Hi Richard, I was equally shocked as much as you are, when the 2.5A fuse carried 26A. I was under a slight confusion on that topic since my meter had a max range of 10A and that too had a fuse in it. That fuse had also blown a few times some years back, when the current was much greater. In those old situations, IIRC it blew up at 18 or 19A. Strange that the meter showed the 26A reading. Wanted to confirm whether the meter was acting up. Measured values, even DC current with a lamp load from a battery. Seemed perfect. Nothing wrong with the meter. The next thing that had triggered my attention was that, the meter probes were slightly warm, not hot. That was the reason, I had to yank out the power to the setup. So, my thought was that: There were 2 fuses in the path. one within the meter: rated 10A (did not blow) another 2.5A fuse on the PCB (did not blow). Since the previous instance with the 80NF10 and large heatsinks, I had downgraded the heatsinks also to smaller ones just to see the temperature rise. The heatsinks did become warm, lets say to about 60-70C but it was not for long. Luckily I saved the MOSFET's also by removing power. Noticed that the Drain to GND waveforms had very little ringing (maybe all the energy was getting drained and hence). Thats why I wrote earlier that the Drain waveform resembled much like the Gate waveform itself, without the RCD snubber. Also, since the incident, nevertheless I wanted to see whether my mind was making things up, retried applying power in the same setup, after everything cooled down after about 30mins. The previous situation was replicable 100 percent. Also the waveforms, very little ringing. But did not have much time to take a snapshot or adjust the rigger .. :-( Regarding the snubber, IIRC I did start here IIRC http://www.cde.com/resources/technical-papers/design.pdf did more reading found an excel sheet to calculate the snubber values http://www.nxp.com/documents/application_note/AN11160.pdf NXP's document was not very clear on the parameter terminology, was a bit lost, but found this guy doing a bit of explanation: http://goo.gl/hQpqsZ Infineon states that the RCD snubber is more efficient than the RC snubber in one of their Application Notes: http://goo.gl/tU2vg8 The diodes and the resistors are in perfect shape. No degradation compared to new components. Seems fine. Just for the record, the capacitors are EPCOS box type film type ones rated at 700VAC. Personally, I have never seen these caps fail previously. Now that I have. I must add that, while in the running state I moved the load slightly, before the tick sound probably I heard a "hiss" but not sure from where. Probably, I moved the load accidentally a bit. But don't know whether it is related. But the converter ran for a few more seconds after that sound. But after that the current started to climb fast, before the FET's failed. I am now wondering of a possibility, whether the load got disconnected completely, a large spike introduced in the secondary, causing secondary insulation failure probably ? Need to still see what happened to the transformer. It is indeed a puch-pull step up converter Effective permeability:1600 Effective cross section area (Ae):125mm^2 Core window area (An): 178mm^2 Magnetic path length: (le): 92.2 mm Core volume (Ve): 11.525 cm^3 Primary 2 turns, Secondary 60 turns Calculated primary Inductance was 10.9uH for each segment Measure inductance before tests were 10.6uH. Measured it again today, shows 11uH Flux density: Min:0.074 Rated: 0.099 Max 0.124 (T) Switching frequency:120kHz Winding fill factor:0.213 Current Density:6A/mm^2 Input voltage: Min:9V Rated:12 max:15V Secondary Voltage: 350V Secondary Current:1.4A Core loss: 1.298W Max transformer power:773W Output Power:515W Primary inductance:10.9uH Primary magnetization current: 2.268A Max Primary current amplitude: 45A Primary turns: 2+2 (x41 strands) wire skin diameter: 0.38mm wire dia: 0.4mm Secondary turns: 60 (x2 strands) converted the primary number of turns into an equivalent foil width In none of the situations, the transformer was never more than lightly warm. Far from warm. In the no load situation, both the FET's and the transformer were very lightly warm, with very small heatsinks. But why do you want to have a gap in a transformer ? If it were an inductor, I would agree that a gap could help to avoid core saturation, but in a transformer the energy is extracted through the secondary. If the gap existed the transformer would be inefficient. Right ? The only exception to this rule, I can assume is a flyback transformer. but in many situations people consider a flyback transformer to resemble an inductor more than a transformer. Previous history: Previously worked up a small converter, based on a SG3524 with 2 IRFZ44's and a pair of BC557 to pull the gates down during turn OFF. I had a small PQ26/25 N87 same material from EPCOS. 