Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Troubleshooting an inverter

  1. Jun 9, 2017 #1
    Hi all,

    1500W 12Vdc to 110Vac modified sine inverter.

    Two output MOSFETs shorted and took some of the driver components with them. I removed all the MOSFETs, and fed in some power carefully - once the voltage rose to 12V, I could see the large output caps were being charged to over 200V, and the input oscillator circuit was working. I concluded the input side was OK.

    However, now I've put in some new MOSFETs, replaced their gate driver transistors and two ICs - TC4093BP. These are quad NANDs with Schmitt triggers, I guess for wave shaping. Sadly, the unit only powers on for a brief time then beeps and lights the "overload" LED. The output caps are charged for the brief time before overload kicks in.

    There is no short on the output, so what is happening? I have a blurry schematic, and my plan is to follow the overload sense circuit, but it's hard going as the schematic refers to an older model and not everything is the same.

    How would you approach this?
    Many thanks
  2. jcsd
  3. Jun 10, 2017 #2
    Don't all answer at once, folks!

    I think I'm some way towards solving the problem. The output MOSFETs are driven by BJTs, which in turn are driven by the wave shaping TC4093BPs, which are driven by LM358 op amp oscillators. Output monitoring is via another LM358 whose two halves monitor the HV DC and AC rails, and this is fed into opto isolators. This chip's outputs were shorted.

    This forms the output section. When the output MOSFETs fail, they generally fail short, cause overcurrent and take some or all of the above components with them. I'm in the process of replacing these and will post an update for anyone else looking for a fix.

    I have fixed 5 or 6 inverters in the past, and nearly every one had a shorted MOSFET or two. Regardless of brand or quality, there appears to be no effort made to protect components upsteam. WIth hindsight, it would be cheaper and quicker just to order and replace all the semiconductor components in the output section as a first response. Here in the UK, an LM358 can be bought for 24p. It's not even worth my time getting the meter out and probing the pins - I can potentially rejuvenate a £500 inverter with £10 worth of components and in very little time with this approach.
  4. Jun 11, 2017 #3


    User Avatar
    Science Advisor
    Gold Member
    2017 Award

    Well you haven't given us much to go on !

    circuit info would be extremely helpful for us to help you
  5. Jun 11, 2017 #4
    Fair point.

    Here is the schematic, in two parts. Looking at 0101, I have replaced MOSFETs Q 34-41, a couple of their gate driver transistors, and ICs 11 and 12 NAND gates.

    The machine now goes straight into overload shutdown on power-up. I have since found ICs 14 (a 393 comparator) and 3 (op amp) have shorted pins 1 to 7.

    These will be replaced, of course. In the meantime, I have a funny feeling this won't solve the problem, so was wondering if anyone could help explain the overload circuit. As far as I can see, IC7 opto latches itself with a 1M resistor and feeds into IC5 flip-flop, which lights the OL LED and shuts down the unit via Q42.

    The overtemp LED does briefly flash too - this is another mystery. On the board, there is a TO-92 packaged object splurged on to one of the input transformers with heat sink compound. I guess this is the temp sensor but where is it on the schematic? Am I right in thinking that overtemp signals are relayed to trigger the overload circuit, since only the overload side can shut down the unit?

    Any help would be much appreciated, since:
    1. I don't know what I'm doing, but would rather not throw the thing away.
    2. The schematic doesn't exactly match the board.

    Many thanks.

    Attached Files:

  6. Jun 11, 2017 #5


    User Avatar
    Science Advisor
    Gold Member

    Have you checked the circuit board traces to see if any are bridged together by something foreign; like maybe a flake of solder?
  7. Jun 11, 2017 #6
    I'm having a hard time following it (the OT circuit is bisected by the fold), but my guess is the TO-92 part is a conventional transistor, and the designer used a variation of this circuit as input to the overtemp comparator. Can you get at the flat part of the TO-92 to read a part number?

    inverter power side IOL.jpg

    I'm not sure what is going on either (what is the rectangular box between point E and K, a diode?), but the second op amp in IC3 appears to be comparing voltages between pin 5 and 6. I'd work backwards from there.
  8. Jun 12, 2017 #7


    User Avatar
    Science Advisor

    Q27 thru Q30 are also prime suspects due for replacement; as is the 0.25Ω, 2W resistor at Q37-Drain.

