Diagnosing and Repairing a Faulty Inverter: Tips from a Scientist

[Guineafowl]

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

 

[Guineafowl]

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.

 

[davenn]

Don't all answer at once, folks!

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

circuit info would be extremely helpful for us to help you

 

[Guineafowl]

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

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.

 

[dlgoff]

Have you checked the circuit board traces to see if any are bridged together by something foreign; like maybe a flake of solder?

 

[Asymptotic]

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?

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?

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.

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.

 

[Tom.G]

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.

 

[Guineafowl]

Have you checked the circuit board traces to see if any are bridged together by something foreign; like maybe a flake of solder?

Yes, but I'll check again.

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

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

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

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.

 

[Guineafowl]

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.

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.

 

[Guineafowl]

Update: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.

Thanks

 

[Tom.G]

Might as well replace both of those transistors.

Found a datasheet -- kind of.
http://www.datasheetarchive.com/pdf...d9d46a9f50b2efc5e509f86fc&type=M&query=2SC124

Ratings:
Silicon NPN, 40V, 25mA, 250mW, V_{CE sat}=10V, h_fe=90
Apparently discontinued about 1975. But the board is dated 2004. Strange.

Pinout::
http://www.pinout.net/element_browse.php?conid=25694
(This site also shows 2N2218 as a substitute w/pinout below, but in a metal case.)

1 = Emitter
2 = Base
3 =CollectorSubstitution Guides:
http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA801000-64.pdf
2SC536, 2SC403B, 2SC372, 2SC645, 2SC300

http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA868000-91.pdf
2SC1359, 2SC1740S

http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA1203000-103.pdf
BC546

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".

 

[Guineafowl]

Might as well replace both of those transistors.

Found a datasheet -- kind of.
http://www.datasheetarchive.com/pdf...d9d46a9f50b2efc5e509f86fc&type=M&query=2SC124

Ratings:
Silicon NPN, 40V, 25mA, 250mW, V_{CE sat}=10V, h_fe=90
Apparently discontinued about 1975. But the board is dated 2004. Strange.

Pinout::
http://www.pinout.net/element_browse.php?conid=25694
(This site also shows 2N2218 as a substitute w/pinout below, but in a metal case.)
View attachment 205435
1 = Emitter
2 = Base
3 =CollectorSubstitution Guides:
http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA801000-64.pdf
2SC536, 2SC403B, 2SC372, 2SC645, 2SC300

http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA868000-91.pdf
2SC1359, 2SC1740S

http://datasheet.datasheetarchive.com/originals/distributors/Datasheets-X2/DSA1203000-103.pdf
BC546

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.

 

[Tom.G]

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.

 

[Guineafowl]

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.

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...

 

[Tom.G]

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.)

 

[jim hardy]

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 don't 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.
EDIT AGAIN
See if 2N3904 looks like a potential substitute. It's a pretty stout transistor and very common,
https://www.onsemi.com/pub/Collateral/2N3903-D.PDF

 

[Tom.G]

Avoid the RX10K scale because it might use a 9 volt battery , typical limit for reverse E-B is 6 volts..

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

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

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!

 

[Guineafowl]

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.

 

[Guineafowl]

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.

 

[jim hardy]

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!

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.

RX1 Ohmmeter Test would not be safe for this oddball device.

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..

 

[Tom.G]

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.

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:

http://www.st.com/resource/en/application_note/cd00003778.pdf

BTW, the photo of your schematic has some "interesting" lighting. How come the multi-color lights?

 

[Guineafowl]

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:

http://www.st.com/resource/en/application_note/cd00003778.pdf

BTW, the photo of your schematic has some "interesting" lighting. How come the multi-color lights?

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.

 

[Tom.G]

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.)

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.

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.

 

[Guineafowl]

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.)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.

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.

 

[Tom.G]

Pair 34/35 average 13V, on the scope a 0-40V wave.

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.

 

[Guineafowl]

Scope DC coupled, baseline in the centre, ground lead attached to MOSFET body, probe to gate. 5ms and 50V/div.

