SMPS bench supply - keeps on popping output rectifiers

[Prehistoricman]

I've been working on repairing/upgrading an SMPS bench supply from Manson. Rated for 20V at 5A. However, with original hardware, it can output up to 30A without completely breaking.

The first time I broke the supply, I replaced its original MBR2045CT (45V, 10A per leg) output rectifier with a TO-247 packaged rectifier from a PC supply. It didn't last 20 seconds before popping the fuse and the mains-side MOSFETs.

More recently, I installed new MOSFETs and a STPS61H100CW (100V, 30A per leg) output rectifier. This lasted about 10 minutes and broke in exactly the same manner. It was under about 100W of load, which is what the unit is rated to output.

Am I doing something wrong here? Do I need to purchase a specific kind of output rectifier for this job?

Edit:
schematic (last 2 pages)
Pictures below

[madires]

A schematic (or pictures if not available) and a list of all your modifications would be helpful.

[Prehistoricman]

It's a MANSON MODEL NSP - 2050 or TENMA MODEL 72-8340. Can't find a schematic, however the last page here has some useful stats.

First mod:
Replaced output rectifier (MBR2045CT) with MBR3045PT
Attempted to increase display update rate (removed, appeared unsuccessful)
1M resistor to ground on output

Second mod:
Replaced output rectifier with STPS61H100CW
Replaced 2x MOSFET (IRF840) with 2x IRFP460
Replaced bridge rectifier (KBL406G) with KBU807


It appears in the picture that the low-side heatsink is touching the output rectifier. However, there is an insulating pad between them and the heatsink is not attached to any traces on the board.

[madires]

Hmm, the IRFP460 has about three times the gate charge of the IRF840 and also switches slower. If the gate drive circuit can cope with that, it might be ok. Replacing the Schottky diode on the secondary side and the bridge rectifier on the primary with higher rated types should be fine. The 1M resistor shouldn't cause any PSU failure, but why have you added that resistor? When the PSU failed, did you draw more current than the rated 5A? How much more?

[Prehistoricman]

The 1M resistor is to stop the output from floating way above ground. I've had an issue in the past where connecting devices to this power supply would cause sparks as the HV output caps discharged (I assume, correct me if this seems wrong).

I was drawing just below 6A at around 16V. Normally, the PSU would go into CC mode at 5.33A, but the current shunt was shorted to test beyond the original limits. After the first mod's failure, I noticed the output rectifier was too hot to touch. After this failure, I noticed the high-side heatsink was vaguely warm and the low-side heatsink was cold. Also the bridge rectifier was cold.

Can I test the gate drive circuit? I hadn't thought that the gates could be an issue until now. I'd assumed since the first mod that the output rectifier failed short, which then caused the MOSFETs to fail short as well. Interestingly, both times, the bridge rectifier has survived.

[madires]

Usually there's a cap between primary and secondary to suppress EMI, but also passing a small current which could be the cause of the sparks. Was the connected device grounded somehow? Via an USB cable to the PC for example?

I don't know how the current sensing/limiting ties into the control circuit, but it might be possible that the missing current sensing at a high output current drives the control including the MOSFETs ouside of the safe operation area. Hence the failure of the MOSFETs. You can check the drive circuit with a scope and differiental probes, but that's something for someone experienced in SMPSUs. The new bridge rectifier is rated for twice the current and about three time the surge current. So it will take more abuse before it breaks. Fortunately it's saved by the fuse. I'd assume that the MOSFETs fail shorted first, causing a surge current through the rectifier (which broke the old lower rated rectifier) and popping the fuse.

Does PSU run fine when you don't short the current sensing resistor and overload the PSU?

[Prehistoricman]

Yes - the exact case I'm talking about was powering a laptop from the PSU and then connecting a monitor to the laptop. There's 5 suppression caps on a board connecting to the output terminals. 2 from negative to ground, two from positive to ground, and 1 between negative and positive.

Hmm... I wouldn't suspect the missing current shunt. Before any other mods, the supply would deliver 8A continuously and 30A for a good 10s or so. After this, it would make a squealing noise until the output was turned off. Before the first mod, this supply has been really abused.

