MeanWell SP-320-48 Repair Thread

amberwolf

Administrator
Staff member
Joined
Aug 17, 2009
Messages
40,479
Location
Phoenix, AZ, USA, Earth, Sol, Local Bubble, Orion
i just got this less than a month ago from mdd0127 along with a pack it's to be used to charge, and the mini-limiter board (not yet installed then), and i've already done something to it to cause it to fail. i don't knwo what, yet, but it should be something simple.

i hadn't tried it out yet, so first i powered it on, got the green led, and 48.8v out across the main terminals. i can't remember if the fan spun up or not.

unplugged it and took the lid off to take pics and to prepare for installing the mini limter board.

next, i researched the right spot to tie the sense wire to from the mini limiter board, at the resistor end that gets 2.48v over by the voltage adjust pot. in at least one of the posts specifying this info, it calls that r25, but on mine it is r75 (EDIT: this is R73, I just can't read), with r55 right next to it. mine is a 1/8w resistor, rather than the 1/4w seen in photos i ran across, but it's in the same physical place and it's got the right voltage. so....

i can't quite get in there with the soldering iron and wire, so i ahve to take the pcb out to do it. unscrewed the pcb from case, and then the heatsink clamps from all three sets of components that mount to the case.

took pics as i got it apart, to note the poor assembly process it went thru, where they didn't even bother to do anything more than dab the paste on there, no spreading or anything. they appear to have just put a dollop on each point it would need it, then screw the stuff down hoping it'd spread out (it didn't, not enough). plus these really thick rubber thermal pads/bags around three of the five case-heatsinked parts. figured i'd worry about that later, once i decided how i'd use it.

soldered on the sense wire from the ml board, then reassembled the meanwell, including putting the lid back on. plugged it back into the wall, and nothing. no light, no voltage, no motorcars, not a single luxury. :(

i checked first for broken leads or bad solder joints, and found neither. i reflowed a few critical ones anyway, for any parts i'd touched/bent during disassembly or install of the sense wire. and removed the sense wire, to no avail.

my three hour tour of all the es threads i could find on meanwells later, and poking and probing it, i found that basically the ac side works all the way up to the first switching stage, and then i'm not apparently getting any switching. haven't figured out why, yet, and am just too tired to continue right now; i keep dropping the probes and tha'ts gonna fry something for sure if i do it on the board. :(


so, below will be pics i took before i broke it, and later more troubleshooting hwen i have time and energy. hopefully someone else has had this same problem and i just missed their solution to it in the threads.
 
pics below. note there are no shunts on this one, at least not as a wire typel i assume it's using one of the large power resistors as a shunt instead, dunno which one yet, probably the last one just next to the positive output terminals.
 

Attachments

  • DSC05800.JPG
    DSC05800.JPG
    16.5 KB · Views: 8,888
  • DSC05801.JPG
    DSC05801.JPG
    30.7 KB · Views: 8,888
  • DSC05802.JPG
    DSC05802.JPG
    53.7 KB · Views: 8,888
  • DSC05807.JPG
    DSC05807.JPG
    42.4 KB · Views: 8,888
  • DSC05808.JPG
    DSC05808.JPG
    33.9 KB · Views: 8,888
  • DSC05809.JPG
    DSC05809.JPG
    38.3 KB · Views: 8,888
  • DSC05812.JPG
    DSC05812.JPG
    42.9 KB · Views: 8,888
  • DSC05813.JPG
    DSC05813.JPG
    36.9 KB · Views: 8,888
  • DSC05814.JPG
    DSC05814.JPG
    44.2 KB · Views: 8,888
  • DSC05815.JPG
    DSC05815.JPG
    28.3 KB · Views: 8,888
  • DSC05818.JPG
    DSC05818.JPG
    33.3 KB · Views: 8,888
  • DSC05821.JPG
    DSC05821.JPG
    23 KB · Views: 8,888
copied from my battery repair thread:
http://endless-sphere.com/forums/viewtopic.php?p=524036#p524036

cor said:
Cool that it is a PFC supply. It should have 3 functional blocks: first the AC input with rectifier and PFC switcher, ending in the big caps that store the upconverted DC (usually 200 or 400V). Then comes the PWM controller and switching FET(s) to turn the HV DC into a square wave that drives the transformer.

everything up to the hvdc works fine; it has ~160VDC on the caps. i haven't gotten back to it yet to know about the next step.

