s3c2440(2410) Causes and solutions for USB HOST instability

Today, tpu took out a 2440 board for debugging and found that despite various measures, USBHOST would occasionally not work. UCLK was brought out through CLKOUT0 and checked with an oscilloscope. It was found that when it did not work, UCLK was not stable at all. So after careful consideration, there are several places that affect USB:

1. External crystal oscillator 16.9344Mhz

2. UPLL LOCKTIME of LOCKTIME register

3. UPLLCON

4. UPLL switch of CLKSLOW register

Next, exclude them one by one:

1. The crystal oscillator is shared by MPLL and UPLL. I have never heard of MPLL being unstable.

2. I suspect LOCKTIME is too large (0xffff), but reducing it has no effect.

3. When UCLK is unstable, it is useless to repeatedly set UPLLCON.

4. When UCLK is unstable, switching UPLL is effective! When UCLK is stable, switching UPLL will also cause instability.

      That is to say, UPLL may fail to synchronize when it starts. But why? UPLL and MPLL should have the same structure, and MPLL never has problems. UPLL just has a lower frequency than MPLL. Let's look for it from the frequency.

      Now set a relatively low frequency (such as 7Mhz) for UPLLCON. At this time, no matter how UPLL is turned on and off, UCLK is still not stable. Well, the dawn has appeared! Then set an ultra-high frequency (112Mhz). Haha, no matter how UPLL is turned on and off, or powered on, UCLK is very stable.

       The problem can be described as follows: UPLL cannot be reliably started at low frequency.
       Solution to the problem: First give UPLL a higher frequency, and then set the required frequency after it works stably.
       Note: If you want to turn on and off the UPLL switch in CLKSLOW, you must also follow this setting procedure.

       After this modification, the TPU switched on and off the power supply for more than 100 times, and UCLK was always very stable. This is only the case for 2440, but it should also apply to 2410. You can actually test and verify it.