Something I forgot to mention before was that during testing the 2nd mod, odd things were happening. The CC mode was acting real funny and slow. I was powering an inductor as a dummy load, so you would naturally expect the current to be wandering all over the place. The volt/current knobs could be adjusted so that the unit would switch a couple of times per second between CV and CC.

I don't have any spare original MOSFETs, but I do have two of the ones that most recently broke. I could install those along with the original output rectifier and test the unit, as well as removing the current shunt short. How about that?

[madires]

Overloading the PSU causes excessive heat and increased ripple. Two things you normaly don't want to have. Then add the inductive load and watch the magic smoke escaping. Please read https://en.wikipedia.org/wiki/Flyback_diode to understand why you have to be careful with inductive loads.

Yes, try the IRFP460s with the short for the current sense resistor removed.

[richnormand]

Your board looks vaguely familiar. I think I may have repaired a variant of this type a few months ago.

The symptoms were random burning out the output FETs and (on the primary side) burning a control TRIAC and sometime the output fuse. It turned out to be several dried out electrolytics in the feedback control circuit and PSU management IC. If you have an ESR meter I would check the capacitors in the red areas. My issue turned out to be three caps in the middle circled area near the IC. The PSU control circuit and feedback pre-regulation was most likely cutting off and on at random or inappropriate times creating surges.

The PSU used to emit a whine as the load on it increased, and then shut off. Sometime it would not start at all, sometime it would just click once and sometime, under heavy load, the FET would burn out while you heard the inverter transformer emit a harsh shreak and high frequency noise for a second.

No problem after replacing three cap in the center area near the control transistors and IC.

Might be worth your while to listen to the transformer for switching noises while increasing the load gradually and check these caps.

[Prehistoricman]

@madires
Ah yes, I'm well aware of the back EMF with inductive loads. I've had a few mild shocks from inductors.
For the supply's safety, I believe it's fine. There is an output MOSFET which has a protection diode built in. And it's taken many hits from large inductors.

@richnormand
This PSU hasn't whined on anything that's within its spec. I'll take a listen when I power it up next. The only audible noise I've noticed so far is 50Hz humming.

Unfortunately I don't have an ESR meter. My multimeter does have a capacitance range but I understand that may not highlight an issue.

[richnormand]

@richnormand
This PSU hasn't whined on anything that's within its spec. I'll take a listen when I power it up next. The only audible noise I've noticed so far is 50Hz humming.

Unfortunately I don't have an ESR meter. My multimeter does have a capacitance range but I understand that may not highlight an issue.

Most of the time when a cap dries out it might not show outside signs (like a bulging can or leakage). This is true for the smaller, lower voltage ones in particular. The ESR is great because the AC voltage applied is low allowing to do in-circuit tests. To measure the capacitance reliably you have to remove them. Most of the time a high ESR will also be accompanied with a lower capacitance but that is not always the case.

Some ESR meters are quite cheap on ebay and would highly recommend them for this type of work.

Best of luck finding the culprit

[Prehistoricman]

Buzzed loudly and popped as soon as I turned the power on. Same components failed short.

[madires]

The next step is to check the caps as richnormand suggested.

[bktemp]

Loud buzzing could be either failed mains capacitors, or a shorted component, therefore a high peak. So probably the diode failed first, shorting the transformer and causing the mosfets to fail.
It could be a shorted choke (the toroidal inductor) causing a high peak current and therefore overloading diode + mosfets. Or it could be a failed (open) snubber near the diodes, causing overvoltage spikes.

[CJay]

So, you replaced the primary MOSFETs and nothing else there? I'd be spending some effort making sure everything else is good before I replaced the FETs again because you've probably managed to send a couple of hundred volts through all the primary driver components including the driver transformer.

OK, so the control circuit is on the secondary, it drives the MOSFETs with a transformer that has a pair of flying leads.

You can power that controller chip (I suspect it's going to be a TL494) with 12V and check you have a clean switching waveform to the transformer and then on the other 'hot' side of it.

Unfortunately I have seen that kind of arrangement fail because the driver transformer goes faulty.