    Output load current is sensed as the voltage drop across the 0.25Ω, 2W at Q37-Drain.
    Both IC3-7 and IC14-1 go low if load current is too high.
    IC14-1 low removes drive to the output stage by driving IC11-3 low.
    When an overcurrent condition exists, diode coupling from both IC3-7 and IC11-3 drive IC7(opto coupler).
    IC7 drives the error logic, turning on the "OVERLOAD" light.

    NOTE: Both an LM358 and an Optocoupler are labeled as IC3.
    IC3(opto coupler) is the voltage feedback to IC1 on the low voltage primary side.
    The approximately 200VDC (V1) supplying the output stage is adjusted with VR3 on IC3-3 (LM358).

    Voltage "VH" should be approx. 12.3V, and "VS" should be about 0.7V less than "VH".

    From your description, either something is keeping some of the output transistors, Q34 thru Q41, turned on; or the 0.25Ω, 2W at Q37-Drain is open or changed value.
    If you have and know how to use an oscilloscope you can signal trace to see what is not switching and triggering the overcurrent.

    I do NOT recommend this, but to assist troubleshooting, you COULD short IC7 pins 1&2 together and see what smokes. This disables the overcurrent protection for the output stage and could also damage components on the low voltage side. You will likely destroy that 0.25Ω, 2W at Q37-Drain.
  9. Jun 12, 2017 #8
    Yes, but I'll check again.

    On the schematic, it's Q3 2n2222A, but on my board it's an LM335 temp sensor.

    I think there's nothing in the box - there's an area of the board that is unpopulated, perhaps for a soft start system or something.
  10. Jun 12, 2017 #9
    I've replaced or checked all the Qs 27-30.

    The two big resistors appear OK too.

    Yes, many of the IC numbers are different on my board. Makes it harder to trace the fault.

    Anyway, ICs 3 (LM358) and 14 (393) had shorted pins 1 and 7 so I'll replace those and move from there.

    I DO have an oscilloscope; I'm learning how to use it (!).

    Thanks for everyone's replies so far - I wish I could read a schematic that well.
  11. Jun 14, 2017 #10

    Replaced IC3 LM358 and IC14 (393) and now the machine powers on, draws about 0.5A at 12.6V, and has 140V on the HV DC rail. Sounds good.

    However, no output. I scoped the output MOSFET gates and worked backwards to find the output oscillator, not shown on the schematic. It's on a daughter board:


    Its a 16-pin IC, KA3525A:


    No Vref, and the Vref pin and outputs are shorted to GND.

    I will replace this, but also, the transistor N1, labelled C124 TSN which I think is an NPN, has an open right leg, which I believe is an emitter. I can't find data on this particular package. Can anyone help with a pinout or datasheet? I may just sub in an S9013 and hope for the best.

  12. Jun 14, 2017 #11


    User Avatar
    Science Advisor

    Might as well replace both of those transistors.

    Found a datasheet -- kind of.

    Silicon NPN, 40V, 25mA, 250mW, VCE sat=10V, hfe=90
    Apparently discontinued about 1975. But the board is dated 2004. Strange.