Third, at the output of the transformer a rectifying diode, filter and shunt with current measurement, feeding back to the PWM controller, together with the voltage feedback. I did not see opto couplers on the pictures. Are they hidden next to the transformer? On the NES they are right next to the transformer.
yes, there are two. in this pic
http://www.endless-sphere.com/forums/download/file.php?id=74372
they are to the upper right corner of that top transformer in the main switching section. i will have better pics of both sides of the board later, tonight if possible. have to set up the camera on tripod cuz i cant hold it still enough.


For fixing, you can check if the PFC stage is still making high voltage DC and then check the final stage for voltage (and current) control and other feedback (on the NES there is also overvoltage and overtemp protection feedback) before you try to attack bringing the PWM stage back to life, since the other parts affect the operation of the PWM controller.
Success!

i've been following your reverse engineering of the other mw with great interest. :) i've had to do that kind of thing before, but don't like it much. :lol: if i have time i will compare your existing schematic as drawn into program by neil, and update it for this version where i find differences. i definitely appreciate you dropping into this thread to help out. :)

link for others to the section of the first thread where schematic capture began
http://endless-sphere.com/forums/viewtopic.php?f=14&t=27199&start=255
and conversion into machine format
http://endless-sphere.com/forums/viewtopic.php?f=14&t=36114


and more reference data:
amberwolf said:
regarding the model designations, i found:
http://www.meanwell.com/webnet/search/seriessearch.html#open
seems to show that sp series has pfc while s series is not listed, but probably is same without it. i found a spec sheet for the s-340 series that doesn't mention pfc
http://www.soaring-cn.com/UploadImages/Specification-350W%20Series.pdf
sp series doesn't show a 350, though, just a 320 then a 480, so i dunno.
http://www.meanwell.com/search/sp-320/default.htm
well, mine says it's from 2008 on the pcb, so they might not make that line anymore. assuming it's not a clone from somewhere else.


amberwolf said:
oops. my meanwell is an sp-320-48, not 350. i cant' read. :oops: so it is the pfc-type series i linked above. (note: fixed thread title)

mine has no shunt at all, at least, not as a wire-jumper type, unless i am just blind, as well as illiterate.


amberwolf said:
now that i'm seraching for the right model number, i've found info on this one going back to at least two years ago jan 2010:
http://www.endless-sphere.com/forums/viewtopic.php?p=228838#p228838
as one example.

so they're not new and untried, but i havent yet found enough info to tell me what i need to know to fix it and use it.

unfrotunaely most of the pics i would like to have for reference are missing or trashed. :( edit: but aussiejester's site still has the hires ones he took:
http://members.ii.net/~aussiejester/index/pictures/Custom%20RC%20Cruiser/Hi-Rez%20Meanwel%20SP320-48v%20PSU/

even some partial schematis
http://wookware.org/pics/meanwell/002-IMG_2371.JPG
 
i'm not sure u\yet, but it looks like the main chip, ml4800cp, may be internally shorted. when i trace out the power, i get about 60vdc at the pre-switching section, up to the diodes, xfmr and fets, then nothing. tracing to the low-voltage kickstart supply, a large 2w 47k and a zener (can't read the p/n right now), i get 60v at one end of the resistor, and zilch at the other.

measuring resistance i get 47k for the resistor and 19ohms for the zener, which goes from that resistor to this section's ground. i started desoldering one end of components in that path to isolate which one is dragging it down so far, first disconnecting the thermal breaker that's in series with the zener output and next a jumper zerohm resistor, and then pin 13 of u1 (ml4800cp). at that point pin13 is at 18ohms to ground, and the rest of the circuit is now at a normal high resistance of 1.3megohm.