[Prehistoricman]

The next step is to check the caps as richnormand suggested.

I checked 4 just for their capacitance and they all measured slightly high. Maybe up to 10% high.

You can power that controller chip (I suspect it's going to be a TL494) with 12V and check you have a clean switching waveform to the transformer and then on the other 'hot' side of it.

It is a TL494, nice guess. I put 12.5V into the Vcc and the chip output 4.94V for its 5V reference. C1 and C2 of the chip oscillated at 113kHz with a 61% duty cycle, 0.5VAC. Interestingly, the frequency setting capacitor oscillated slightly on the CT pin - 103kHz 43% 0.002VAC.

I did get an anomaly once. C1 put out 169kHz while C2 was still 113kHz.

Now, reading the datasheet (page 8), I see that C1 and C2 should cycle between 0V and Vcc. And it doesn't appear the chip is doing anything like that.

[h.bashar]

According to the data-sheet, the PWM outputs on C1 and C2 are open collector type so the output waveform should be swinging between VCC and the GND if it was bypassed externally by some resistors to either rails, these are two transistors inside the PWM chip and if current was excessively drawn from them for some reason they might fail. page 9 of the datasheet shows the typical load capacitance on these two pins to be 15pF, now I am not aware of the circuit itself but if this was driving a mosfet with high input capacitance it may fail to reach the rail voltage at the given switching frequency.  I would try to replace the output mosfet with original part number the device came with. And if the input circuitry was intact you can still test the PWM chip basic operation by at least monitoring the outputs of C1 and C2 without the installing any ouptut mosfet but again I am not an exert on this subject but maybe someone can weigh in on this.

[CJay]

You can power that controller chip (I suspect it's going to be a TL494) with 12V and check you have a clean switching waveform to the transformer and then on the other 'hot' side of it.

It is a TL494, nice guess. I put 12.5V into the Vcc and the chip output 4.94V for its 5V reference. C1 and C2 of the chip oscillated at 113kHz with a 61% duty cycle, 0.5VAC. Interestingly, the frequency setting capacitor oscillated slightly on the CT pin - 103kHz 43% 0.002VAC.

It seems to be a pretty standard sort of configuration to have a transformer driving the power switching devices when the control chip is a TL494, I've seen it in various places all the way back to power supplies in power supplies for clone 286 PCs and it just looks 'familiar'..

You can find dozens of PC power supply schematics all over the web which are all using the TL494 in extremely similar configurations (I'd bet the circuit is almost identical around the TL494, it'll just have a few differences on the hot side to accomodate the MOSFETs):

http://320volt.com/wp-content/uploads/2010/12/atx-smps-schema-200x-atx-power-supply-tl494-lm393.png
http://www.eleccircuit.com/wp-content/uploads/2008/08/200w-pc-power-supply-110v-220v-by-tl494.jpg
http://www.eleccircuit.com/wp-content/uploads/2008/07/old-power-supply-computer-by-tl494.jpg
http://danyk.cz/s_atx01g.png

The chips are dirt cheap so you can check it by substitution

Usually drives a couple of transistors from C1,C2 output which in turn drive the transformer to get the switching signal to the power MOSFETs.

Check the drive is clean all the way to the transformer, that both drive signals are the same at the transformer and that you get a clean switching signal out of the other side of the transformer.

On the hot side of that driver transformer there should be a small circuit that 'rectifies' the drive signal to make sure it doesn't go negative and a little filtering. Check those diodes.

As it's using MOSFETs for the main power switching there will probably also be a Zener clamping the gate drive to something like 15V, check they're good as well.

You should be able to follow the PWM drive all the way from the TL494 to the gate of the power MOSFETs with only the 12V supply to the TL494, same again, check the drive signal all the way from the transformer to the power MOSFET gates.

Once you've got that clean, I would remove your mods and get the supply working as it was designed first, draw out the schematic if you haven't found one yet, then try to mod it on paper before you add any solder.

[Prehistoricman]

Usually drives a couple of transistors from C1,C2 output which in turn drive the transformer to get the switching signal to the power MOSFETs.

