series pump speeds" thread here:
http://forum.corsair.com/v3/showthread.php?t=106031
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
600
(16 items)
 
Adapters
(9 photos)
My custom water
(4 photos)
CPUMotherboardGraphicsRAM
AMD Phenom II x4 955 BE @ 3,8 GHz ASRock 870 Extreme3 Sapphire Radeon 6850 1GB GeIL 1600 CL8 2x2GB 
Hard DriveHard DriveCoolingOS
Corsair Force GT Samsung Spinpoint F3 Custom water. Very custom. Windows 7 x64 Pro 
MonitorKeyboardPowerCase
Philips 232EL natec Piranha 2 gaming :) OCZ ModXStream Pro 600W Corsair Graphite 600T 
MouseMouse PadAudioOther
A4 X-748K A4 X7 500-MP Asus Xonar D1 Roccat Kave 
  hide details  
Reply
post #160 of 1734
Larhleolynx- you might want to try what I did when I had the same issue, and try putting 2 diodes in series rather than 1. Mine is silent now and a little over 1700 rpm.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Other Cooling Discussions
Overclock.net › Forums › Cooling › Other Cooling Discussions › Corsair H100 / H80 / H60 noise, grinding pump - *fix* + official Corsair response