according to the datasheet
http://www.fairchildsemi.com/ds/ML/ML4800.pdf
http://www.fairchildsemi.com/pf/ML/ML4800.html
pin13 is supposed to be vcc for the chip, and a 7.5v internal reference voltage. many other pins also have a very low resistance to ground, that i would not expect.
ml4800.PNG


i gotta stop for now, cuz i am having trouble holding probes without shorting stuff out. later on i will see if any dead psus i have around here have the same control chip. probably not, cuz i don't think i have any pfc units.


what i want to know is what killed the chip, because nothing was even connected to the system except ac power, for the time it was powered on successfully. the time it did not power up at all, the mini limiter was connected but i can't see any way it would have killed the chip--there's nothing i can trace back from the sense lines to the switching section where the chip is. it's all isolated completely.

the only thing i can imagine is esd damage, which is easily possible as i was not observing proper precautions like i usually do, because i have misplaced my wrist strap (or a dog ate it :lol:), and rather than spend the time rigging up another, like i should have, i just went ahead with the disassembly and limiter installation.

later i will take the whole chip out of the board, and remeasure the pads to ground and see what they are without it, then measure it's pins to it's ground.

i wish i could think of a way to power up the rest of the psu without the chip to do it, but it pretty much does all of the control and switching. it's the only chip on the board besides the optos.
 
i desoldered the chip and most of the pins are less than 50 ohms to the ground pin. no way that is supposed to be that way. :(

so gonna have to find another of these chips. i can get one from mouser
http://www.mouser.com/ProductDetail/Fairchild-Semiconductor/ML4800CP/?qs=sGAEpiMZZMvxMWXNOhNKbi7cIyZV0kXKMujlSKdBktc%3d
for $3.37 each, plus $6.95 postage.

digikey has them for $3.59 each, plus $8 postage.
http://parts.digikey.com/1/parts/626617-ic-pfc-ctrlr-average-curr-16dip-ml4800cp.html
but when you actually attempt to add one to the cart it charges $3.99 for it instead!

i'm gonna see if anything i already have around here has one i can salvage before i order anything. also gonna wait until mdd0127 comes down, and do some comparison measurments between his units and mine to see if anything else is cooked that i would have to order.
 
copied over from my battery repair thread:


amberwolf said:
cor said:
If you get only 160V on the caps then that tells me that the PFC is not working, as that is the normal voltage of rectified 110V AC, not the boosted voltage of a PFC stage, so I would certainly recommend to test if the PFC switching FET is actually working and whether the ML4800 PFC and PWM controller is working.
i found earlier and posted in the repair thread that the the pfc and pwm chip is not working--it appears to be internally shorted:
:(
just posted a minute ago in there that it's definitely shorted, and links to where i might be able to buy a new one.

cor said:
I noticed that you already found the datasheet with a reference schematic here: http://www.fairchildsemi.com/ds/ML%2FML4800.pdf
The 900V 9A PWM MOSFET 2SK2082 from Fuji Electric can be found in many places, for example: http://pdf1.alldatasheet.com/datasheet-pdf/view/60874/FUJI/2SK2082-01.html and the PFC FET IRFP460A from IR can be found here: http://www.irf.com/product-info/datasheets/data/irfp460a.pdf it is a 500V 20A N-channel MOSFET.
Please note that the voltage of the caps C5 and C?? is 400V, which is a typical PFC boost voltage from anything between 90-260V AC.
ah, ok. i have a lot to learn about pfc, having rarely fixed any psus with it, and so far only having to replace bad caps that boiled from high temps in sealed units, or blown fets or output diodes from the caps having died.

regarding the mosfets i actually have some spares of these out of other dead psus that had leaky caps and such, or were simply scrapped by someone else due to age, or not being needed, so i harvested all the parts i could.



The location of the optos and the array of diodes under the transformer looks like the NES setup, but due to the extra complexity of PFC, the controller section is about twice as busy and you have an additional switching FET and rectifying diode.
i was following your schematic on the nes version, and this one is definitely a lot different, presumably because of pfc stuff and using the combo chip rather than just the simpler nes method. still, i will probably be able to take your schematic and modify it to match this one, with the output section being mostly the same, i think. i havent' tested or measured or traced anything on the output area yet.


