@mava70
I guess you're referring to this YT video :-
I tried this mod just this week on the four digit display version I'd purchased from BG some 16 months ago (virtually identical to the main board in the three digit version shown - three 330R 5W instead of just two). However, I was a little leery about emulating the ultimate 33R with 47nF caps mod so compromised with 33R and 22nF. Using these values the input power for 320W output went up from 394W to 397W and the IR thermometer readings of the power fets showed just a modest ~4 or 5 deg C rise.
A week or so later, I decided to test the 33R 47nF combination (I couldn't find any suitable 33nF in my collection of salvage). For exactly the same output power (32V@10.08A), the input power rose to 404W and the IR thermometer readings on the power fet heatsinks went up another 7 to 10 degrees. After 40 minutes, I finally got to prove the efficacy of the overheat protection which simply disables the output until the cooling fan has taken the temperature back down below the cut-in temperature point.
This gave me the motivation to pick through my collection of circuit boards in my basement which had been awaiting the Round Tuit needed to recover another batch of 'salvage' from which I managed to find just three suitable 33nF 100v mylar capacitors. Swapping the 47nF caps for these 33nF caps reduced the input power to just below the 400W mark, allowing it to run flat out for over 45 minutes without cutting out. In fact I simply reduced it to quarter power (16V 5A) at the 45 minute mark which allowed the fan to shut off after another 5 minutes before cycling to keep the temperature in the safe zone over the next 15 to 20 minutes before I resumed the full loading state.
This result answers (to some extent) nctnico's then untested speculation of the 10A version being less reliable than the 5A version he'd been playing with way back in 2015. Considering the abuse I'd applied to this 4 digit version, the reliability (so far!) seems to be holding up. Mind you, I entirely agree
with his thoughts and observations about the safety and EMI/RFI issues (but this is par for the course with most cheap Chinese kit - opening them up to defuse the various time bombs contained within[1] is SOP for any seasoned 'cheapskate' collector of such "Sows's Ears"
).
In this case, turning it into a "Rayon Purse" will be difficult enough and you'd best give up all hope of ever converting it into some facsimile of a "Silk Purse" - it is what it is, a cheap 'n' cheerful bench supply to experiment with less critically sensitive projects or testing the immunity to noisy DC power of kit designed to be powered via a DC input jack.
As for my own efforts, as well as optimising TheStuffMade gate drive mod, I also threaded about 40uH's worth of small ferrite rings over both output wires between the main board and the banana binding posts for common mode filtering, along with a 470nF 250vac mylar cap at the board end and a couple of 2.2uF 100v mylar caps across the banana binding posts plus about 10uH's worth of ferrite rings over each individual wire to provide additional differential mode filtering.
I also replaced the rather weedy earthing link wire connection to the C14 PE tag with a toroidal choke to reduce pollution of the mains wiring with common mode RFI and tied the negative to the case ground with a 3k3 1W resistor to suppress the 25vrms 50Hz imposed by the various Y caps linking to the half mains live voltage from the HT rectifier pack down to half a volt (also neatly providing a relatively lo-Z 'static drain' connection). That choice of resistor still permits it to be safely stacked in series with another psu up to 50v (60v at a push).
These measures reduced the switching spike noise from its original 1.5v ppk down to a "mere"
50 to 60mV at the full 680MHz BW of my 'scope (the 20MHz BW setting reduced this by about another 12 dB), leaving the 'ripple' unaltered (varying between 2 and 14mV ppk depending on voltage and loading). It's still far from being a 'stellar result', more a case of being just "meh" rather than "absolutely horrendous".
I did try nctnico's Y cap mod to link the output switching noise back to its source in the HT rectifier pack but I couldn't detect any improvement in spite of using a couple of 8nF caps in parallel for "maximum smoke"... er benefit. I think the rather spread out circuit layout, in spite of my linking the centre tap of the secondary to the nearby negative rail of the HT connection to the low side FET's source to minimise unwanted inductance, rather defeated this attempt at reducing this capacitively coupled switching noise between the primary and secondary sides of the transformer.
I'm beginning to think this switching noise is being magnetically radiated directly from the transformer rather than via just the usual conductive pathways. Fitting a copper shield around the whole transformer to short out leakage flux might provide a significant level of mitigation in this case.
I might try making up such a copper 'shield' to fit over the transformer to test this hypothesis when I'm next in the mood for yet another modding session. Anyway, I thought it was worth mentioning just in case anyone else fancied trying out this possible solution.
Incidentally, for anyone looking for a cheap smpsu based bench supply where the cheapest or cheaper option is "The wrong mains voltage type for their region", the conversion from 220 to 110 v mains is merely a matter of linking the junction of the two smoothing caps to the neutral side AC connection on the bridge rectifier or removal of this link to convert a 110v only unit into a 220v unit. Adding this link converts the full bridge rectifier into a half bridge full wave voltage doubler (+/- 170vdc wrt neutral/capacitor junction - a total of 340vdc from the 110vac supply). Removing this link on a 110vac unit converts it back into a full bridge circuit giving the same 340vdc total off a 220vac supply. This is the standard way to add a dual mains voltage selector switch to smpsus which came into general use with the advent of the IBM desktop PC (and clones) some 40 years ago (and it wasn't a new idea even then!).
One of the very first things I did with this bench psu was to disconnect the voltage selector from the main board to eliminate the risk of the selector switch being accidentally (or deliberately) moved to the 110v position since I have absolutely no plans whatsoever to operate it outside of the UK.
I did mark the original connection points to cover the extremely unlikely need to operate it from a 110v supply at a later date. I'd had it in mind that if I didn't, Sod's Law would guarantee that such a need would arise no matter how extremely unlikely this seems (second guessing the Lords Sod and Murphy is a losers' game).
[1] The 'time bomb' in this case being the power fet's heatsinks' tending to lean towards each other to varying degrees of (undesirable) 'intimacy', some to the point of touching each other at the slightest hint of vibration (as a few unfortunates discovered to their cost). All those YT tear down videos provide a valuable service in forewarning the prospective purchaser over what needs sorting out to avoid the "Tears Before Bedtime" effect.