Oh yes! That makes a lot more sense. I thought that the driver transformer was sending power to the low side!

I found a schematic on the last 2 pages. It appears to be for version 1.1, and mine is version 3.0. I can't find any differences on the board, however.

Signal's getting through fine. Peaks at about 10v on the MOSFET gates. 56kHz or so with a 5% duty cycle. Interestingly, the input to the gate driver is 1VAC 50%. As soon as the signal goes past the PNP transistor, the duty is cut down. Or at least that's what it appears to be.

Checked two diodes - they were fine. I've ordered a component tester that would probably be able to tell me if the zeners are fine.

[oldway]

For me , the problem seems very easy to understand.

With higher current than the nominal, L4 goes to saturation and inductance became very low.

For this reason, peak current in the output rectifier and in Mosfets became very high, as well the rms value of ac ripple,  and  overload heavily  those components.

You must substitute L4 by a bigger one.

[Prehistoricman]

For me , the problem seems very easy to understand.

With higher current than the nominal, L4 goes to saturation and inductance became very low.

For this reason, peak current in the output rectifier and in Mosfets became very high, as well the rms value of ac ripple,  and  overload heavily  those components.

You must substitute L4 by a bigger one.

Good call - if I want to get more power out of this thing, I need a bigger inductor. However, this unit has spent a lot of time outside it 5A rating with original components. And then I place in more heavy-duty components and they fail very quickly under relatively normal conditions.

According to the schematic, L4 is rated for 10A. I don't know if that means it saturates beyond 10A or if the coil will overheat beyond 10A.

Edit:
10A must be its saturation - L4 has 2x 0.7mm solid copper wire.

[oldway]

Upgrading an SMPS bench supply requires to have good knowledge and practice in developping SMPS.
You must also measure all the parameters of the modified SMPS.

Manufacturers don't waste money with oversized components.
If they seems bigger than needed, it is because there is a security margin for reliable working.

[singapol]

Buzzed loudly and popped as soon as I turned the power on. Same components failed short.

This could indicate mosfets oscillating which led to their failure. The problem is not understanding device
characteristics like Vds ( drain to source voltage), see the graphs in the respective datasheet. IRF840 vs IRFP460. Drain current (continuous) for IRF840= 8A. while IRFP460= 30A. Look at Vds vs gate voltage and see how much each device is putting out. For IRFP460 it's way above it's rated 30A (close to 100A). Since the inductor is rated for only 10A so it is saturated currentwise and does not act like an inductor. So you have thermal runaway plus oscillation. A deadly combination. It's not easy to alter the components to stop oscillation (perhaps by simulation) and feed the proper gate voltage to IRFP460. Best is to restore back original IRF840 ,at least you have a working power supply, lessons learned.

http://www.vishay.com/docs/91070/sihf840.pdf

IRFP460- http://www.vishay.com/docs/91237/91237.pdf

[Prehistoricman]

Okay so... I'm trying to understand. Is this right?

The newer MOSFET has a lower Rds-on, therefore more current flows through the transformer. Therefore there is a much higher current on the output rectifier - so much so that it fails.

I'd always assumed that the transformer winding would limit the primary side current.

[singapol]

Okay so... I'm trying to understand. Is this right?

The newer MOSFET has a lower Rds-on, therefore more current flows through the transformer. Therefore there is a much higher current on the output rectifier - so much so that it fails.

I'd always assumed that the transformer winding would limit the primary side current.

Just know that your power circuit was designed around IRF840 so a higher powered mosfet needs
a different gate voltage(less) and  different component values because of device capacitance but I'm not qualified to comment. The thing about mosfet is it's different to a bipolar power transistor in that once turned on it outputs full power so that's why I recommend sticking to original device. I learned this the hard way with bipolars so it depends on what is driving the power devices. Unless you know how to design such circuits. A lower Rds means it's more efficent ,less power loss during switching.

A transformer is a passive component ,it tries to satisfy demand ,failing that it gets hot because the wire thickness and the size of the magnetic core ( the square cross section which the coil is wound around determines the current output can't cope so either the coil wires short (wire insulation melts) or fuse opens. Please read up online for more info.