Old Mosfet question-new problem

Dippy

Moderator
Glad you've got it working... now you know why they make drivers :)

And good points on heatsinking. And some good caveats on Real World versus Data Sheet.

If you have genuinely found that a 'good' device cannot operate at specification you should report it to the manufacturer. Are you confident enough to say to the manufs "You've got it wrong."?

But you have to be careful.
Are you driving the gate at the voltage defined in the manuf's test ciruit? (10V in their spec sometimes pulsed @XHz @Y% Duty).
Are you using the correct capacitors on the driver?
Is your PCB layout and track geometry good?
Is your load reactive?
When you say 'continuous' do you mean DC or a steady continuous PWM?


There are numerous things that need to be mastered to get the best out of a device and the manufs have this knowledge and probably do the measurements when the wind is blowing in the right direction :)
70A down to 21A is a long way down... I'm genuinely surprised that, under the correct conditions, that it is that much different.

And pulse swich speed IS important. And that is affected by your gate drive quality.

Anyway, bottom line. Your project works and doesn't explode. Good stuff.

EDIT: Oh, sorry, some of mine has crossed with BBs.
 

manie

Senior Member
Yes BB, I'm not refering to the speed of switching on/off (x nSecs), I'm refering to the chip's running freq. Tried it today at 8MHz,4Mhz and 500KHz frequency. This as you said affects the PWM freq. At 4MHz the lowest PWM freq is 3900Hz or there abouts. It gets down a lot lower at chip 500KHz. At all three chip freq's and hence at higher/lower pwm freq's, there was no real change in Amps delivered and no real change (HAPPY) in mosfet temp. at the SAME Amps passed in each case. Also, at lower duty% and thus lower Amps passed, the temp. dropped. At 94% duty (where max. Amps setting occurs) the Fet temp. was highest in all cases, and in fact in all cases the temp was "x" at Amps "y". A perfect correlation. That is what point 4 tries to say.
Once again thanks
Manie
 

manie

Senior Member
Ah Dippy, saw your post when I replied to BB. Agree on all the points you say and NO, I'm not confident enough, just that if I take this new approach, as you say, no more explosions. "Continuous" means a long steady pwm stream of pulses at varying duties according to external stimulants that will decrease/increase the duty% as required. In short, I could improve the project after all this posting back and forth, and that is what matters.
Manie
 

Dippy

Moderator
You should go careful with those "external stimulants" ;)

Anyway, to quote some cliches:
Job done.
Lessons learnt.

So, the next project will be controlling Space Shuttle rocket nozzles with a MOSFET and string...
 

jglenn

Senior Member
A 70A FET will carry 70A all day long, if you have a very large heat sink, or small one with a fan. Practical products "derate" the specs, by 30% perhaps, so you have a safety margin. Switching loss will add heating, and it goes up with chopping freq.

You should look at putting in a current limit circuit if you haven't. This is essential to preventing destruction of the fet. It has to work fast. Normally it is a shunt in the fet source ground, and a differential amp, then comparator, fed back to the chip making the pwm.
 

manie

Senior Member
Jglenn, I run feeder wires through a 200A closed loop Hall sensor with 3.125mV/Amp output (after subtracting the reference voltage = 1/2 supply V = +- 2.5V). With wires looped through twice, like I have it, you get 6.25mV/Amp output to chip ADC. Measured against good quality 400A DC clamp meter it differs only in the decimal eg. 50.4 meter = 50 calculated via Hall an 50.8 meter = 51 Hall etc. This is used to control PWM duty%. A bit pricy but worth the steady accurate ADC readings. Even so, a few days back I would blow/explode/sublimate TO220 package fets at low duties. It's better now. I think I can safely recommend the IRFP064N in TO247 livery as I saw this morning that it can handle a bit of abuse and survive ! (60-65A for about 30sec to 1 minute). I did not have a chance to measure surface temp's, I was busy switching off/disconnecting etc !

What would you think is the effect/efficiency of a 92mm 12V fan blowing directly through the fins of the heatsink ? Is it safe to say +- 20cfm is pushed through by this fan ? (No data found on it yet). It produces GOOD air movement and you can sense the removed heat on your skin at the air outlet.

Looking back, like BB/Dippy/ yourself and everyone said, thats why driver chips are made. My first try with chips had something wrong and I thought it might be the driver chips fault, hence the embarkation on a futile search for a better (worse) driver ! Lots have been learnt and things are really looking better now !

Thanks all !
Manie
 

BeanieBots

Moderator
Always nice to hear of a success story. Even if the path was painful.

Please note though, it was NOT the high currents that were killing your FETs.
If they say they can handle it, they can.

