I notice the data sheet for the AMS1117-5.0 is
different from the data sheet for the LM1117-5.0. The
manual calls out an LM1117-5, IC11 on page 8 in the
parts list but the board may be using an AMS1117-5.0
which are not the same chip design and have different
requirements.
It seemed to me likely that AMS1117 and LM1117 were
supposed to be functionally equivalent, in much the same way
as many other parts are, which are produced by different
manufacturers who prefix their own letters...
But to make things even more confusing, there are at
least two different versions (refer to Google) of AMS1117
datasheet, which also contradict each other in certain
areas...
The actual problem is the quality of the OUTPUT
capacitor. The regulator chip is likely unstable and
going into oscillation causing that big spike seen on
the input. The AMS1117-5 calls for a 22uF solid
tantalum and the LM1117-0S calls for a minimum 10uF but
low ESR type with a specified ESR between 0.3 and 22
ohms for the OUTPUT capacitor.
I'm not totally convinced yet. The AMS1117 datasheet says
in the paragraph on Stability, that the output capacitor
22uF covers all operating conditions, including the use of
bypass capacitors across the potential divider which feeds
the Adjust input. It goes on to say that if there are no
bypass capacitors then smaller capacitors can be used with
equally good results. The AMS1117-5.0 has no bypass
capacitors because it has no Adjust input, it is a fixed
regulator. Therefore it follows that smaller capacitors
ought to be fine.
Additionally note that there is a 470uF capacitor (C47)
hanging on the 5V rail, admittedly via 47uH L6 which will
change things of course.
I did try additional capacitors at the output, and I tried
removing the 470uF capacitor completely, and nothing made
any difference to the spikes that were seen at power-up. A
10uF electrolytic or 10uF tantalum (both were tried) solved
it extremely comprehensively. I tested both with extreme
intermittent power connections, lots of
connecting/disconnecting over extended time, and nothing I
could do could fry the system. When I removed the input
capacitor, my 20m QCX-mini got smoked within seconds.
I did also remove a 10uF SMD capacitor and measure it; it
was a bit on the low side but these things have quite wide
tolerances anyway and my measurement method (time decay
across a resistance) wasn't carried out particularly
carefully - so I felt that within reason the capacitor
actually IS a 10uF capacitor. But the quality, ESR, type etc
is a different question.
Anyway that was my night time's work from midnight until
4:30am this morning - except for a few nappy-changing
(a.k.a. diaper) interruptions - and in the end, I had
managed to convince myself that spikes were seen at the
input, with the stock capacitor; I could kill a QCX-mini by
application of power intermittently; and by adding a 10uF
electrolytic or tantalum, I could comprehensively cure the
issue such that on both my 5A and 20A bench supplies, no
amount of messing about would kill the QCX-mini.
It goes without saying that in all the prototype QCX-mini
units I built, and the three production units from the first
batch, and in all the other times I have used the identical
AMS1117-5.0 chip in an identical way from the exact same
supplier etc., I had never seen this failure mode.