    (This site also shows 2N2218 as a substitute w/pinout below, but in a metal case.)
    1 = Emitter
    2 = Base
    3 =Collector

    Substitution Guides:
    2SC536, 2SC403B, 2SC372, 2SC645, 2SC300

    2SC1359, 2SC1740S


    p.s. edit: "2S" is the Asian default prefix for transistors with the third character being "A", "B", "C", or "D". That third character denotes the possible combinations of 'PNP', 'NPN', 'low (audio) frequency', 'high (RF) frequency'. Unfortunately I don't remember which letter belongs to which combination, though the "C" obviously shows 'NPN'. That's how I found your 'C124' part number. In the USA the default prefix is "2N".
    Last edited: Jun 14, 2017
  13. Jun 14, 2017 #12
    Thanks! I found the 'tin can' datasheet too, but nothing relating to this odd little package. It's smaller than a TO-92, without the rounded back. Instead, the back edges are chamfered.

    There is a diode drop from pin 2 to 1. The open pin 3 is connected to ground. Can I assume the pinout is CBE?

    I'll try a BC546.
  14. Jun 14, 2017 #13


    User Avatar
    Science Advisor

    We overlapped each other; see edit of my previous post.

    If that second transistor, N2, is the same type as N1 and functional, use that to at least find which is the base lead. If you can't find the pinout for the transistors then order several spares so you can try different pin assignments after letting all the 'Magic Smoke' out of them.:rolleyes:
  15. Jun 14, 2017 #14
    This board is most certainly made in China, judging by the schematic. I did try 2NC124, not knowing about the 2S variant.

    Sadly the N2 is a normal TO-92 so I can't use it as a reference. Plan B it is...
  16. Jun 14, 2017 #15


    User Avatar
    Science Advisor

    What is the part number on N2? It is likely N1 and N2 are driven directly from the KA3525A outputs. If they are, and N2 is also an NPN Switching xistor, then use the N2 type for the N1 replacement.

    Would you be able to draw a schematic of the circuitry connecting to N1, N2, and the output pins of the KA3525A? Since it's a single sided PC board it shouldn't be too terribly difficult. You will learn a fair bit by trying to draw the schematic and a schematic will help us to help you.

    Perhaps you could also post a good photo of the back side of the board. I'm willing to spend a little time confirming your partial or complete schematic..

    (BTW, from the shrink tubing on the wiring to that board, it looks like it is already a transplant from another unit.)
  17. Jun 15, 2017 #16

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    I've had that trouble, most transistors are EBC but some are ECB . You can use your analog ohm meter on RX1k scale to figure it out.
    NPN you say ?
    identify base, it'll be the anode side of both base-emitter and base-collector junctions . So it'll read a diode drop to both of them in one direction and open in the other.

    Now to tell which is the collector.

    Connect your ohm meter between the other two leads, one is collector and one is emitter but you don't know which is which. Resistance should be near infinite.
    Connect about a megohm to base, touch it in turn to the other two leads. One should produce no effect, the other should make the needle move upscale a little bit. Write down what is that ohm reading .
    Now reverse your meter leads. Repeat that megohm to other two leads step, writing down the reading.

    What you've accomplished is to bias the transistor with your ohm-meter and check its current gain with the one megohm resistor.
    You did that with both polarities.
    A transistor that's biased backwards will still amplify current but not very well. When it's biased forward it works a lot better.

    So whichever polarity gave you the bigger current gain (lower ohm reading) is the correct polarity.
    The lead to which you touched the one megohm base resistor to get current gain in that step is the collector. The other is emitter, connecting base to emitter shuts the transistor off..

    That's a "poor man's" transistor decoder .. Try it a few times and get a feel for your meter.
    Avoid the RX10K scale because it might use a 9 volt battery , typical limit for reverse E-B is 6 volts..

    I've never tried it with a digital meter but i'd think it ought to work.

    Good Luck !

    old jim

    EDIT: PS - be aware that on oriental analog meters when set to ohms the red lead is the negative one. I dont know why.
    On US meters like Simpson and Triplett(not the imported ones) red lead is positive on ohms.

    So an oriental analog meter can confuse you when reading diodes. It'll show conduction when red lead is on the cathode.
    See if 2N3904 looks like a potential substitute. It's a pretty stout transistor and very common,
    Last edited: Jun 15, 2017
  18. Jun 15, 2017 #17


    User Avatar
    Science Advisor

    Unfortunately the 2SC124 datsheet says VEBO=1V so even the RX1 Ohmmeter Test would not be safe for this oddball device.

