Trace SW3024E MOSFETS blown again after sucessful repair

simat
simat Registered Users Posts: 12
Hi,

My Trace SW3024E has gone and popped again, this time on a different H bridge, last time it was the H-Bridge driving the middle transformer, this time its the 4 x MOSFETS in the H-bridge for the smaller transformer. Gate-Drain-Source shorted together on 3 of them, the other has Gate+Drain shorted and has blown the PCB track leading to the drain pin.

The motherboard, display board and relay board still power up ok with the mosfet board removed and +24v injected directly. Relays works, control logic seems ok, voltage measurement ok. No shorts on the 240v side.

I'd replaced all 52 MOSFETS, all the TC1427 low side drivers, all the TC4804 high side opto isolated drivers, any burnt gate resistors, and rewound the common mode choke L1 on the relay board that had gone short, as well as add a couple of 275V MOV's to the input to catch any spikes.

Why would the MOSFETS go dead short ? Does the gate pin need pulling down to ground via a resistor ? My guess is that the low and high side of the H-Bridge were on at the same time thus causing a fatal short that literally melts the inside of the MOSFET. All 4 drive pulses looked fine on my scope with no overlaps.

Does anybody have a circuit diagram for the main board ? I need to see how the low and high side drivers are driven, are they independant of each other ? or just the inverse of each other ?

The unit was charging the battery at the time.

Anyone got any ideas ?

Regards

Simon.

Comments

  • Joe_B
    Joe_B Solar Expert Posts: 318 ✭✭✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    Why would the MOSFETS go dead short ? Does the gate pin need pulling down to ground via a resistor ? My guess is that the low and high side of the H-Bridge were on at the same time thus causing a fatal short that literally melts the inside of the MOSFET. All 4 drive pulses looked fine on my scope with no overlaps.

    I have not been following your thread closely but I can offer some insight. There are lots of ways that a MOSFET can short all three terminals but it boils down to a couple of things. One is if Vgs goes over 20 volts (+ or -) you can blow the gate oxide layer and then all hell will break loose. Another way is if the power dissipation gets so high it melts the die. This can occur if the gate drive does not fully enhance the FET and it goes ohmic or something in the drain - source circuit that is causing the Pd to go up. Of course there are other ways to do it like if a snubber network failed, then Vds might be violated.

    Oh and yes, the gate needs to be actively driven high and low, it looks like a big capacitor so it must be charged to turn on and discharged to turn off.
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Unfortunately, the TC4804 high side driver wasn't very good. Trace designed those out and completely re-desigined the
    gate drive. Not sure if that was the reason your SW blew again though.

    That must be quite an old SW inverter I am guessing. I am surprised that you were able to find those parts still,
    but I suppose there are some around still or maybe some other company started to make them ?
    The original TC4804 was from Teledyne. Next time, you might want to try something like the HCPL-3120 or equivelent
    but the output side will have to be re-wired slightly.

    Another thing that can break is the gate drive resistors or diodes in the charge pump or who knows what else.
    Gotta check every component.

    boB
  • RCinFLA
    RCinFLA Solar Expert Posts: 1,484 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Did you disconnect any of the transformer feed wires when you did the MOSFET board repair. Polarity of these transformer wires matters (both primary and secondary sides). Having just one of the transformers with their wires swaps will cause this kind of problem.

    You can disconnect the 12v power to the relay power board to prevent the inverter engagement with AC mains, then check the phase syncronization, waveform shape, and amplitude match between the AC grid input and the inverter output on a scope. Use neutral wire as your scope ground.

    I have not worked on that older model but I believe they have phasing calibration adjustments on the controller board. You really should not mess with those adjustments unless you have the service test software to guide the adjustments.

    On AC side, baby bear puts out about +/-15 volts, mama bear puts out about +/- 45v and papa bear puts out about +/- 135 v. Actual voltages and number of steps vary depending on AC grid voltage and battery voltage. Secondaries are connected in series so not easiest thing to look at individually with scope.

