Strange MOSFET driver behaviour - Microchip TC4426
in [Design]
|
Prev: Please, DO NOT forget the only thing that matters on this planet........................……..
Next: Adios, amigos...
From: legg on 10 May 2010 23:33 On Mon, 10 May 2010 09:53:57 -0700 (PDT), Didi <dp(a)tgi-sci.com> wrote: >Not so long ago I put them in a design (the other Microchip part which >has >made it into my designs is an I2C EEPROM). >Worked as expected while driving tiny MOSFETs (two in an SO-8). > >But when it came to an IRF540 - plain old IRF540 - they failed >miserably. >I first looked at them driving the IRF "empty", i.e. the drain >hanging, source >practically grounded (well, via 0.1 Ohm or so). Never got past this >with this >driver. > >The edges got really bad - 500 nS if not worse, way beyond spec. I >checked >what happened with a series 10 Ohm resistor, nothing worth noting. OK, >at >the driver output the initial perhaps 1/5th or 1/4th of the voltage >excursion >got better. > >Before I began trying things out etc. I replaced the TC4426 with a >good old >MC33151. Worked same as it has worked last 20+ years for me, well >under >100 nS (was something about 50-60). > >I did not pursue it further, there was no point stopping at it. Got it >working >with the 33151, which was designed in times when apparently someone >would have noticed if a design did not work :-). > >But a few days later I am still curious what was that. Looked as if >some >foldback current limiting - wanted by the designer or not - took >effect, perhaps >I could have eliminated it with a larger current limiting resistor so >things would >get usable (much worse than with the 33151, obviously, it would have >taken >perhaps 50+ Ohm). > >Any thoughts? > >Dimiter As was related in an earlier thread, the input hysterisis of the 4426 and similar parts is not enforced on the input signal pin, but after a linear input stage comprised of an nmos fet. This means that input hysterisis cannot be specified for these parts as a percentage of the input signal logic level - so input noise and ground bounce on the devices' supply pins may be an issue. In a dual driver situation, you might even find that one driver is upsetting the other, so check the the input and output state of the static side when the funny behavior is occuring on the other. There may be symptoms there, that you wouldn't normally look for, or expect. RL
From: Didi on 11 May 2010 03:29 On May 11, 6:33 am, legg <l...(a)nospam.magma.ca> wrote: > On Mon, 10 May 2010 09:53:57 -0700 (PDT), Didi <d...(a)tgi-sci.com> wrote: > >Not so long ago I put them in a design (the other Microchip part which > >has > >made it into my designs is an I2C EEPROM). > >Worked as expected while driving tiny MOSFETs (two in an SO-8). > > >But when it came to an IRF540 - plain old IRF540 - they failed > >miserably. > >I first looked at them driving the IRF "empty", i.e. the drain > >hanging, source > >practically grounded (well, via 0.1 Ohm or so). Never got past this > >with this > >driver. > > >The edges got really bad - 500 nS if not worse, way beyond spec. I > >checked > >what happened with a series 10 Ohm resistor, nothing worth noting. OK, > >at > >the driver output the initial perhaps 1/5th or 1/4th of the voltage > >excursion > >got better. > > >Before I began trying things out etc. I replaced the TC4426 with a > >good old > >MC33151. Worked same as it has worked last 20+ years for me, well > >under > >100 nS (was something about 50-60). > > >I did not pursue it further, there was no point stopping at it. Got it > >working > >with the 33151, which was designed in times when apparently someone > >would have noticed if a design did not work :-). > > >But a few days later I am still curious what was that. Looked as if > >some > >foldback current limiting - wanted by the designer or not - took > >effect, perhaps > >I could have eliminated it with a larger current limiting resistor so > >things would > >get usable (much worse than with the 33151, obviously, it would have > >taken > >perhaps 50+ Ohm). > > >Any thoughts? > > >Dimiter > > As was related in an earlier thread, the input hysterisis of the 4426 > and similar parts is not enforced on the input signal pin, but after a > linear input stage comprised of an nmos fet. This means that input > hysterisis cannot be specified for these parts as a percentage of the > input signal logic level - so input noise and ground bounce on the > devices' supply pins may be an issue. Well the input is 5V CMOS, well below 10 nS edge. If that is not enough for them I can't imagine what is. Ground bounce if any will be solely inside the chip, like I explained in an earlier post a 1206 1uF (ceramic) is pinned to the power/GND pins, at the bottom side of the board. The connecting vias also connect the thing to the power/GND planes respectively. So it is neither of these, I would not have posted this if things were that easy :-). > In a dual driver situation, you might even find that one driver is > upsetting the other, so check the the input and output state of the > static side when the funny behavior is occuring on the other. There > may be symptoms there, that you wouldn't normally look for, or expect. The other driver is driven and unused. Does not change state at all, the input is either 0 or 5V (don't remember which it was). Nothing worth noting at it. Dimiter ------------------------------------------------------ Dimiter Popoff Transgalactic Instruments http://www.tgi-sci.com ------------------------------------------------------ http://www.flickr.com/photos/didi_tgi/sets/72157600228621276/