3 turns primary, 89 turns secondary, designed for 77W. It worked very well. In fact running that with no load and no heatsinks, the MOSFET's were hardly warm. With heatsinks, I was able to drive a load of upto 60W. It ran for many hours without issues. Wanted to push it to more than max limits to see the state, tried a 100W lamp. The transformer started making a huge chatter like a relay and then the MOSFET's exploded. The tracks on the PCB burned and the PCB also burned. I was pretty happy even though it was damaged, it handled the load, the destruction occurred due to the chatter. The 2 sections were secured strong, but it was hard for me to believe chatter could arise still. After this, wanted to change the SG3524 with a microcontroller, do a larger power as well. An old picture: https://app.box.com/s/55cu54tiofcaixwt1w1f56izvhf5r5dk On Tue, Jun 28, 2016 at 2:27 AM, Richard Prosser wrot= e: > Hi Manu, > At 26A I would have thought that your 2.5A fuse would have lasted less > than 5 seconds but there you go. It sounds like you have cooked the > transformer - shorted turns. So the transformer was getting hot. If it > hit its curie temperature then that would have the same effect as core > saturation. > I can't see how both MOSFETs could have turned onat the same time so a > DC component seems unlikely. The measurements on the snubber caps are > of concern though. Are the diodes OK? > > I'm still not sure about the snubber arrangement. When the top MOSFET > turns on the voltage on the drain of the lower one will increase to 2 > x supply voltge by transformer action. This will cause the lower > snubber cap to charge through the diode and then discharge through the > resistor. So the caps will charge & dischrage on every 1/2 cycle. > Looks to me to be quite lossy but there may be a cunning plot behind > it I can't see. > > Can you give some indication regarding the history of the converter. > Is one you are desiging from scratch or one that is adapted from other > equipment? If designing from scratch, can you provide a background > design you've based this on. > I think I remember that it is a step - up unit to ~200V so a shorted > turn on the secondary appears most likely. Your scope plots show > primary voltages - did you take any of the secondary? > Is the core gapped ? Any idea of gap width if it is. How many primary > turns? Did you measure the primary inductance as 10.6uH or is that a > calculated value? > > I'm starting to run out of ideas - I worked on switchmode supplies > about 10years ago but have now moved on. I do remember that the secret > of a good design was the magnetics and that was generally left to the > experts to sort out! As much an art as a science! > > Richard > > > > On 28 June 2016 at 04:32, Manu Abraham wrote: >> I did not realize that the term "normalised was a key player in there. >> Thanks for clearing that one. >> >> Removed both the snubbers, wanted to measure values of the >> components. The 4n7 on one side showed 3.9nF and the other one >> showed 20pF. Strange indeed. The capacitors had no short though. >> Really weird. >> >> As for the MOSFET's didnt have any more spare one's, but had very >> similar but slightly lower voltage ones at hand: STP80NF55 >> Datasheet: http://goo.gl/gtoDJy >> >> Replaced the MOSFET's, replaced the 10A fuse with a 2.5A fuse >> instead. On the transformer, I do have a small auxiliary winding, had >> that initially just in case "if" I required an auxiliary power, but had = not >> come across the need for it yet. >> >> To pull out the transformer from the PTH PCB had been a pain, so left >> it alone for the time being. >> >> Powering it up, the meter showed 26Amps being drawn. I had >> connected the scope across both the FET's drain-source. The waveform >> showed much like the gate waveform, but all the cables were getting >> heated up due to the large current. I thought the FET's went up in smoke >> much like the 80NF10, but the 80NF55 seemed much resilient. But there >> was no output, all this happened in 5 seconds, as I yanked the power >> out of the setup. The auxiliary power was available but nothing on the >> 300V secondary. Seemed like a short in the transformer secondary? >> Wonder, what could've caused that. It cant be the primary, as it is real= ly >> well insulated and thick copper foil, So primary failure is unlikely. Th= e >> auxiliary output had power, so likely the secondary winding alone failed= ? >> >> Pulling the transformer from the PTH PCB turned out to be a nightmare. >> Took many hours to remove it alone. >> >> I have not yet taken the transformer apart, to identify the damage cause= .. >> It is a ETD39 core from EPCOS with a N87 material >> >> core datasheet: http://en.tdk.eu/inf/80/db/fer_13/etd_39_20_13.pdf >> material datasheet: http://en.tdk.eu/blob/528882/download/4/pdf-n87.pdf >> >> 2 turns of the copper foil (~0.4mm thick) , approximated equivalent to >> 41 strands in parallel (hope so). The secondary sandwiched between >> the 2 primaries. >> >> At 120kHz, that core can handle a maximum of about 650W. >> Do you think a larger core is necessary for such small power outputs >> ~=3D20-300W ? When it failed, the load was a mere 23W. >> >> >> >> On Mon, Jun 27, 2016 at 1:40 PM, Richard Prosser w= rote: >>> The key word here is "normalised". Whatever value Rds(on) is at 25C is >>> taken as unity and the plot shows how this changes with temperature. >>> So a graphed value of 1.1 would indicate a resistance of about 15mOhm >>> x 1.1 =3D 16.5mOhm. If the actual value for that device is 10mOhm, then >>> you could expect it to increase to 11mOhm. >>> >>> The data sheet should give you thermal conductivity values that will >>> let you estimate the die temperature but a heatsink temp of 34C or so >>> would normally indicate that there is plenty of headroom. >>> >>> Another thing to check would be the drain peak voltage. Make sure the >>> snubbers are working properly and can handle the power levels >>> reqyuired. If the snubbers go o/c then the peak voltages could be >>> punching through the MOSFET & destroying it. Putting addional R-C >>> snubbers wouldn't do any harm. The MOSFETs are 100V types so you >>> should be OK here provided things work as expected. >>> >>> Also, the snubbers seem to be low values. 