Introduction: Fixing Clicking Noise Problem on Apple 27" Display
Ever have one of your beloved display start to make a lot of noise when you are using it? This seem to happen after the display has been in use for several years. I debugged one of the display thinking there was a bug trapped in the cooling fan, but it turn out the root of the failure is a lot more complicated.
Step 1: Power Supply Design Overview
Here is the instruction on how to identify and fix the clicking noise problem experienced on certain model of the Apple Thunderbolt display and IMac computer.
The symptom is usually a fairly annoying noise coming from the display that sound like crashing leaves. The noise usually comes on after the display has been in use for a while. The problem tend to go away after the machine is unplugged for a few hours but will come back in minutes after using the device. The problem does not go away if the machine is put into suspend state without being unplugged.
The source of the issue is caused by the power supply board as I'll try to walk though the process of identify the issue. With enough knowledge, it is an issue that can be fixed for a few dollars worth of components.
WARNING!!! HIGH VOLTAGE!!! WARNING!!! DANGER!!!
Working on power supply unit is potentially dangerous. Lethal voltage exists on the board even after the device has been unplugged. Only attempt this fix if you are trained in handling high voltage system. Use of isolation transformer is REQUIRED to prevent ground short. Energy storage capacitor takes up to five minutes to discharge. MAKE MEASUREMENT OF THE CAPACITOR BEFORE WORKING ON THE CIRCUIT!
WARNING!!! HIGH VOLTAGE!!!
The design of the most majority of the Apple display's power supply module is a two stage power converter. The first stage is a pre-regulator that convert input AC power into a high voltage DC power. The AC input voltage can be anywhere between 100V to 240V AC. The output of this pre-regulator is usually anywhere from 360V to 400V DC. The second stage converts the high voltage DC down to the digital voltage supply for the computer and displays, usually from 5~20V. For the Thunderbolt display, there are three outputs: 24.5V for laptop charging. 16.5-18.5V for LED backlight and 12V for digital logic.
The pre-regulator is used mainly for power factor correction. For low end power supply design, a simple bridge rectifier is used to convert the input AC to DC. This cause high peak current and poor power factor. Power factor correction circuit correct this by drawing a sinusoidal current waveform. Often, power company will place a restriction on how low the power factor a device is allowed to draw from the power line. Poor power factor incur extra loss on the power company's equipment hence is a cost to the power company.
This pre-regulator is the source of the noise. If you disassemble the display till you can extract the power supply board, you will see there are two power transformer. One of the transformer is for the pre-regulator while the other transformer is the high to low voltage converter.
Step 2: Problem Overview
The design of the power factor correction circuit is base off the controller produced by ON Semiconductor. The part number is NCP1605. The design is base on boost mode DC-DC power converter. The input voltage is a rectified sine wave instead of smooth DC voltage. The output for this particular power supply design is determined to be 400V. The bulk energy storage capacitor consists of three 65uF 450V capacitors running at 400V.
WARNING: DISCHARGE THESE CAPACITORS PRIOR TO WORKING ON THE CIRCUIT!
The problem I observed is the current being drawn by the boost converter is no longer sinusoidal. For some reason, the converter shuts off at random interval. This leads to inconsistent current being drawn from the socket. The interval where shutdown occurs is random, and is below 20kHz. This is the source of the noise you hear. If you have an AC current probe, connect the probe to the device and you should be able to see the current draw by the device isn't smooth. When this happens, the display unit draw a current waveform with large harmonic components. I'm sure the power company is not happy with this kind of power factor. The power factor correction circuit, instead of being here to improve the power factor, is actually causing a bad current flow where large current is being drawn in very narrow pulses. Overall, the display sounds awful and the power noise it throw into the power line will make any electrical engineer cringe. The extra stress it places on the power components probably will cause the display to fail in the near future.
Combing though the datasheet for NCP1605, it appears there are multiple ways the chip's output can be disabled. Measuring the waveform around the system, it become obvious one of the protection circuit is kicking in. The result is boost converter being shut off in random timing.
Step 3: Identify the Exact Component That Cause the Issue
To identify the exact root cause of the issue, three voltage measurements should be performed.
The first measurement is the voltage of the energy storage capacitor. This voltage should be around 400V +/- 5V. If this voltage is too high or low, the FB voltage divider is drifted out of spec.
The second measurement is the FB (Feed back) pin's voltage (Pin 4) with respect to the (-) node of the capacitor. The voltage should be at 2.5V
The third measurement is the OVP (Over voltage protection) pin's voltage (Pin 14) with respect to the (-) node of the capacitor. The voltage should be at 2.25V
WARNING, all the measurement nodes contain high voltage. Isolation transformer should be used for protection
If the voltage of the OVP pin is at 2.5V, the noise will be generated.
Why does this happen?
The power supply design contains three voltage dividers. The first divider sample the input AC voltage, which is at 120V RMS. This divider is unlikely to fail due to the lower peak voltage and it is consists of 4 resistors. The next two dividers sample the output voltage (400V), each of these dividers consists of 3x 3.3M ohm resistors in series, forming a 9.9MOhm resistor that convert the voltage from 400V down to 2.5V for FB pin, and 2.25V for the OVP pin.
The low side of the divider for FB pin contains an effective 62K ohm resistor and a 56K ohm resistor for the OVP pin. The FP voltage divider is located on the other side of the board, probably partially covered by some silicone glue for the capacitor. Unfortunately, I don't have a detail picture of the FB resistors.
The trouble occurred when the 9.9M Ohm resistor start to drift. If the OVP trips under normal operation, the output of the boost converter will turn off, resulting in sudden halt of the input current.
Another possibility is the FB resistor start to drift, this can result in output voltage start to creep above 400V, till the OVP trip or damage to the secondary DC-DC converter.
Now comes the fix.
The fix involve replacement of the defective resistors. It is best to replace the resistors for both the OVP and the FP voltage divider. These are the 3x 3.3M resistors. The resistor you use should be 1% surface mount resistor size 1206.
Make sure you clean the flux left over from the solder as with the voltage applied, the flux can act as a conductor and reduce the effective resistance.
Step 4: Why Did This Fail?
The reason this circuit failed after some time is due to the high voltage applied to these resistors.
The boost converter is on all the time, even if the display/computer is not being used. Thus, as the way it is designed, there will be 400V applied to the 3 series resistors. Calculation suggest 133V is applied to each of the resistors. Maximum working voltage suggested by the Yaego 1206 chip resistor data sheet is 200V Thus, the designed voltage is quite close to the maximum working voltage these resistors are meant to handle. The stress on the material of the resistor must be great. The stress from the high voltage field might have accelerate the rate the material deteriorates by promoting particle movement. This is my own conjuncture. Only a detailed analysis of the failed resistors by a material scientist will fully understand why it failed. In my opinion, using 4 series resistors instead of 3 will reduce the stress on each resistor and extend the life of the device.
Hope you enjoyed this tutorial on how to fix the Apple Thunderbolt display. Please extend the life of the device you own already so less of them end up in the landfill.