BTW, I wonder who Willis Lu is, who put his name on the PCB.
i'd guess willis lu was the layout designer, so he or she "signed" the work. :) might be the actual engineer designing the whole psu, too.
 
Hi Amberwolf,
from the pics it looks like the output (secondary 48V) side is also different than the NES, because the voltage and current limit were compared to the 2.5V reference in a dual opamp U100 on the NES, but I do not see that on this SP. Of course the principles of voltage and current limit stay the same, it can be implemented in many different ways...
BTW, in your first post you mention the voltage feedback resistor that you tied the mini limiter to. Was this R73 (1k5) instead of R75?
Where else did that board attach - that may be the first indication where trouble was started and hopefully explain how the ML4800 got damaged.
 
yes, r73. :oops: i did a number on my numbers for much of this project, apparently. :roll:

the only other connections of that limiter board are to main + and main - terminals of the meanwell.
http://endless-sphere.com/forums/viewtopic.php?f=31&t=21768
it would be mounted like this:
file.php


the v2 schematic is this:
file.php


schem layout bom
http://endless-sphere.com/forums/viewtopic.php?p=496760#p496760

which is prety much what i have.




future versions below:

v 4.2.5 schematic
file.php


v 4.2.6 schematic
file.php

http://endless-sphere.com/forums/download/file.php?id=73737
 
gmuseless suggested the internal zener on the ml4800 chip could've failed shorted to ground, essentially grounding vcc of the chip, which would cause all sorts of bad/wierd currrent paths inside the chip.

makes sense, but i wish i knew why, because afaict without powering it up the zener that pre-regulates the power to the chip is working.

i'd have to power up the system without the chip in it to be sure, though, and i'm not sure that's safe to do for other components. probably, but my inexperience with doing that....

if no one else posts a reason why i shouldn't be able to do that, i'll probably try it later tonight after work. and maybe roll video, just in case.... ;)
 
Hi Amberwolf,
There is a small risk in powering the board without the LM4800, but it depends on the actual design. The reference design in the datasheet has no other protection than a zener to ground for the FET drive of PWM and PFC, so if one of those gates floats up then the MOSFET might blow when it starts to conduct in its linear region.
The reference design has an external 15V zener across the Vcc but it may be that Meanwell removed it since there is already an internal zener.
I can't think of a common way that this controller should fry in this circuit, other than one of the FETs blowing and fusing its gate to the high voltage Drain, which will certainly blow the controller as there is usually only a small resistor between them. If there is a gate protection zener than that should catch most of the hit, but I can's see how Meanwell implemented that protection.
Certainly measure the FETs for shorts before going any further. The S-D channel should be high resistance until you make the Gate positive (between 5 and 15V) with respect to the Source, for example with an old 9V battery or an isolated power supply. But often when the FET fails, all 3 terminals are fused together internally. Success!
 
My SP-48's all have output over voltage (around 62V) protection and when it kicks-in the unit appears quite dead. Disconnect power to reset but if the pot is asking for too much voltage it will immediately shut off again.

Are you sure the volt adjust pot didn't get cranked up and/or damaged during sense wire install? Lemme know if you need to compare anything with a known working SP-320-48?
 
this pot has a tiny hole for screwdriver in it's face, which faces outward on the pcb so no way to bump it and turn it when working near it, unlike others like the pic from another thread linked above. i don't know for sure that it isn't changed, but afaics it's still in the same position it was when working, just left of center. i haven't measured it's resistances, but as it wasn't touched during the install, it's unlikely to have been damaged.

it definitely didn't reset after power disconnect, even after being left for over an hour, and verifying there was no charge left in any caps.

all i can think of at the moment to compare is the resistance of pin 13 of that chip to ground (pin 10, I think), after you've unplugged teh mw from ac power and any load on it's dc out, and all voltage is drained from caps and such. i'm pretty sure it can't be a super-low resistance.