It was excess power dissipation caused by driving them in linear mode.
Assuming a resistive load, when half on, they will have about half current at about half the voltage. Now work that out in watts and see what the datasheet says about it. They don't like being like that for very long!
 

BeanieBots

Moderator
I'm sure someone will trawl through the thread, get the numbers, download the datasheet and do the sums, but I sure ain't going to;)
Needless to say, it's enough to go "puff".
(and I'm quite sure the datasheet will confirm that)

Oh yes, nearly forgot, that's where the PWM frequency bit comes in.
How often in a given period they are in that state.
 
Last edited:

Dippy

Moderator
"Oh yes, nearly forgot, that's where the PWM frequency bit comes in.
How often in a given period they are in that state."

- Absolutely, I've been saying this for years, well months.
People seem to forget the proportion that it spends in transition compared with the absolute on/off times.

Never mind BB, the same question will pop up again in a few days time and you'll have to go through the whole thing again...:(
 

BeanieBots

Moderator
If it was a sticky then it would be the "next" problem which gets repeated.
Besides, who reads the stickies anyway?
Most questions have been thrashed out before, all that is required is a forum search. Hardly anyone bothers with that either.

Did YOU read the FAQ before posting?
(not that there's anything on FETs in there, but there might have been)
 
Last edited:

manie

Senior Member
BB, I actually own a PRINTOUT of both FAQ and Technical FAQ. How else did I survive my first 5 months with 28x1's already before venturing a question on the forum ? Although I like your quick response times, maybe point people like you did me towards reading spec sheets more. Some people can understand data sheets, some not, sometimes one just needs a pointer or PLAIN ENGLISH explanation to make that quantum leap...
Manie
 

BeanieBots

Moderator
You're quite right.
It's very easy to point people at the datasheet, but there's little point if they are not going to understand what they are reading or don't know which bit of data to look up or, more to the point, WHY.

That's also what makes it hard to have a "general" answer. Each time the same question is asked it's usually based on a different understanding (or lack of) and requires tailoring to suite.

Besides, it keeps me out mischief;) .... maybe.
 

manie

Senior Member
Besides, it keeps me out mischief .... maybe

The amount you post you must be seriously mischievous !!!!!
 

Dippy

Moderator
Well he does run around clean rooms in the nude.

I agree with the sentiment of what's been said.
And I'll say again; a nice set of sheets on basic electronics, applications and simple examples on some of the popular components like MOSFETs would be really handy.
Even some hints on what to look for in Data Sheets would be very revolutionary educational.

Sometimes a search can turn up all sorts of stuff which can be annoying. For example if this thread was 30 pages long and discussed flower arranging and I typed "MOSFET" then the search would pick this up and the poor sop would have to trawl through 29.x pages of nonsense before realising they'd wasted their time.

A permanent link to a PICAXE Manual called BASIC Electronics (which hippy could write ;)) would be one solution but will never happen.

I'm retiring from MOSFETs to concentrate on selling ice-cream.
 

manie

Senior Member
In my case it was like this:
Last I soldered anything to anything it was a 50g 28gauge wire wound "miniature" resistor to the 6V heater of an EF47 Mullard valve ! Then the incredible happened. Someone discovered Germanium and the OC28 "transis whatsisname ?" was there. Thats when I opted out, it won't last I said, Mullard valves are too good to be replaced.... well this is now, the old lust for masochism was reborn and you know what ? No one in any of the electronic shops has even heard of Mullard.. Mullard ?? Who's he ? I have been active in computer software, specifically VB/VBA and MS Access etc. So when I came accross this "microprocessor" site called "The Revolution Has Happened Education" near where some old guy who wrote plays and sonets used to live, I thought, HECK I CAN DO THAT ! That was 5 months ago. The "easier" part was getting the 28x1 and board, the bad part was 100+ Mosfets up in smoke. Hell, I'd forgotten most of Kirchoff's, they package seeds for retail don't they ? Then the real revelation ! No more BULKY BOOKS with outdated component specs, you could get them "on demand" off the web !! WOW !
Then I realised, I will have to learn, and QUICKLY, how to read and interpret datasheets again. I must have passes every senior school certificate, every Poly Tech Diploma and every Bachelors there is in the past few weeks. The learning was STEEP ! but worth it, and you know what ?