    Don't know what kind of snubbers that model has but a bad cap or blown resistor in the snubber can repeated subject the mosfet's to avalanche breakdown. You can look at overshoot. Scope ground to battery negative, scope probing at transformer lead, one side at a time (or both at same time with dual channel scope).
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Thanks for the replies.

    I hadn't swapped any of the transformer wires, they were left in place during the MOSFET replacements. The unit ran fine (when it was running) even hung a 2kw kettle off it at one point to test under a reasonable load. Snubber networks had been replaced as the two resistors had burnt, these resistors are still ok and the capacitor reads correctly.

    RCinFLA - do you have a schematic for the SW motherboard ?, I need to find out how to MOSFET drivers are driven, are they individual or inverse of each other per H bridge ? I'm also thinking of building a new MOSFET board with modern components that could be retrofitted - to get rid of the TC4804 high side drivers.

    As the later models didn't have the TC4804, is this what the 2966B daughter board was designed for ? (i know it won't fit mine)

    S.
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    simat wrote: »

    As the later models didn't have the TC4804, is this what the 2966B daughter board was designed for ? (i know it won't fit mine)

    S.

    I don't remember the board number, but that's what the daughter boards are for in the more recent units.

    boB
  • RCinFLA
    RCinFLA Solar Expert Posts: 1,484 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    I don't have a schematic for the controller board.

    The usually way the sequence of the H-bridge MOSFET is as followed.

    Defining H mosfets a left highside, left lowside, right highside, and right lowside.

    There are three states to each transformer primary driver.

    1) Battery + on left transformer lead, Battery - on right side transformer lead.

    2) Zero (near short) across the primary side of transformer

    3) Battery - on left transformer lead, Battery + on right side transformer lead.

    The original low frequency SW inverter system uses three transformers with their secondaries wired in series. The system relies on transformers with tight coupling between their primary and secondary windings, yielding a low leakage inductance. What this means is if you short the primary, the secondary looks like a near short in the series stacking. It actually has series resistance a little higher then the secondary wire resistance and a small amount of series inductance.

    As indicated in the previously attached spreadsheet, these three voltage states (+, zero, -) on the three transformer primaries allows 3^3 = 27 possible output voltage steps across the series connected secondary stack. (inverter will not necessarily use all 27 steps. The number of voltage steps used depends on AC line voltage and battery voltage. Lowest steps per full AC cycle occurs with highest battery voltage coupled with lowest AC line voltage, which yields the 'rattiest looking stepped sinewave)

    So the H bridge is putting +24v, short, or -24v across the primary of transformer for the 24v model.

    N-channel MOSFET's are used for all four H bridge MOSFET's because they provide the lowest Rds-ON for smallest semiconductor die area and thus cost. This creates the need to take the gate voltages on the high side MOSFET's ten to twelve volts above the battery voltage to fully switch them on.

    Older high side drivers used opto-isolator on input to high side driver stage. Newer high side drivers use a differential amp with current source outputs (within the I.C) to transfer the logic state to the upper floating, highside driver circuitry

    Whether old or new type high side driver, they both rely on 'bootstrapping' to create the floating 10v to 15v DC supply to run the highside driver stage. There is typically 12v to 15v fed through a blocking diode to the high side driver Vcc+. The "ground" of the highside driver is connected to the junction of the upper and lower MOSFET's on half of the H bridge. When the low side MOSFET is on, the bypass cap across the highside driver supply is charged through the blocking diode to 12 v range. When the lowside MOSFET lets go and high side driver switches on the highside MOSFET, all the powering of the high side driver circuitry and MOSFET gate charging is supplied by the bypass cap on the highside driver.

    This highside driver supply bypass cap is, in effect, a temporary power supply for the high side driver circuitry. Because the highside driver is running on the bypass capacitor charge it will 'fizzle' verses time the high side MOSFET is switched on so it needs to have its charge refreshed by the next lowside MOSFET conduction cycle. The Trace inverter, being a low frequency design, requires this high side driver supply bypass cap to hold enough charge to power the high side driver for a much longer time then in high frequency switching power supply designs, which is the normal target market for high side drivers. Also, with so many parallel MOSFET's connected together there is a large gate charge capacitance that need to be charged from the bootstrap supply cap. The Trace inverter needs a high 'uF' cap for this bootstrap cap.