@cor:
i can easily add a grounding resistor on the fet gates so they don't switch on. didn't think of them going linear but that's a good point.

the fets are not shorted because they don't have any voltage on their gate or output, only on their input pin. i haven't tested to see if they switch or not, but can easily do so. my fluke meter has enough voltage on it's ohm test to switch most fets; i've used it to test quite a few per this 4qd page:
http://www.4qdtec.com/mostest.html
(he added the fluke there after i found that it works and let him know)

i'm hoping to have time tomorrow to diagram some of the mw.
 
I run mine very close to highV limit (62V) and sometimes the HVC kicks in and sometimes it doesn't. I do know those pots change a little when bumped around. Dunno where yours might've been left, just suggesting to crank it back a little + cycle power, to make sure it's not something that simple?

Gotta head out to work now but will pull one apart and check those pins when I get back in a few hours.
 
amberwolf said:
all i can think of at the moment to compare is the resistance of pin 13 of that chip to ground (pin 10, I think), after you've unplugged teh mw from ac power and any load on it's dc out, and all voltage is drained from caps and such. i'm pretty sure it can't be a super-low resistance.
@cor:
i can easily add a grounding resistor on the fet gates so they don't switch on. didn't think of them going linear but that's a good point.

the fets are not shorted because they don't have any voltage on their gate or output, only on their input pin. i haven't tested to see if they switch or not, but can easily do so. my fluke meter has enough voltage on it's ohm test to switch most fets
Just make sure that you don't test it while the MW is powered or you will blow the PWM FET.
Correct that the resistance between pin 13 (VCC) and 10 (GND) should be pretty high on the part. Otherwise it would never startup (bootstrap) from a small current via a high resistance directly from the AC input.
To test the basic operation of the part you can apply voltage to pin 13 directly and check that pin 14 supplies 7.5V reference. But if the part draws more than a few mA then you can presume it is bad and needs replacement.
Success!
 
Pins 10-13 begin around 20k range and climbed to over 100k. I guess a cap is charging? Reversed probe polarity and it started out low and went high again.

And just to confirm the sense wire is tapped on R 73 the end closest to the LED.
 
between the posts above, it verifies that the chip is definitely toast, and that i did have the mw limiter hooked up right. i still need to check everything else that goes to the chip to make sure a new one won't cook as soon as it's installed. :(
 
Earlier today I did some more poking and prodding, including verifying the FETs are open when no gate signal, and turn on with gate voltage, using the meter test I usually use for this (from 4qd I linked on the previous page).


Then this evening, Mdd0127 was down in the area for other stuff, and brought by one of his identical MWs and ESD-safe mat/wriststrap, so i could compare components/values/readings/etc. between the two.
DSC05874.JPG


All the components I can check with resistance measurements in-circuit are within component tolerances AFAICT, comparing one to the other. Can't do power-up comparisons, since the brain of mine is not installed, but that tells me it is unlikely anything is toasted completely besides that chip.

So I soldered some 1Kohm resistors out of my scrap bins from the gate to source of the FETs to keep them from drifting on into linear operation and smoking. Then I plugged it in to AC and turned it on...and the fuse blew spectacularly.

I pondered on this to see if i could figure out why it would do that if everything was essentially "off", and then when I was about to put the board away, I realized I am a moron.

I soldered the resistors from gate to DRAIN. :shock: :oops: :roll:

So I probably blew up the FETs, too, though I haven't rechecked them yet. If I am very lucky, the fuse stopped anything serious from happening by blowing fast enough, so I am going to go check the FETs and if they are ok, I will resolder the resistors from gate to SOURCE instead :roll: and retest the power-up. But I will put an even smaller fuse in there than normal, just in case. ;)
 
Hi Amberwolf,
Such things happen - even to the best.
You have a good chance that the FETs are toast, since having a resistor to Drain will *force* them into their linear region, though the voltage across should be in the order of their Gate turn-on voltage (often around 3-4V) so even when they drew a short 50A spike or so, their power should have been under 200W which is a lot, but depending on how they were heatsinked and how long it lasted, they might be OK or have molten internally...
Good that everything checked out OK for resistance, that narrows the problem down considerably. There is of course always the odd problem with a transformer winding shorting together or a component failing under high voltage, but it is likely that if you install a socket for the LM4800 (just so you do not have to desolder it again and ruin the board if it might fail again) and install a new LM4800 then it might work...
Success!
 