    Just a FYI. The few handheld digital meters I've used have a special "Diode" range because the normal Resistance ranges use a voltage less than 0.7V Silicon threshold voltage. Don't recall what the benchtop meters use, its been too many years since I've used one!
  19. Jun 15, 2017 #18
    Thanks, Jim. That's a useful 'rolling road' test of a transistor that I'll bear in mind - I do have a little analog(ue) meter. It does things digital meters can't. But this transistor has a fault so I'm having to infer the pinout from where the pins go.

    My digital meter has a diode test function, and I normally screen transistors in-circuit by probing +1-2 then +1-3 then +2-3; reverse probes, repeat. This will always find the base. In an nPn transistor the base will be the common Positive lead, in a pNp the base will be the common Negative. Inspired by thinking of the transistor as two diodes commoned to the base.

    Tom, many thanks again. Reverse engineering is on its way. That shrink tubing was me - since that daughter board looks so bodged in I concluded it was an soft start add-on and 'deleted' it in an attempt to locate the original short. Then I read the chip number, found the short somewhere else and put it back. For shame.
  20. Jun 15, 2017 #19

    The black dots are the connecting wires to the main board. If N1 is an NPN with EBC pinout then it's looking to pull the Vref to ground if the output from pin 7 of IC3 LM358 (main board) goes too high.
  21. Jun 15, 2017 #20

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    The ones i've used try to force a trickle of current through the unknown and measure voltage that results, just placing the decimal point where it belongs.
    My old Beckman applies 1 ma on 200 ohm scale, 0.1 ma on 2K scale, etc..
    On diode scale it tries to force 2 milliamps and reports voltage . It has 2 volts of drive behind it, ie that's all it will apply.
    I had another DMM, much cheaper, that applied only 1 volt max.
    It's easy to check how much voltage a DMM applies - just measure it with another one..

    That's a strange transistor - 1 volt max Vbe ? Do they mean forward ?

    Anyhow, 3904 looks to me a viable replacement.

    PS - Never use analog meter set to RX1 on a small transistor. Meter puts out in excess of 100 ma and will burn out those teeny little internal wires
    That's why better analog meters with a RX1 scale have a D cell instead of AA like the cheap ones with just RX10..

  22. Jun 16, 2017 #21


    User Avatar
    Science Advisor

    Got It!
    N1 pinout is:

    1 = Emitter
    2 = Collector
    3 = Base

    That took a few hours. The KA3525A is a Fairchild part and they have only an eight page datasheet; no signal descriptions at all, just electrical specs & pinout.
    I found an Application Note (AppNote) covering the SG1525, an early version of the SG3525, made by SGS-TOMSON that has a detailed description of the internals and pin functions. See page 6 for both SS (pin8) and Shutdown (pin10). The AppNote is available at:


    BTW, the photo of your schematic has some "interesting" lighting. How come the multi-color lights?
  23. Jun 16, 2017 #22
    Aha, so I had the pinout wrong. I agree the Fairchild datasheet is not much use - just "this is what the pins are, knock yourself out".

    As far as I can see, then, N1 responds to pin 7 of IC3 on the main board going high. It pulls pin 10 of the SMPS chip low and initiates shutdown.

    C4, R7 and N2 seem to be a timing set-up for the soft start. I'll just leave that bit alone.

    The picture was taken under a standard incandescent desk lamp using an iPad. Some sort of strobing/interference? Line frequency is 50 Hz here, but I would have thought the filament would smooth any flicker. A question for another thread, perhaps!