    One thing you might want to replace are these highside driver bypass caps. If they degrade due to age, they might not hold enough charge to supply the highside driver for the relatively long period of time in the Trace SW inverter design.

    Getting back to your original question, the cycling of the H-bridge always cycles through +battery, zero, -battery on primary of transformer.

    The zero, shorted state, can be accomplished by both lower MOSFET turned on, or both upper MOSFET's turned on. The decision to use both lower MOSFET, or both highside MOSFET's for zero voltage transformer state depends on several consideration. 1) Last 'ON' state (which side of transformer was pulled high or low), amount of possible cross conduction (releasing highside MOSFET conduction is longer then activation of lowside MOSFET conduction which creates potential high current 'shoot-through' current), and finally not asking for too long time since last bootstrap capacitor refresh. The low side driver usually has an intentional delay built in to compensate for the longer high side MOSFET turn off time. Changing out the type of high side driver is not a good idea since the circuitry might have been delay compensated for the particular low side/ highside driver combination.

    The bottomline is you may see the 'zero' state as seemingly random occurance of both highside and both lowside MOSFET conductions.

    The best way to look at the differential voltages is with a dual channel scope using one channel for one side and other channel for other side of transformer (or gate driver output observation) and putting scope in ch A - ch B mode. You have to make sure you do not reduce the V/div scale to point you put the scope channel into clip.

    The physical layout of the SW inverter is not the easiest thing to trouble shoot. Particularly the placement of the MOSFET board. Be careful trying to probe the buried connections that you don't accidently touch something that might bite you.
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    RCinFLA wrote: »
    One thing you might want to replace are these highside driver bypass caps. If they degrade due to age, they might not hold enough charge to supply the highside driver for the relatively long period of time in the Trace SW inverter design.

    I had replaced all of these with standard 22uF 25V elect caps, would increasing the capacitance help ? I should probably scope this part to see what the bootstrap voltage does.

    I was also considering adding a 2k2 resistor between the Gate and Source on each bridge to help with the turn off, but I guess this could alter switch on / off times and may have a detrimental effect.

    I had noticed from my initial testing that if the transformer leads don't have a dummy load connected and they are allowed to float around, that the TC4804 must over voltage somehow and destroys itself. Working TC4804's that had an output on Pin 8 later stopped having that output, may be the ground pin of the TC4804 requires a 10k resistor to always keep a ground connection, I say this because when the MOSFET blew the TC4804 GND was no longer connected and that caused further damage.

    I have a self draw basic schematic of the MOSFET board - see attached.

    Attachment not found.
  • RCinFLA
    RCinFLA Solar Expert Posts: 1,484 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    The load on the driver is a short rush of current to move the total MOSFET gate load. With all the parallel MOSFET's and their Miller effect there is probably the equivalent of about 0.05 uF to charge or discharge on the transitions. Putting a good quality 0.47 uF mylar cap in parallel with the 22 uF's may help supply the impulse current as the 22 uF electrolytics have a higher series resistance. 2.2k in parallel with gate won't do much of anything to help the turn off. The diode bridging the 220 ohm resistor is to accelerate the gate discharge.

    From your schematic there is no snubbers across the transformer. You might look at the edge voltage spikes during the switching to ensure the spikes don't exceed the 100v breakdown spec of the MOSFET's.

    Chances are the inverter phasing adjustment is out of wack causing the inverter to be slightly bucking the grid AC.
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    From your schematic there is no snubbers across the transformer. You might look at the edge voltage spikes during the switching to ensure the spikes don't exceed the 100v breakdown spec of the MOSFET's.

    The only snubbers on this inverter are the ones on the relay board post L1 and L3 common mode chokes (R1+R2+C3 and C1 and C2), would adding snubbers across each of the transformers in series help ?

