Someone with a theory at apple discussions.
http://discussions.apple.com/thread.jspa?threadID=1852329&start=105&tstart=0
darrylh
Posts: 2
From: MI
Registered: Jun 21, 2009
Re: screen flicker resolved?
Posted: Jun 21, 2009 7:45 AM in response to: Jeff Storm
Click to reply to this topic Reply email Email
Hopefully I can shed some light (no pun intended) on the backlight flickering issue at low brightness. I have designed fpga controllers for LCD panels, and the issue is a combination hardware/firmware issue.
I purchased a uMBP in October and never saw any backlight flickering at low brightness with either GPU. I just received a newer uMBP and it shows the backlight flicker problem only on the 9400. It was immediately obvious what the problem is. The LCD panel has an underlying repetition rate as does the LED backlight controller since it uses PWM techniques to dim the lamps. Apple did not provide enough filtering on the LED PWM controller so the LED's are actually slightly flickering on their own at the rate of their PWM controller. This is not a problem on it's own as the PWM frequency is always high enough that the human eye can not perceive it. The problem is when this frequency or it's harmonics are close a frequency of the LCD, patterns will be created in the light intensity called beat patterns. Since these beat patterns are dependent on the absolute frequency of the LCD and LED PWM, and since each GPU and the LED PWM are sourced from different clocks, the issue will be different for every motherboard, and will vary with such things as voltage and temperature.
I verified my suspicions by slightly shifting the operating frequency of the LCD (not something anyone can do) and the problem went completely away. As I continued to change the frequency the problem would reoccur and subside as the two fundamental frequencies would beat against each other.
The solution would have been to provide better filtering of the PWM supply, but this is a typical trade off that even I would have made when designing a PWM dimming circuit. Also, typical white LED's have a fairly slow response time, so this was probably relied upon for some of the filtering. The other solution is to just shift the frequency of the LCD or LED PWM so that they don't beat. The problem here is again with the absolute frequency (over time, temp and voltage) on each motherboard.
So the bottom line is an LCD replacement will not affect this issue at all (although it may address some of the other issued raised in this thread), and a motherboard replacement is a hit or miss proposition at best, (unless Apple has made a hardware changed, but based on my new uMBP, they haven't) and even if it does appear to fix it at first, it may return later. My guess is (and this is only a guess) a firmware change could be made to shift the clocks to alleviate this problem. This is something I have done to my firmware designs, but is not always possible for other engineering reasons. Certainly the LCD clock rate can be shifted easily, but what complicates the solution is it needs to be universal across all notebooks and that can be a difficult thing to achieve.
So can anyone confirm that they have spoken to someone at Apple that has confirmed they know of the issue and are actively seeking a solution? Without this confirmation, I don't intend to send my MBP back as it can be a futile exercise as pointed out by others in this thread.
MacBook Pro Mac OS X (10.5.7)