    Thanks again for taking the time to do this. I guess you'll be leaving a reasonably detailed study of a presumably common inverter drive circuit for posterity, anyway. I'll update soon.
  24. Jun 16, 2017 #23


    User Avatar
    Science Advisor

    This post is a bit hard to follow, but explains the route taken to sort out all those error signals and their interactions.. (Also I'm creating it while doing the laundry.:frown:)

    Ahhh, almost. That was the main confusion factor. Pin names are given by the function or action taken when the signal is logically TRUE. The over- bar, or negation, indicates that the function is active when the signal is logically FALSE. Positive logic is assumed unless otherwise noted, i.e. HIGH =TRUE.

    The Fairchild data sheet labels pin 10 as SHUTDOWN with a bar over it, indicating that the function is asserted when the signal is FALSE, or, following convention. LOW.

    Your board schematic around N2 and N3 did not support the above labelling. The device pin names required both SOFTSTART and SHUTDOWN to be TRUE; but N2 does an inversion so that the two signals can never be the same.

    At first, I suspected your schematic might have an error. But tracing the datasheet device schematic, this didn't seem reasonable either due to the way SHUTDOWN is internally connected to SOFTSTART. That's why I searched for a circuit description, to see if it was the schematic, the datasheet. or the pin labelling that was wrong.

    Result was that the pin-10 name was wrong (maybe somebody in the drafting dept. had a hangover that day) -- instead of SHUTDOWN-bar, it should be SHUTDOWN; i.e. it shuts down when TRUE or HIGH.

    One of my earlier posts evaluated IC3-7 (LM358) as Hi = OK, Lo = Error
    4N25 pin 4
    on the controller board: Hi (On) = OK, Lo (Off) = Error
    N1-C(pin2) (and KA3525A-10) Hi = Error, Lo= OK
    N2-C Hi = OK, Lo = Error
    (resets C4, SoftStart, on Error)
    N3-C Hi = OK, Lo = Error (error signal to off-board ckt.)

    And all the polarities and pin numbers finally work out.

    Looking forward to you updates.
  25. Jun 17, 2017 #24
    Well, that's not fair. As if it's not hard enough to troubleshoot without datasheet mistakes!

    New KA3525A and BC547 in. Unit powers up, but output is uneven:


    Output voltage: 77.2 Vac (true rms meter)
    +V1 rail: 143.7 V
    VH and Vs rails both around 12.4V
    The 4V line into IC12 hovers around 3.6V

    The KA3525A chip voltages:
    1 INV 2.4
    2 NINV 2.4

    8 SS 4.8

    10 SHUT 0

    15 +Vin 11.77
    16 Vref 5.0

    The OSC outputs are a 0-12V square wave with approx. 25% duty cycle.

    However, the MOSFET gate drives are all over the place:
    Pair 34/35 average 13V, on the scope a 0-40V wave.
    Pair 36/37 average 8V and are a 0-12V wave.
    Pair 38/39 average 50V and are a 0-150V wave.
    Pair 40/41 average 8V and are a 0-12V wave.

    From previous inverters I've looked at, the 8V readings are the normal ones. The abnormal readings are from those MOSFETs on the VH rail. Since the VH voltage is normal at source, could the excess voltage be backfeeding through the 4148 diodes D29 and 31? They test normal on a static diode check.
  26. Jun 17, 2017 #25


    User Avatar
    Science Advisor

    Please use DC coupling on the scope and document where the zero baseline is on the screen and where the scope GND lead is connected.

    How much load, if any, was connected to the device output?

    Q34/Q35 is the only one that looks bad. It should match Q38/Q39 of 150V.

    Most likely suspects are open D32 and/or any/all of the resistors associated w/ Q34/Q35. Possibly a shorted 1N4148 there.

    Possibly IC12-pin11 and Q31. Check that the waveform and voltage on IC12-11 is similar to IC12-10. The waveforms will be out of phase with each other.

    BTW, the IC12 inputs labelled "4V" are most likely the ORG and RED leads from the KA3525A outputs. Hang the scope on them to verify, if needed.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook

Have something to add?
Draft saved Draft deleted