Mosfets in parallel?

This sounds odd to me.
There are plenty of Mosfets which will deal with vast currents.
5A is pretty low, so is it the voltage that's the problem? If so, putting several in parallel won't help.
i know that there is a lot of posts on this subject, but still I would like to hear some opinions.
My opinion concurs with the previous threads.
Del

You must have a very limited source for Mosfets where you are. A quick search through any online electronics supply company will give you thousands of different sizes and capacities to choose from.

Or consider using an IGBT device instead. They are voltage controlled like Mosfets but have much higher power handling capacities and are ideally suited to carrying higher currents and higher voltages without problems. There are many that have specs that allow them to be used as direct drop in replacement devices for many power Mosfets while having in many instances greater voltage, current, and power handling capacities as well.

I have stacked IC CMOS DIPs in parallel to lower the on resistance. It works best if you match the FETs as close as possible. Funny a push-pull output vacuum tube configuration of (4) tubes does basically the same thing. The tubes are matched at the factory and both sets of the (4) tubes conduct 180 degrees out of phase. Yes I believe it will work if you hand match your N MOSFETs. But of coarse I am a Maverick and try different things that people think are strange.

Thanks for all replies. Here are my comments:

@Del the cat

5A is low, but that's on 150V between drain and source and it is 750W on a single MOSFET. There are plenty of MOSFETs that can work on 1kV, but their SOA is heavily derated above 50 or 100V.

@tcmtech

I'm in the US and I went through mouser, digikey and newark catalogs, but as I said all MOSFETs above 100V are heavily current limited, even they can 'carry' 50A at lower voltages. I directly contacted IXYS (most good MOSFETs mouser and digikey don't have in stock) and they found me one that can have 5A@150V, but that still means that I had to use at least 2 of them since their ratings are on 25C and each one was 60$. In contrast, some other MOSFETs that can have 1.5A@150V are ~3$ and I as wondering how I can used them. I tried connecting these cheap MOSFETs in parallel without any source resistors, but didn't work. One that first starts to conduct 'blows' instantly before other start to conduct and I blown ~20 of them so far.

@maveric_manic

I was thinking about hand matching, but is this the only way to go?

I'm looking for a way to allow them to have Vgs voltages slightly different. Putting a small resistor in their drains may work, but if I use 5 MOSFETs that means that each one on average carries 1A, and their individual Rds is 150V/1A=150 Ohm. So, how small resistor in drain is actually small to offset mismatching?

Hi,

usually you do not need to match if you select by rds-on.

Consider the maximum current that one of n has to carry if symmetrising with resistors.

I am not at all sure that exceeding the DC-specification that blew your first approach. Your timing should be very accurate and symmetric, so that none of the transistors start to conduct (or stop) earlier than the others. So inductance and capacitance of layout shall be near perfect: no fast spikes please, do you see 1µs spikes?

I had similar problems in 35 years ago with the then "modern" 2N3055, paralleling 3x24 pieces in a variable frequency and amplitude 3-phase motor-drive.

For those MOSFETs in series you must be sure that switching is synchronous, how do you control the necessary voltages? Do you select also for U_GS-on? And gain?

I estimate that it is much easier to parallel many high-voltage-low-current transistors than try to operate with serial connection!

RHABE

Have you considered other regulator topologies?

Is there some reason you need a "linear" regulator design with 750W dissipation in the pass device?

One of the switching regulator topologies could be better choice and many common MOSFETS would run warm switching under your stated conditions.

If you feel you must do it this way there is no reason not to put as many in parallel as you need to achieve the required dissipation. You may want to use fans or even water cooling but MOSFET's current share well just give yourself a healthy margin and stay within the ratings of each device. Please post a picture when you are done. I'm sure we would all get a kick out of this monstrosity.

Even matched devices when connected in parallel need resistor in series in each emmiter/source.

This is a general practice I have noticed in most of cases and is supported theoratically as well.

<< emmiter/source. >> Please read emmiter/Drain

If you switch "on",

then resistivity is giving the current distribution between the FETs.

So if you have a variation of 10% in R_DSon you will have 10% variation in the currents!

This is possible and I did it successful, so if you insist on using resistors this may originate from much worse matching requirenents.

RHABE

doesn't apply to mosfets

if you use 7 1A devices to distribute 5 A between them

each devices share of the current is 5/7 Amperes, which puts each device at approx 70 % of its rated load,

if the typical on resistance of the mosfets is X ohm, then your drain should be at least 30% of X ohm if the mosfets are closely matched, mosfets typically have < 0.1 ohm on resistance, by using a larger resistor than ~x/3, say 1 ohm resistor you will achieve accordingly a larger tolerance for mismatched devices

resistance in each mosfets drive loop, you will achieve automatic foldback per device

UJT transistors do not work well in parallel, however, MOSFETS love it.They tend to share the load, not to hog it, as regular transistors do.When they heat up, they shed load,so parallel as many as you need for current demand.Very much like a vacuum tube in operation.

However, there are many high current mosfets out there readily available.Google a little and you will see.

Hope this helps.

EZStreet

That is correct they have a positive temperature coefficient and will tend to load share. But it is important to match them for best operation, also circuit layout and pc board copper tracks should be the same for each device especially if they are to be operated at high frequencies. This ensures they all switch at the same time. If they are working with a steady dc load layout is not to critical.