    I'll add the 0.47uf mylar caps across the TC4804 bootstrapper cap, would replacing the diode bridging the 220 ohm in the gate drive with something better than a 1N4148 help ?

    Also can I measure if the phasing is out with the scope ?
  • RCinFLA
    RCinFLA Solar Expert Posts: 1,484 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    simat wrote: »
    Also can I measure if the phasing is out with the scope ?

    You should be able to remove the 12vdc power to the power relay board. This should keep them running independently to allow you to look at grid AC waveform and inverter waveform on ch1 and ch2 of scope.
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Nice job on your schematic, Simat !! I don't seem to remember any snubbers across the primaries but don't think they would be necessary.
    Unloaded, things should work fine as any inverter like this needs to be able to put out a waveform when unloaded.

    Phasing should not have to be measured as long as you didn't re-wire things. The 0.47 uF cap is a good idea.

    But those Teledyne high side drivers had problems and SW's using them often broke. How old are the ones you replaced ?? Do they have a date code or year printed on them ?

    Also, how long are your battery cables ? Too long of cables can cause high ripple voltage across the battery terminals when drawing lots of amps.

    RC, nice waveform spread sheet ! Did you calculate those or maybe read the table ROM in the SW ?

    boB
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair
    boB wrote: »
    But those Teledyne high side drivers had problems and SW's using them often broke. How old are the ones you replaced ?? Do they have a date code or year printed on them ?

    Also, how long are your battery cables ? Too long of cables can cause high ripple voltage across the battery terminals when drawing lots of amps.

    All of the TC4804's were replaced with new, the old ones where 1995 week 44 so they lasted well. My battery leads are 5 metres long and taped together so that they run in parallel.

    I'm not sure if I've complicated things by replacing the RFP70N06 with PSMN6R5-80PS mosfets, as the gate charge times differ, see below

    MOSFET RFP70N06 PSMN6R5-80PS
    Vds 60V 80V
    Id 70A 100A
    Rds(on) 0.014 0.0059
    Qg(tot) 185nC 71nC Total Gate Charge
    Td(on) 12nS 26nS Turn-On Delay Time
    Td(off) 40nS 57nS Turn-Off Delay Time
    Tr 50nS 24nS Rise Time
    Tf 15nS 22nS Fall Time

    I figured that at this low frequency a few nS wouldn't make much difference, but then the specification of the 220 ohm resistor in the gate drive was probably matched to the MOSFET specification ?
  • Joe_B
    Joe_B Solar Expert Posts: 318 ✭✭✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Your substituted FET has a much lower Qg than the original, I think your gate drivers will handle it just fine if they drove the 70N06's. Rise and fall times look OK, if you add Tr and Tf, that will give you the total time it is ohmic and the new FET looks faster overall.
  • RCinFLA
    RCinFLA Solar Expert Posts: 1,484 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    As MOSFET technology evolves newer MOSFET's are generally getting lower Qg (lower junction capacitance) for given Vds breakdown and Rds-ON specs.

    The relationship that always hold in relative though is the lower the Rds-ON the greater the Qg, the greater the Vds breakdown the greater the Qg for the same technology/generation of parts.

    Lower Qg is generally better as it requires lower gate drive power. However there maybe cases in a particular design the too fast a rise/fall time might cause problems with some other section of circuits. For example, if the stacked MOSFET react with the old driving circuits to increase the period during transitions where both MOSFET's in the series momentarily are ON at the same time due to the faster turn off or turn on. This puts a short time period short circuit across the supply through the MOSFET's and is called shoot-through. Very high current will pass through MOSFET's for short period of time that may damage them.

    The most fool proof way to avoid shoot through is to purposely have the drivers create a short dead space (both MOSFET's off). For these sized MOSFET's the dead time might be something like couple hundred nsecs. Downside of dead space is releases the transformer to unloaded ringing that might cause overvoltage spikes on the MOSFET Vds.
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    I think I may have it, the trace on the outputs of the gate drivers especially the TC4804's had a spike that varied in height at the edge of each pulse, the power supply was also noisy, so I've replaced all of the electroyltic capacitors on the main board (theres not that many) and the output is now very defined - nice. Just need to fit the mosfets and give it a try.