4n7 & 100ohms could use >>> quite a lot of power (Although maybe not too bad on a 12V supply). Not >>> sure about the return to the source of the MOSFET either, across the >>> transformer to the plus line would be more conventional?? On the plus >>> side, the snubber current is going to pass through the transformer and >>> so assist the output. Maybe. It may also contribute to the ringing >>> you're getting. Certainly it will show up as a component of the soure >>> resistor current. >>> >>> Anyway, It sounds to me like the inductor is saturating. Can you try a >>> larger core? >>> >>> RP >>> >>> >>> On 27 June 2016 at 19:21, Manu Abraham wrote: >>>> Thank you guys for the info. >>>> >>>> I watched the video. The probe issue according to the video, exhibits >>>> anomalies by around 400Mhz. I am nowhere near that frequency at >>>> 120kHz. The next thing that dawned upon me was to adjust the trimmer >>>> on the probe. but that didn't seem to make much or any effect at all. >>>> Maybe it did make a bit change, but it was not visible enough for me. >>>> The scope probe came with a spring clip, that could be placed on to th= e >>>> barrel. Tried looking at the Gate waveform, but no obvious change at a= ll. >>>> >>>> With a bit of impatience, not seeing the other end, wanted to give it = a run >>>> for a while. >>>> >>>> With a 23W load at the output, ran it for a while at 15% dutycycle wit= hout >>>> the feedback loop. With a thermometer on the heatsink. Starting from 3= 1C, >>>> the heatsink warmed up to 42C. It ran well for 10min. The current draw= n >>>> was climbing, but became steady at around 3.63A. The current drawn was >>>> ramping up and down, but it was hovering around the same values. The >>>> output voltage was slightly lesser than what I expected it to be (~ 28= 0V. >>>> At some point it fell as low as 230V as well.). >>>> >>>> Retried the setup with 20% dutycycle, in the same state. The current w= as >>>> climbing up slowly from 3.6A. Some 3-4mins down the lane, the current >>>> started climbing up wildly, reached about 9.5A (heard a tick sound). >>>> MOSFET's blew, fuse (10A) blew. The temperature on the heatsink was >>>> starting to climb from 33C to 34C. But I presume, the die temperature = was >>>> much higher and hence the result ? The Drain voltage at that time was >>>> about 11.8V, IIRC. >>>> At 20% duty cycle, the out voltage remained nearly moreover the expect= ed >>>> value of 300V. >>>> >>>> Anyway, even with an ample heatsink, need to have a proper and good >>>> feedback loop if the converter is to behave. >>>> >>>> >>>> One thing that I do not understand, the vendor specifies a Rds=3D 15m >>>> But, then they specify in Fig 11: that the normalized ON resistance vs >>>> temperature, Rds @ 25C is 1Ohms. Agreed that Rds is temperature >>>> dependant, but what cant fathom is that if the Rds is nowhere near 15m >>>> in those charts. Wonder how they made up that figure. ?! >>>> >>>> So, given that Rds=3D1.2Ohms. No wonder why the MOSFET blew up. >>>> >>>> Back to measuring drain current again. >>>> >>>> >>>> On Sun, Jun 26, 2016 at 6:36 PM, Denny Esterline wrote: >>>>> Make sure >>>>>> that all probe leads are as short as possible and that there is >>>>>> minimal loop area between the earth lead and the signal lead at the >>>>>> probe. >>>>> >>>>> >>>>> ^^^^^ this. >>>>> And "short as possible" does not mean "clip lead from the probe to gr= ound". >>>>> >>>>> [looking for reference picture...] >>>>> There's a relevant picture on this thread: >>>>> http://electronics.stackexchange.com/questions/221292/tip-barrel-test= -of-oscilloscope >>>>> >>>>> Actually, the whole thread is kinda relevant... >>>>> >>>>> -Denny >>>>> >>>>> P.S. there's a black hole link on that page for the interested. >>>>> https://youtu.be/2vzvWUqUtb8 >>>>> Leads to a series of talks by "Bob Pease and friends" that just swall= owed >>>>> me for a couple hours. >>>>> This specific video at about 8:00 is talking about this specific scop= e >>>>> probe problem. >>>>> -- >>>>> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive >>>>> View/change your membership options at >>>>> http://mailman.mit.edu/mailman/listinfo/piclist >>>> -- >>>> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive >>>> View/change your membership options at >>>> http://mailman.mit.edu/mailman/listinfo/piclist >>> -- >>> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive >>> View/change your membership options at >>> http://mailman.mit.edu/mailman/listinfo/piclist >> -- >> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive >> View/change your membership options at >> http://mailman.mit.edu/mailman/listinfo/piclist > -- > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive > View/change your membership options at > http://mailman.mit.edu/mailman/listinfo/piclist -- http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .