开云体育

Well... I just finished my second QMX, the first one was toasted by increasing the supply voltage from 6V to 12V wiith a ramp-up time of 3milliseconds.
The full story is here:?/g/QRPLabs/message/106237

That one was? running firmware 1.07. Image. D109 was destroyed (short) so I removed it.
After that my QMX turned on but power? consumption was200mA @12V input.
Image rejection was nonexistant and bandpass filters totally wrong, so I concluded IC402 and/or IC403 were damaged.
I decided to stop there and ordered a new kit

Now the second QMX, finished last night:
I loaded firmware 1.09 without problems.
Everything was fine and I was just about to write something on the forum of a successful build.
I wanted to put in some figures of RF power output and current consumption.
Power output was fine 4W on all bands @12V supply. Image rejection and filter sweeps were also fine. Supply current about 100mA in receive and 600mA in transmit
But while I was measureing supply current I heard a "crack" and the QMX turned itself off.
I turned it on again by pressing the left rotary switch.
Now I smelled burnt electronics.
I opened the case and out came the holy smoke...
C107 was now black and with a large crack in it.
I took out the power board and removed C107 and measured the resistance between Vcc and GND.
733 Ohms in both directions with the DMM, so i removed D108. Now I had 75 kOhm in one diection and about 7 kOhm in the other.
I also checked Q108, no short between terminals, so I guess it survived.

So I took the 5V power board from my first QMX and put it into the second.
And voila, everything back to normal.

So what happened?
During the current measurement It is possible that I interrupted the power supply line shortly when I disconnected the DMM without powering down the QMX before.
But htis case was addressed by Hans in his article about SMPS behaviour during input voltage step...

There is a saying in swedish "en g?ng ?r ingen g?ng, tv? g?nger ?r en gong-gong"
Free translation: One time is OK, two times is an alarm clock.
I can see that several QMXs have blown up C107.
I think this is no coincidence. There is something going on here
I recall that all who reported blown C107 were running 12V supplies with high current capability

I suspect a HW/SW interaction:
During a power interruption glitch, is it possible that the processor can leave the the PWM signal high, thus creating a voltage increase that kills C107 and D108?
I would suggest that the PWM-signals should be AC-coupled so that if stuck high, the SMPS would turn itself off.

73 de SM5EIE /Gunnar
?












































So I ordered?

Hi Gunnar

I noticed that C107 was getting warm at a supply voltage of 6v, and that D108 ceased working as a zener. C107 is only rated at 6.3v according to the parts list...which is OK til something goes wrong and it gets the full supply voltage across it. Tantalums always put on a good show!?

I had no PWM5V from the processor in my case and Q106 was full on, causing? the "full 6v" to appear across C107...at 13.8v I cant imagine.

73 Steve G4EDG

Gunnar,
I am beginning to think that C107 has too low a rating (6.3 V), and it will occasionally self destruct for no apparent reason. Unfortunately, there isn't a lot of room for alternative pieces.?
--
73, Dan? NM3A

Gunnar,?
If this was a high voltage HW/SW glitch, you should ?have fried the whole 5 V bus, or at least IC402/403. My similar failure occurred at 13+ V but did not take out the 5 V bus, so I don't think 13+ V ever got past the SMPS, although I can't prove that.?
--
73, Dan? NM3A

Agree

It is definitely the case that Tantalum caps do that, surprisingly frequently and without much provocation
Unfortunately when they let go, like a Lithium battery there is a self fuelling exothermic reaction that has nothing to do with the applied power

Bad batch of caps ?

I would prefer to see an aluminium electrolytic to supply the uF and a parallel ceramic to handle the higher frequencies

Rick

Daniel, I dont agree...
The resistance of L101 is 2 Ohm and with a good short circuit in C107 and D108 the voltage rise will be limited by the voltage drop over L101.
If the supply voltage is disconnected fast enough the damage should hopefully be limited to C107 and D108
To operate a 6.3V tantalum capacitor at 5V should not lead to a failure, unless the capacitor is faulty

73, Gunnar SM5EIE?

Rick,
I will repair the board with an axial lead zener of higher power rating and put a ceramic capacitor in place of C107.
Then I will fit a large electrolytic somewhere on the main board.
I will do a simulation of AC-coupling the PWM signal before deciding if I do that modification

73 de SM5EIE /Gunnar

Hans, I agree about the component failure scenario being unlikely.
I assume that the PWM signal is generated by a HW timer, you are not running a 166 kHz interrupt...
What if the software crash writes to the timer configuration registers and changes the function to a normal output?
Do you have a comment to making ghe? PWM signal AC-coupled?

73 de SM5EIE /Gunnar

Does anyone have a scope plot of what the first second of Voltage looks like across C107 when you power ON , do it multiple times ON-OFF, ON-OFF, etc.

470uF 6.3V Tantalums may only be available with a 7343 (LxW) footprint but so are 220uF 16V Tantalums.?

73 Kees K5BCQ

Kees,
?I believe 470uF 10 V tantalums are also available in 7343, but are 1mm higher.?
--
73, Dan? NM3A