From: Didi on 11 May 2010 03:48 On May 11, 1:45 am, "Tim Williams" <tmoran...(a)charter.net> wrote: > .... > > Powered iron is horrible for switchers; AC losses get huge as flux > > density increases. MPP is great (more energy storage per volume, much > > less loss) but expensive. KoolMu is almost as good as MPP and a lot > > cheaper. Somebody else, maybe Micrometals, has an equivalent to > > KoolMu. > > Avoid yellow/white mix 26 if possible. It's cheap, so you can give up flux > density for dollars, if space and weight allows. I see green/blue mix 52's > on motherboards a lot, which apparently has about half the losses of 26, for > slightly more price. I haven't seen any other grades used in consumer > equipment, just those and ferrite. > > I think Micrometals' products are all numbered. I'd guess 52 is the > equivalent you speak of, but I don't have a reference for that. I use their 52 material. The people at Micrometals were as kind as they could get, sent me some samples etc. I am not sure I can do that on a ferrite core, at 500 kHz I manage somewhat above 10 (I think I tested at 12) W through a 10mm outer diameter, 3.8mm high toroid. It does get hot at this power though. Can't measure it, too tiny for the IR thermometer, but it is above 100C. Not like John's smoking paint, though :-). I guess it will be there if I push it for another 5W or so. I have some margin as things will not get to 10W, I expect them to stay within 5 to 7 where it works OK; I am more after saving some battery power if things get to that. Can you make a guess how much there is to gain by using MPP (presumably the lowest possible loss)? (right now my efficiency is around 60% to 70%, I would guesstimate 65-70% as far as the inductor is concerned). > At high frequencies, gapped ferrite looks best. I have been using ferrite cores for that sort of thing last 20 years, but at 100 kHz. I could not locate anything which could do the job at 500 like that 52 material toroid does. I am not sure why I did not find anything and opted for the powdered metal, was some years back. Then I may simply have made a mistake. ------------------------------------------------------ Dimiter Popoff Transgalactic Instruments http://www.tgi-sci.com ------------------------------------------------------ http://www.flickr.com/photos/didi_tgi/sets/72157600228621276/
From: Nico Coesel on 11 May 2010 03:57 Jan Panteltje <pNaonStpealmtje(a)yahoo.com> wrote: >On a sunny day (Mon, 10 May 2010 09:53:57 -0700 (PDT)) it happened Didi ><dp(a)tgi-sci.com> wrote in ><33785897-ee61-4e81-ae30-cf767ff3b81b(a)g21g2000yqk.googlegroups.com>: > >>Not so long ago I put them in a design (the other Microchip part which >>has >>made it into my designs is an I2C EEPROM). >>Worked as expected while driving tiny MOSFETs (two in an SO-8). >> >>But when it came to an IRF540 - plain old IRF540 - they failed >>miserably. >>I first looked at them driving the IRF "empty", i.e. the drain >>hanging, source >>practically grounded (well, via 0.1 Ohm or so). Never got past this >>with this >>driver. >> >>The edges got really bad - 500 nS if not worse, way beyond spec. I >>checked >>what happened with a series 10 Ohm resistor, nothing worth noting. OK, >>at >>the driver output the initial perhaps 1/5th or 1/4th of the voltage >>excursion >>got better. >> >>Before I began trying things out etc. I replaced the TC4426 with a >>good old >>MC33151. Worked same as it has worked last 20+ years for me, well >>under >>100 nS (was something about 50-60). >> >>I did not pursue it further, there was no point stopping at it. Got it >>working >>with the 33151, which was designed in times when apparently someone >>would have noticed if a design did not work :-). >> >>But a few days later I am still curious what was that. Looked as if >>some >>foldback current limiting - wanted by the designer or not - took >>effect, perhaps >>I could have eliminated it with a larger current limiting resistor so >>things would >>get usable (much worse than with the 33151, obviously, it would have >>taken >>perhaps 50+ Ohm). >> >>Any thoughts? > >Yep, stay clear of Microfip. This sounds like the Pope converted to the Islam. Everything allright? -- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... nico(a)nctdevpuntnl (punt=.) --------------------------------------------------------------
From: legg on 11 May 2010 07:39
On Tue, 11 May 2010 00:29:25 -0700 (PDT), Didi <dp(a)tgi-sci.com> wrote: <snip> >> In a dual driver situation, you might even find that one driver is >> upsetting the other, so check the the input and output state of the >> static side when the funny behavior is occuring on the other. There >> may be symptoms there, that you wouldn't normally look for, or expect. > >The other driver is driven and unused. Does not change state at >all, the input is either 0 or 5V (don't remember which it was). >Nothing worth noting at it. > As long as you did look...I've only seen trouble on an MIC4424. I realize that a range of these drivers has the same nominal inputconfiguration, but this similarity may encourage a lack of dilligence in characterization. My reaction, like yours, was simply to change the chip, after noting the pin waveforms accompanying the behavior.. RL |