    I'll still fit the 0.47uF polyesters at the driver supplys.
  • simat
    simat Registered Users Posts: 12
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    It's alive again !! and I actually found the cause in the process. When I first repaired this inverter due to a short on the common mode choke on the relay board, i removed all of the TC1427's and TC4804's as well as other items detailed in this thread, well it looks like one of the plated through holes must have got damaged whilst removing the TC1427 on the small transformer drive, It was on PIN 7 - basically open circuit and not connected to the gate at all, this would have had the effect of not switching off the gate on the mosfet (or leaving it floating around) thus causing the shoot through and the bang.

    All tracks have been checked and wire added to risky sections. I've also added a 10k resistor to pull down the output of the mosfet drives (Pin 5 and 7 to GND on TC1427, between pins 6 and 8 on the TC4804's) so that if one of the drivers fail the mosfet gate isn't left hanging around too long.. I've also added the 0.47uf across the elect capacitors near each TC1427 and TC4804.

    So just to recap. I've replaced the following

    1, All electrolytic capacitors on MOSFET and Main Board - not that main and cost is low
    2, Added 15uf 25v Tantalum capacitors to the 5v, 12v, 16v (TC1472), 16v (TC4804) and 24v rails on the main board (PSU section is top right hand of board) - add across the capacitors near the transformer with 4 x diodes side by side.
    3, Added 150uf 63V Electrolytic capacitor to the 16v rail that feeds the TC1427 as it didn't have one on the main board and it relied on the 22uf 25v ones on the MOSFET board.
    4, All Mosfets on Mosfet board
    5, Damaged 47r resistors (burnt)
    6, All TC1427 on Mosfet board - DIL sockets added to aide future works.
    7, All TC4804 on Mosfet board - DIL sockets again added.
    8, Added 0.47uf polyester (mylar) across TC1427 and TC4804 supplies.
    9, Cleaned and resealed the boards.

    The lot was tested without the transformers connected 1st, it's was run off a 24v supply with a 3A fast blow fuse, the transformer leads from the mosfet board were connected to a 20r 50W resistor to take the place of the transformer load. All seemed ok so I've connected the transformers, connected a much larger battery and bit the bullet - no bang yet. Battery full charged now and inverter section works ok too, output voltage as expected, in fact the VAC voltage reading on the inverter display is much more stable since replacing the electrolytic capacitors on the main board.

    Thank you to the others who contributed to this thread, I hope this thread can be off use to someone else in the future.
  • solar_dave
    solar_dave Solar Expert Posts: 2,397 ✭✭✭✭
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    simat perhaps you should offer a repair service!
  • lindsley
    lindsley Registered Users Posts: 6
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Hi members,
    I have a SW3024E that has just gone bad.
    When its switched to inverter mode with no load it hums heavily and draws in excess of 160 amps from the battery.

    It turned out that the choke on the output of the inverter output got shorted.

    How can I secure a copy of the servicing Manual for this unit?

    Grateful for any help available.

    Lindsley
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Even for Inverters long out of production--The manufacturers have never released their service manual to the wild.

    And parts can be difficult to find.

    I guess you are out of Georgia?

    If anyone knows of a company that services these inverters in Lindsley's region--Welcome to leave recommendations in thread (or PM to Lindsley).

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • lindsley
    lindsley Registered Users Posts: 6
    Re: Trace SW3024E MOSFETS blown again after sucessful repair

    Thank you Bill for the information.
    I am not from Georgia, I live on the Island of St.Lucia in the CAribbean.

    Lindsley

    BB. wrote: »
    Even for Inverters long out of production--The manufacturers have never released their service manual to the wild.

    And parts can be difficult to find.

    I guess you are out of Georgia?

    If anyone knows of a company that services these inverters in Lindsley's region--Welcome to leave recommendations in thread (or PM to Lindsley).

    -Bill