IT KEEPS ME OUT OF S.... STREET TOO !
Manie
 

BeanieBots

Moderator
Nice to know I'm not the only one who knows his pentodes from his triodes.
Oh happy days! At least they gave a warning before exploding.
When they started to glow purple, back it off quick.
 

manie

Senior Member
Nice to know I'm not the only one who knows his pentodes from his triodes.
Oh happy days! At least they gave a warning before exploding.
When they started to glow purple, back it off quick.
Purple just before burning BALCK ! From the inside... Imagine telling someone now that in those days installing a Diode meant that first you had to build a multi voltage power supply, 6V for heaters, 200V etc for etc etc.. Then the old round 5 or 7 pin socket, then some suitably heavy current capable wiring, then you had an operational diode. Oh how nice 1n4148's.... and dont forget the 1/8th inch galvanised chassis...........
Manie
 

BeanieBots

Moderator
Indeed, look at the problems they have with a 5v supply. And they have regulators to help as well. Would love to see them knock up an HV + Heater supply:eek: (and stay alive with a live chassis version)
 

manie

Senior Member
BB wrote "(and stay alive with a live chassis version)" Yikes ! I forgot about those bites ! And they were NOT 8-bit either.........
Manie
 

jglenn

Senior Member
To do thermal design right, you need a few temperature coefficients. One for the fet junction to the case, then case to the heat sink, then heat sink to air. Fans really help. If you don't have all the specs, just try some suitable parts, put a fast temp sensor on the source or drain lead of the fet (isolated hopefully), and try it.

Your current measurement circuit sounds ok, but can it shut off the fet rapidly? Have you done any overload tests? Does it go thru the PICAXE? Might take a mS, which is a LONG time in the world of fets blowing up. :eek:
 

manie

Senior Member
Jglenn, think I've "overloaded" enough fets thanks... I've just decided to go the "overkill" route, enough fets distributing the load seems to work fine now, haven't blown one up for a whole week now ! And passing good current with really fine (1% duty) control. I have a h/sink of 200x120x75mm with 6 fins top and bottom and a 92mm fan blowing directly down the fins. Seems to do the work..

EvanH, thats some neat wiring there. I've also done the heatsink thing with some fairly heavy wire on the legs, it seemed to have helped.

Manie
 

jglenn

Senior Member
AOK. In the future you may want to try a current limit circuit, they are very useful. Sometimes the load changes, or even shorts, and that does not need to destroy the fets. In switching power supplies, you must have it. Things happen with transformers, like saturation, where the inductance goes away, and then you are driving a short circuit! Beware rapid load changes. Better safe than sorry. An ounce of prevention is worth a pound of cure. :rolleyes:

All you need is one 10A fet for a test circuit. When the feedback amplifier is working correctly, you should be able to short the load, and the current will be sensed and drive reduced or shut off, depending on how you set it up.
 

evanh

Senior Member
On that subject, the customer literally dropped a spanner on the slip-rings between those units I built and the heater elements they were powering. No problem at all. Only the fuse blew. With the idealistic isolation and parasitic power supply there was no where on that board that the fault could travel other than through the on-state mosfet.

As a result, I've concluded that many of the typical component blow-ups in fault conditions are likely due to earth-loop type current paths dumping high currents through low current paths.

EDIT: Although, it would be fair to say the actual peak fault current was prolly somewhat limited by the 10 volt transformer and subsequent wiring runs to the elements. So, my example will prolly not work for more extreme faults.
 
Last edited:

manie

Senior Member
I am fiiting some 25A fuses between the Mosfet Source and supply (-). I have however, when current exceeded 25A blown that particulat fuse...... and that particular Mosfet too ! Was this "transient" damage maybe ? If so, will fitting of "Transorbs"(18V) accross the Source/Drain help ? It never rains..............
Manie
 

BeanieBots

Moderator
Hard to say without seeing 'scope traces.
The very act of a fuse popping can kill MOSFETs due to spikes caused by the sudden drop in current on even a very small inductance. Transorbs might help. A proper investigation plus a designed snubber is the only way to make it bullet proof.
 

BeanieBots

Moderator
A 'snubber' is (more often than not) just a simple RC often connected via a diode to the output of a switching transistor. Its purpose is to absorb any spikes and dampen any ringing to prevent the transistor from over-voltage damage. The values of the R & C and diode type are designed to reduce power losses in the 'snubber' yet still maintain a voltage safety margin for the semiconductor(s) it is protecting.

