Some folks who have the clock have noticed that the PM light will sometimes flicker slightly, or in 24 hour mode, the 10s-of-hours segments other than B and C (those that form a "1") will flicker. The flickering is in cadence with the 10th-of-a-second digit. This is, so far as I can tell, consistent across the entire manufacturing run - it's not just on some of them.

It's a mystery to me why the 10s-of-hours segments are affected, but the PM digit makes sense. From a wiring perspective, the PM LED is actually the decimal point for the 10th-of-a-second digit. When the 10th of a seconds change from 0 to 1 to 2 you go from 6 segments to 2 to 5 lit. When the digit changes from 6 to 7 to 8, you go from 6 segments to 3 to 7. Each segment is pumped with 40 mA when it's on (keep in mind that the duty cycle is quite low because of the display rastering), so 4 segments' worth of current is 160 mA. The MAX6951CEE has a current regulator in it, but there's probably a limit to how effective it can be.

The 10s of hours digit is number 0. The 10th of a second digit is number 6. There is a virtual digit number 7 which supplies only the AM light. The rastering sequence is sequential, so there's a gap of one digit between the 10th of a second and the 10s of hours digit. That's why the flickering of that digit doesn't make any sense to me.

If I look at the waveforms on the scope, I do see variations in voltage that run in a 1 Hz cycle. What's more, disabling the 10th digit in firmware makes the problem go away. Sort of. What actually happens is that what variations there are are stable for 1 second at a time. That makes whatever brightness variations result get lost in the changing digits.

I've tried increasing the bypass capacitance on the MAX chip, but doubling it didn't make a difference. I can't do more than that because there's a limit on how much capacitance the buck converter can supply before it has problems with startup. If you do the calculations, the trace width of the power bus only contributes 4 mV or so of drop at 150 mA. I'm seeing variations on the supply bus of more like 40 mV - still well inside 2% regulation - fine for digital logic. Those variations are in cadence with the display raster, so clearly they're directly related to the LED draw. I built a special clock for my wife with green LEDs, and it shows this problem as well, so it's not just the choice(s) of LEDs.

I've decided to add a menu item to allow you to disable the 10th of a second digit entirely. For me, I prefer to have the 10th digit and put up with a tiny bit of flicker, but if the flicker is bothersome, the only way to get rid of it that I've found is to not have it.