34/35:

36/37:

38/39:

40/41:

No load on output. Output was measured with ground on neutral and probe on live. If I put a small load on (40W 230V glue gun) it looks much the same.
IC12 pins 11 and 10 are 0-12V antiphase square waves that grow beautifully in duty cycle upon power-on - the soft start, I suppose.

The bases of transistors 28 and 29 are not being driven evenly. The base of Q28 matches IC12-10, but that of Q29 is in phase but is supressed to 2V, not 12. Similarly, Q31's base only reaches 2V and the low portion is very blurred:

Grounded at main board ground, top trace is on base of Q31, bottom Q30. 10V and 5ms per div.

I'm going to swap out Qs28-31 and the signal diodes, but this won't be for a week or so. I'm taking my old Dad fishing on the Isle of Lewis. In the mean time, thanks yet again for your help.

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.

 

[Tom.G]

You're welcome. Actually it's things like this that keep my brain working in retirement. (Itself not an easy task!)

Scope DC coupled, baseline in the centre, ground lead attached to MOSFET body, probe to gate. 5ms and 50V/div.

There be an instrumentation confusion. The IRF640 (Q34 - Q41) has the mounting tab internally connected to the Drain. You are measuring the Gate drive with reference to the Drain, that's why the 150V offset. The measurement reference should be on the Source pin of the MOSFET you are measuring.

The Base drive voltages on Q28 - Q31 appear normal, not worth replacing at this time. Notice that Q29 and Q31 are NPN with the Emitter grounded and the Collector driving the MOSFETs, yielding an inverting, Common Emitter circuit. Q28 and Q30 are PNP with Collector grounded and the Emitter driving the MOSFETS, yielding a non-inverting, Emitter Follower circuit.

The output stage MOSFETs **may** need a minimum load to switch as expected. I would have expected the glue gun to be adequate. However, try using a heaver load to check for a reasonable output waveform. When checking the output waveform, you will have to measure across the load or the inverter "output" terminals.

If the waveform still isn't good, follow the suggestions in my previous post (https://www.physicsforums.com/threads/troubleshooting-an-inverter.917080/page-2#post-5785590)

(Hope you catch LOTS of fish!)

 

[Guineafowl]

You're welcome. Actually it's things like this that keep my brain working in retirement. (Itself not an easy task!)There be an instrumentation confusion. The IRF640 (Q34 - Q41) has the mounting tab internally connected to the Drain. You are measuring the Gate drive with reference to the Drain, that's why the 150V offset. The measurement reference should be on the Source pin of the MOSFET you are measuring.

The Base drive voltages on Q28 - Q31 appear normal, not worth replacing at this time. Notice that Q29 and Q31 are NPN with the Emitter grounded and the Collector driving the MOSFETs, yielding an inverting, Common Emitter circuit. Q28 and Q30 are PNP with Collector grounded and the Emitter driving the MOSFETS, yielding a non-inverting, Emitter Follower circuit.

The output stage MOSFETs **may** need a minimum load to switch as expected. I would have expected the glue gun to be adequate. However, try using a heaver load to check for a reasonable output waveform. When checking the output waveform, you will have to measure across the load or the inverter "output" terminals.

If the waveform still isn't good, follow the suggestions in my previous post (https://www.physicsforums.com/threads/troubleshooting-an-inverter.917080/page-2#post-5785590)

(Hope you catch LOTS of fish!)

Back now! I highly recommend the Outer Hebrides if you like fishing. Hundreds of lochs and no-one about - just stop the car, hike across and cast in a fly. No licences/permission/hassle. The wild trout are beautiful there.

Anyway... I put in new 2N2222As at Q29 and 31, as the ones in there were different - perhaps someone was here before me. Also replaced D29 and 31 4148s. Output voltage is only 40Vac regardless of load but looks even:

I assumed the body of a MOSFET would be connected to source, by analogy with metal casings and earth/ground, etc. I was obviously wrong. Here I scoped Q38 gate (top trace) and Q40 gate (bottom trace) with reference to circuit ground:

Q38 gate (top) goes from 0V to about 60V (note messy trace) while Q40 has a much neater 0-12V wave, in antiphase.