I got lucky--the FETs are fine (or at least they pass the low-voltage ohmmeter tests). I put the resistors from gate to SOURCE now, and powered up the supply fine even wiht just a 4A fuse temporarily stuck in there till I find one closer to the 6.3A that came out (the other i had handy is a 2.5A).

So now that I can power it up without exploding things, I measured the voltages at the pads for the chip, and most of them are low, from 0V to about 1.5V to 15V (on pin 13, VCC), but pin 15 (Vfb) has the full 160V on it. :shock:

The table in the ML4800 pdf spec, page 3 states that VCC pin can have max of 18V on it, Isense can have -5 to +0.7V, and any other pin can be max of GND-0.3 up to VCCz + 0.3V.

Unless I misunderstand that, the max voltage for every pin should be no more than 18.3V absolute, and even that isn't an operational voltage.

This voltage is coming from a resistor path that leads to (I think) one of the PFC transformers, but I have to re-trace it out and draw it up. I'm really hoping that it is not a transformer that's failed, because while I can probably rewind it if I have to, I'd really really rather not. :(
 
First, I can't count. It's not pin 15, it's pin 2. :roll: upside down when measuring it I screwed up coutning them. Pin 2 is Iac, which is the PFC AC line reference input. It's max current input is 10mA, but it's voltage is still supposed to be really low.

A comparison to the same pin on the working MW Mdd0127 left here temporarily for me to use as a comparison for troubleshooting shows it doesnt' get anything like that kind of voltage, only about 1.4v.


Also, I must've missed some components in my resistance comparison testing. :(

D9's pads read essentially shorted, but D9 itself tests ok out of the board. On the working unit it has the PWM output from the chip on it's cathode.

The resistors that connect to the points that go to pin 2 are all good, when tested out of the board, so they are probably not the source of the 160V, even though they are connected to that on one end.

A small chiclet capacitor that connects to pin 2 also reads ok out of the board.

There isn't anything else that connects there, so the source must be thru the resistors, and most likely that means there should be something at the junction point that ought to be grounded, as a voltage divider, that for whatever reason isn't, so the full voltage just goes right thru (albeit at very low current due to the high resistance, it doesn't matter since the chip can't take that voltage).

pics of the area:
DSC05870.JPG

DSC05872.JPG

So...I still have a ways to go before I get this figured out, and will probably need to trace it out as Cor did on the NES type.
 
Hi-res pics of the PCB and components. (click on attachments to view full-resolution)
 

Attachments

  • DSC05862.JPG
    DSC05862.JPG
    389.6 KB · Views: 8,577
  • DSC05863.JPG
    DSC05863.JPG
    432.8 KB · Views: 8,577
  • DSC05865.JPG
    DSC05865.JPG
    423.1 KB · Views: 8,577
  • DSC05866.JPG
    DSC05866.JPG
    420.4 KB · Views: 8,577
  • DSC05868.JPG
    DSC05868.JPG
    473.7 KB · Views: 8,577
  • DSC05869.JPG
    DSC05869.JPG
    501.2 KB · Views: 8,577
  • DSC05870.JPG
    DSC05870.JPG
    515.7 KB · Views: 8,577
  • DSC05871.JPG
    DSC05871.JPG
    488.9 KB · Views: 8,577
  • DSC05872.JPG
    DSC05872.JPG
    532.4 KB · Views: 8,577
  • DSC05873.JPG
    DSC05873.JPG
    462.4 KB · Views: 8,577
The 160V on pin 2 (Iac input) is normal after removing the chip, because also the reference design shows that it is connected by two series resistors directly to the rectified AC line, so it will be at whatever level the caps charge up to. The chip will have an internal resistor that will keep the voltage at a much lower value. (voltage divider with the 2x 249k)
I can't find D9 on the pictures, though there are a few places with bent-over components and reflection. I see D5-6-7-8 and D12 and D13 but you gotta help me where D9 is.
In an earlier question you asked where the shunt was.
From the pictures it is not clear whether the PCB traces themselves act as shunt or that there simply is no shunt on the secondary side, only primary side protection.
 
