Just a quick addition - the dead Xicon electrolytic had no visible signs of 
failure - no bulging or leakage at all. The DMMs, however good they are, 
were mostly useless in this case as they were simply confused by what was 
happening - no DMM or analog VM would have helped much - an analog VM would 
probably have been worse as it may have reacted as a true-RMS meter which 
would have hidden this issue completely. At least the DMMs went a bit 
bonkers, which made me think a bit!

The 'scope is what nailed it; the gate on the FET was clear, as was the mad 
HT output line.

Nick

On Thursday, 14 September 2023 at 14:22:00 UTC+1 Nick wrote:

> So,
>
> I have had an original NixiSat from new, so a long time now - I detailed 
> it's restoration to a more modern time source and working condition in 
> these two threads:
>
> https://groups.google.com/g/neonixie-l/c/vcNOdSB3nYM/m/3AmeUTtCAQAJ
> ...and...
> https://groups.google.com/g/neonixie-l/c/evzX0noEGXg/m/JrVaZG3YFogJ
>
> However, a couple of days ago it had completely died; the PSU, a 
> switched-mode 12V 2A wall-wart was fine, but the clock was completely dead.
>
> A quick check showed that the tiny 1.5A fuse had blown... and the IRF740AS 
> FET was TOAST, almost literally. It had got so hot that it had burnt the 
> PCB, (1cm charred circle) lifted the track and died to the extent that it 
> no longer identifies as any sort of semiconductor... it's shuffled of this 
> mortal coil and gone to join the heavenly choir  invisibule...  . 
> Interesting smell, well known to EEs (no, not pizza, burning FR4).
>
> I replaced the fuse, did some clean-up of the crispiest bits, removed the 
> CPU, dead FET , the tubes (position recorded for each tube as it was 
> wrapped and stored) and HV drivers and did some voltage checks. The MAX771 
> was running and the 5V rail was fine... the HV drivers and CPU were looking 
> good.
>
> So I replaced the FET with a more modern FDB14N30 which is also in a 
> TO-263 (D2PAK) and is just generally a better FET (and I had some) - 
> specifically lower Qg & RDSon; similar VGS(th). Tracks repaired just for 
> the FET drain with 28G TCW.
>
> Using a good lab PSU (Tektronix PS2521G), 'scope (Tektronix 2465A) and 
> DMMs (Agilient U1272A & U1253B) plus an Aglilent FLIR,  I started looking 
> for trouble...  it arrived quickly.
>
> With 12V in, the current went right up to 260mA and the FET temperature 
> started climbing rapidly - at 70C I cut the power. The HT rail was not 
> easily readable and was the first indicator of what might be happening. The 
> FET gate was looking fine with variable width pulses at about 45kHz but 
> much more "mark" than "space" - not what was expected for an unloaded HT 
> SMPS boost controller - it should have settled really quickly into more 
> "space" than "mark"...
>
> Looking at the HT rail using the 'scope was revealing - it was all over 
> the shop, from about 120V up to 290V - sharp rise time and then decay until 
> the next hit from the inductor...
>
> Just for sanity, I checked the feedback resistors - 1M5 and 12K7 - they 
> were fine, but obviously what was happening was that the wild swings were 
> driving the MAX771 insane, that was keeping the FET mostly on and that in 
> turn was connecting the FET between 12V and  GND with about 0R25 or series 
> resistance (the current sense resistors, 0R125 and the inductor 100uH and 
> 0R11) plus the RDSon of the FET (which is about 0R290). 
>
> What on earth would cause this? It's obvious if you think about it :)
>
> The boost converter output tank capacitor should be 4u7F @ 400V. I removed 
> it and checked - it came in at 5p5F ! Another dead component. For 
> reference, it was an Xicon.
>
> Without a tank, the output will do exactly as I observed and the FET will 
> cook. Lucky that other stuff like the HV drivers and the MAX771 didn't get 
> damaged too...
>
> Replacement cap - I had a nice, good quality Nichcon 10uF @ 300V with the 
> same pin spacing, so used that.
>
> All done. Without the tubes, the NixiSat idles now at about 25mA with the 
> FET stone cold and the gate staying at 0V as the tank was a nice, steady, 
> 182V. Replaced each tube in original position. All good.
>
> Clock fixed for almost nothing, but an interesting lesson for those who 
> haven't fiddled with these sorts of supplies...
>
> Nick

-- 
You received this message because you are subscribed to the Google Groups 
"neonixie-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web, visit 
https://groups.google.com/d/msgid/neonixie-l/27d627c8-162d-47c0-b04f-f94aa0918d81n%40googlegroups.com.

Reply via email to