Since the output waveform is approximately correct, could we assume there's a fault in the output voltage regulation circuitry? IC3 (358) appears to do this. Voltages:
1 - 2.84
2 - 2.50
3 - 2.50
4 - 0.00

5 - 0.18
6 - 0.00
7 - 10.46
8 - 11.68

Pin 5 appears correct since the 2.4V from IC13 voltage reference is divided through R53 and R60 to give about 0.18V. Why is there 0V at pin 6?

 

[Tom.G]

IC3-6 (358) is the output current sense filtered with a 2 second time constant. If there is no load, there won't be any voltage there.

Ah-Ha! For Q29, Q31 I notice '2N2222' is penciled in. Something wrong there. The gates of Q34, 35, 38, 39 should approach "+V1" supply (200V) during normal operation. The 2N2222 is rated for 75V maximum. They are breaking down and clamping the Gate voltage, which means that's as high as the Source voltage of Q34,35,38,39 will go. (The Source won't go higher than the Gate because there wouldn't be any Gate voltage left to turn them on.)

Looks like your next job is to find some 300V NPN transistors.

An earlier scope photo showed Q38,39 having 150V drive (referenced from Gnd.) and Q34,35 having 40V drive. I would hazard that Q31 was a 2N2222 and Q29 had the correct part installed.

 

[jim hardy]

Looks like your next job is to find some 300V NPN transistors.

2N3439 ?
https://www.microsemi.com/document-portal/doc_view/8830-lds-0022-pdf

 

[Guineafowl]

So 'we' (you) might be getting somewhere here. I should have clarified the transistor arrangement, and should also remind you that the schematic refers to an earlier model, although by your reckoning the 2N2222As would not have worked even on that model.

At Q29 there was an A42 (http://cdn2.boxtec.ch/pub/diverse/A42.pdf) - Max voltage 300V
At Q28 there was a 2N2907A - this appears to be OK.
At Q31 there was an S9013, max voltage 40V
At Q30, another 2N2907A.

I'm thinking...

1. The 2N2907A PNPs are correct.
2. Some eejit put in the S9013, a general purpose NPN, where there should have been an A42.
3. The S9013 has been clamping its own Q34/35 as well as Q38/39, negating the efforts of the correct A42.
4. By replacing the S9013 and A42 with two 2N2222As, I have also been an eejit.

I will try to get some A42s, or failing that, will follow Jim's recommendation (thanks for that, by the way Jim.)

 

[Guineafowl]

Update -

Ordered these:

http://docs-europe.electrocomponents.com/webdocs/0f9b/0900766b80f9b314.pdf

Fairchild KSP42BU

They come in convenient packs of 10. I think they'll be OK - there's a 42 in the name, at least...

 

[Tom.G]

Fairchild KSP42BU

Those should do the job. They appear to be clones of the MPSA42 that was in Q29. The 2N3439 that @jim hardy suggested should also work, they are very similiar, perhaps a little more robust; just be aware that the 2N3439 is in a TO-5 TO-39 package (metal can), not the plastic TO-92 package.

p.s. Since the IRF640 is also penciled in, I wonder what the originally part was. Probably doesn't matter if it works!

 

[jim hardy]

perhaps a little more robust;

I sawed one open. Collector was bonded right to the metal can so it should have better thermal capability than the plastic ones.. Feel of them, if not hot to touch you'll be fine with the little guys. And you already know about EBC vs ECB packages...

Been fun watching you guys. Congratulations on your progress !

 

[Guineafowl]

This is the cause of all the trouble. I replaced the whole set with IRF640N.

It's always the output MOSFETs - couldn't you build a more robust inverter with an IGBT or just a much larger MOSFET with proper heat sinking? Isn't a row of piddly little units bolted to the case rather fragile? Seems to be from my experience - most inverters I've repaired (it's an odd hobby) have failed in this way. Some just need new MOSFETs, others have lost some of the drive circuitry.

Another thing - is there a better way to isolate the driver ICs and other semiconductor devices from a rogue short or overvoltage condition? I suppose surface-mount fuses would fatigue, but what about a MOV or two? SCR crowbar? These mass-produced inverters seem to lack provisions to a) stop the MOSFET from failing and b) stop the fault from propagating upstream. In contrast, even a cheap SMPS wall-wart will sit there chirping with the output dead shorted, and manage not to blow itself to bits.