Sorry, D9 is circled in this pic above:
http://www.endless-sphere.com/forums/download/file.php?id=74588
I think I forgot to type in some of what I was thinking at the time, though I don't remember it now either.

I am going to try to get a schematic drawn up, but it will take a little while. Next chance I really get to do it is tomorrow night after work, but mroe likely my next day off on Tuesday.
 
Hi Amberwolf,
I am guessing that it is R34 (820k? I can't see much of the printing as it is hidden under a wire) that supplies the 160V from the rectified AC to pin 2. Since the chip is not in and thee only other connection is a capacitor, that one will charge up to the AC line voltage... With the chip back in the voltage should be low again.

The D9 is a zener (probably 15V) protecting the gate of the PFC FET. That is why it is wired from pin 12 (PFC out) to negative supply (pin 10). From pin 12 the PFC drive goes through BC4 (which is a wire with a ferrite bead around it to dampen high frequencies) and through jumper J12 to a set of two resistors and one diode at the gate of the MOSFET. The diode and one resistor will take care of a faster turn-off for the FET by faster discharging the gate through both resistors in parallel than when turning it on (through just the second resistor).
My question is, did you have the gate-to-source resistor in place when you measure the diode D9? If the gate is shorted to source (which is connected to negative supply) then you are measuring the 22 Ohm resistor between gate and pin 12, because pin 10 is already connected to the negative supply on pin 10 also. So you are not measuring the diode and that is why it tests OK out of the board.
Indeed - tracing the board will clear up a lot of this, but it is a bit of (guess) work. However, your supply might be OK with just a new ML4800.
You can test this chip before giving it AC power (I recommend you put it in a socket, just in case it ever fails again) by supplying +14V on pin 13 (negative to pin 10) and check the Vref and outputs, especially when feeding 2.5V into pin 15 (Vfb) and keeping Vdc low (below 1V) to force the chip to generate 50% duty cycle PWM, though the PFC ducy cycle may be 0% or over 95% and hard to detect, depending on the Vfb voltage (it is nominal 2.45V).
There are some smart tricks in the SP, for example there is a Q4 which has the same function but is wired differently than in the reference design. In the reference, it monitors the voltage feedback circuit (the output of the opto) as the Vdc pin 6 should be pulled low by the feedback and pulled high by a resistor to Vref (7.5V). The voltage should be pretty constant here, as it is compared to an internal reference (but that can be influenced by the RAMP2 input) so if it is seen that this feedback voltage drifts up to Vref (opto does not conduct at all) then Q4 will start to conduct and - in the reference - it will pull the current sense DC I-limit pin 9 high and shutdown the supply. In the SP the Q4 will pull the SS (SoftStart) pin low to negative supply, which also has the effect of shutting down the supply and requesting a fresh soft startup.
 
I haven't been able to sit down and work on this again yet. I keep being pulled away for other things, and getting distracted. :(

I did indeed have the gate-source resistor (1kohm, IIRC) installed when I measured the voltages. I think I measured all the resistances before that, though.

You may well be right about the 160V thing without the chip in there, that it's just a floating voltage. I'll have to trace out at least that part of it to find out for sure.

Or...before I install a new chip, I could put a small resistor to ground from that pad? I don't know exactly what the input impedance of that pin of the chip is (have to check the spec sheet to see if it says).

Let's see: 160V/10Kohm = 16mA, whcih would be 2.56W (assuming there was no other limit to current flow, which of course there is, from the other resistors in line). I guess I can try it. If it drops to some very low voltage there then maybe I'm ok?


I *think* that if all i do is install a new chip it will just work. But it'd be nice to know in advance. :)


Testing the chip functions before applying mains power sounds like a very good idea that I hadn't thought of. :)

I really appreciate your help here!


I guess it all comes down to me having to get back to this and trace it out and test it.
 
Back
Top