Q=C*U? where Q is constant, but C decreases (e.g 10%) under the influence of increasing U -> making U even higher (e.g. 10 times higher).
I didn't get that bit...
Indeed, with L101 about 1.5R or so supply?DC-resistance, the 1ms software control loop is way too slow and should be in the order of 4ns response time to correct the voltage in C107, difficult to achieve in a ADC polling fashion so this is probably a NO GO as it would have a serious CPU cost.
Yes the 1ms control loop isn't a cycle by cycle correctionams this is why it can have trouble with sudden large changes.?
A simple workaround could be to DELAY the use of switching regulators after powering up. This allows the USB PD device to do the voltage jumps just while only the LDO is active, and limits the sensitivity in all these control loops that are anyway too slow to resolve this issue.
The switching regulators are already slow to start up, around 250ms. How long does a PD device need? I could add a configurable additional delay.
As soon as there is a voltage drop (due to a bad cable) you could redo the complete power-up procedure to also be immune in that circumstance.
No because you'd have to detect that and react to it. Same problem.
My view on this is that all effort you put in making the control loops faster will eventually still insufficient to deal with this USB PD voltage steps.
If PD start up somewhat faster than 250ms then there isn't any problem anyway.?
In the history of QRP Labs there have always been cases of people using power supplies that don't go too well even back to the Ultimate2 in 2014. Supplies with unexpected spikes etc. It isn't surprising that we've found with QMX a couple of people who've had trouble with a particular powering arrangement. Such as the 6V to 12V step. Is it indicative of a general weakness? No. Should it be something that if possible gets improved so QMX gets more resilient?
The zeners are in place for dealing with unexpected power excursions. However as demonstrated by Gunnar they can't manage a step 6V to 12V. Even after I had improved the response speed drastically for supply steps in firmware 1_00_009.
So my question is still... Wouldn't just making the zener diode bigger fix the sensitivity to this unusual supply scenario? In my tests, with the standard circuit (500mW zeners) I was able to do 6V to 9V, 10V, 11V steps without issue and the zener ate the brief over-voltage. It even did on 6 to 12V steps but not every time. So given that it almost works even with the 500mW zeners isn't it worth trying a 5W zener say, being optimistic that it may be able to swallow that extra current for a millisecond? Then if QMX has a specified supply voltage range of 7V to 12V say, you'd be able to do anything at all with the supply voltage in that range without any fear of damage.