Here's a document which can expalin it better than I can.
http://www.powersystemsdesign.com/design_tips_nov07.pdf
 

jglenn

Senior Member
I would not have suspected it, but there are videos on snubbers!

http://www.snubberdesign.com/

http://www.smpstech.com/mtblog/snubber_design.html

By way of background, I was influenced in the early sixties by one of my peer designers, who always used a faster transistor than he needed and then looked at the voltage and current waveforms at turn-on and turn-off, shaping them as he wanted with external components to make the switching characteristics dependent more on the external parts than the transistor. None of us had ever heard the word snubber -- it was just wave shaping and EMI control, and we used as many lossy ferrite beads as we did lossy capacitors for this purpose (lossy to damp high-frequency ringing) and made our cores lossy above the switching frequency. (Note that the use of lossy components, including magnetic cores above the switching frequency to suppress ringing and EMI is a very useful design technique.) What I mainly learned from this was never to switch faster than you need to, trading some power to avoid exciting high frequency resonances and to use lossy components to damp these resonances. The Electro-Magnetic Interference (EMI) that the resonances cause is easier to kill at the circuit than with an added EMI filter. I also learned that you should always look at voltage and current waveforms -- many designers fail to look at current waveforms, losing half the insights into their circuits performance.

I tend to use transient suppressor diodes and then augment them with snubbers only if needed. The first environments I designed for included Electro-Magnetic Pulses (EMP) and lightning, and I always placed a transient suppression diode across each input capacitor and across the power transistor for reliability purposes. The spike you are trying to snub is caused by leakage inductance and I try to minimize the leakage inductance causing the spike. For adequate design margin, most snubbers dissipate more energy than stored in the leakage inductance. The transient protection diode has an advantage in that it absorbs just the amount of energy needed to protect the transistor, never more. I use a transient suppression diode as much for later failure investigation reasons as engineering reasons. When a failure report comes back on a transistor or capacitor with the usual electrical voltage overstress reason, you can show the circuit schematic with the transient suppression diode and force a more competent failure analysis of the part. You need to use a fast-acting transient suppression diode, not your normal zener diode for this application, whose clamping action is too slow. This approach may be too expensive for some commercial designs, but in my design career, reliability was always the first consideration, not cost. Reliability has a major impact on life-time costs, and if these are considered, the cost impact of using transient-suppression-diodes changes for the better.

Finally, there is the question of lossless snubbers. One way of looking at them is that the lossless snubber circuits can be viewed as miniature switching-mode powers supplies and much of the some thinking applies.

Notice I did not answer your questions, but used it as a sounding board for some ideas. You will also want to look at my other website, Snubber Design.
 

manie

Senior Member
Jglenn, I'll take some time to assimilate all that ! Good info thanks. Wish I spent more life in electronics AS A PRO and not a HOBBYIST ! Now I'm trying to be a SERIOUS hobbyist (LOL LOL) What the heck, live/try/learn.....
Manie
 

Dippy

Moderator
Blimey JGlenn, that's a big one.
Drawings would be lovely, thats a lot of text to read on this Forum ;)
 

BeanieBots

Moderator
Some good info jglenn.
I think you hit the nail on the head with regard to robust vs cost.
The FAST zener is my prefered choice because it's easier and quicker to implement. Not always less lossy than RCD but a LOT more expensive.

Always better to do whatever you can layout and component wise to reduce the need for snubbing in the first place.
 

manie

Senior Member
Something interesting and POSITIVE for a change.

I bought some ST-W18nb40-18A mosfets under "specials" from Electronic Goldmine a while back. They are TO-247's, rated D-S 400V, Gate 30V, 18A @ 25 deg C, 11 Amp @ 100 deg C. Used the exact same TC4420 driver chips, PWM @ 4KHz, varying Duties, exact same heatsink and fan cooling. Now the good bit....

Running at 9A each (+- 50% of rated) they ONLY reached 57 deg C !!! Was running at varying duties for 4 1/2 hours non stop. Ambient temp is 30 deg C. Fet temps varied between 41 and 57 deg C, low duty = low temp = lower amps. These mosfets are from ST and cost US $1.00 each. The mosfets I get locally (IRFP064N's @ 70A rated) costs equivalent US $4.00 each. They last not more than minutes ! Now the question....

Could they be counterfeit shady types ?

I mean, NO success at 20A (30% of rated capacity at 70A), then change to a much lower rated Fet of 18A and suddenly success running at 50% of rated, with everything else staying exactly the same. Seems to me I might be getting shady components here ?

It is cheaper even if I have to tripple up for higher amps.
Manie
 

BeanieBots

Moderator
I'm sure I mentioned earlier about not using higher rated MOSFETs??
A higher current rating does not always mean it won't get as hot as a lower rated one. The reasons are all burried deep in the datasheets.

Compare the gate voltage / current curves and the gate capacitance of each device and it will probably become more obvious.

Anyway, really glad you've found one that works in YOUR application.
 

Andrew Cowan

Senior Member
When I was researching FETs for a recent project, I found that the smaller FETs (in terms of current) have less capacitance. This may not be the case for all FETs, though.

A
 
Top