However there are lots of MOSFETs that will handle the load required by the poster so I would not go to the trouble of paralleling them.

Hi Guest. If your circuit was digital (i.e. digital control) this should not be a problem. You could drive the mosfets with the maximum (but safe) Vgs (in your case: 5V) and everything would be fine.

But your circuit is an "analog control" one. You'll not only have a problem on "start up" but, also, in the continuous operation of your circuit. As an example: If your P.S. output voltage is larger (than it should be) then the control circuit (via feedback) will drive the mosfets with a lower voltage, in order to make them "less conductive" hence reducing the P.S. output voltage (and the opposite happens if the P.S. output voltage is smaller than it should be). In this case, the control voltage could go down enough making some mosfets (those with higher Vgs) to be "less conductive" while other mosfets (those with lower Vgs) would still be "more conductive" (carrying the majority of the output current). Hence, this is a general problem when you use different mosfets with different characteristics.

(Notice, though, that when a mosfet (with a low Vgs) is -e.g.- in "full conduction" the other mosfets (with higher Vgs) are not in "full cutting off" but in a "half conduction" state instead. This means that the output current doesn't pass through just one mosfet. A smaller part of the current passes through the other mosfets too.)

So, it is recommended to use the same type of mosfets in your circuit. So, the Vgs of all mosfets will be -more or less- the same. Furthermore, you could use one or two more mosfets in order to have a redundancy in the case where there are fairly big differences in the Vgs of the mosfets.

You are right about putting the N mosfet's in series; it is a problem to match them. It is much easier to match them if they are in parallel. The differences in gate to source threshold voltages can be easily balanced with some ballast resisters.

I assume that a series inductor and switching is too noisy for your application; but it would be more economical if noise filtering is possible.

I hope this helps.

Luther M

The world is going green and you're designing a cirquit with 750 W losses! Shame on you.

You cannot state this as you do not know what is driving the request for a linear amplifier.

Did you ever design a good linear and a good switching amplifier?

Try hard and then comment again!

RHABE

Actually have both but lt's go deeper into the problem if we have to. Matcing Mosfets for a linear application is a lot more trickier than you may think. Actually for switching mosfets this is not neaded because Rdson has a positive temp coefficient, gate voltage is driven well over Miller platau and Rdson no longer dominates the actual current, external impedance does. The higher Rdson Mosfet gets a few degrees hotter for the next cycle which then increases it's Rdson and currents balance. The real problem for a linear application is mosfet's Gate threshold voltage variation which can be like +/- 50 % for the same type Mosfet and what's worse it has negative temp coefficient i.e. it drops with higher temp. So the first mosfet that conducts takes all the load and getting hotter with the same Vgs it actually increases the current that makes a disaster at high powers. This looks strange at first unless you understand that Rdson is the minimum resistance the mosfet can have nothing to do with actual momentary Rds and Rds to Vgs relation that is far from linear, and has significance only for switching applications.

--64 6E 69 4D 65 6C 70 6D 69 53--

Oops! another typo sorry! Change the "...higher Rdson Mosfet gets a few... with: ...lower Rdson Mosfet gets a few... and any clarifications needed please ask.
--64 6E 69 4D 65 6C 70 6D 69 53--

As I stated before you may need to step up to IGBT devices. New old stock devices with multi KW dissipation capacities having 1200 volt working voltages and current ratings into the high hundreds of amps are not hard to find or expensive when compared to their new stock costs.

I buy International Rectifier and Infineon IGBTS with 1200 volt, 600 amp, 4500 watt dissipation ratings all the time for around $35. Their smaller and faster cousins are even cheaper!

A 600volt, 100 amp, 1 KW dissipation IGBT is around $15 or less if you know where to look.

I guarantee you wont burn one of them out with a 150 volt 5 amp drop across one of them!

Sure the packages are much bigger but a heatsink that can take 750 watts is not that small either!

Hi,

are IGBTs good for linear operation or "only" switches?

With FETs this amplifier is a true linear one, very low noise and very low distortion if done well!

There are still today linear (pure harmonic amplifiers) to drive servo-motors (not only music). This is to minimise unwanted interference with nearby measurement and to insure smooth operation of (nearly) torque of wanted frequency only!

RHABE

I would assume they must have a fair tolerance to linear operation being that many can have the same voltage, current, and gate control specs of many Mosfets yet still be listed with considerably higher power dissipation numbers.

I use them fairly often in switching applications where they are deliberately driven in soft turn on and soft turn off modes that do produce a much higher level of energy loss during those transitions. They have shown themselves to be able to easily handle it provided a reasonable overhead is built into the design. But no they are not in a high fidelity sound reproduction application as I use them.

However I have read that IGBT's are in fact becoming more common place in audio amplifier power output stage designs now due to their inherent robustness of design and unique characteristics.

Thank you, good to know.

Soft turn on? How long is the transistion time?

RHABE

Depends on the application. In an analog based Grid Tie Inverter systems I have developed for personal use I am using a fairly long transitions time of about 1/600 second on each side. Thats rather slow for the actual devices that are capable of around 100+ times faster switching times or better.

It wasn't my first choice for switching speeds but it cleaned up the odd line noise and harmonics problems that high switching speeds would keep producing at different power levels. Simply slowing down the transition times from on to off and off to on gave me a low enough line noise levels that basic off the shelf line filters could easily handle them from there which allowed for line noise levels that are below whats on my local utility already.

I also have used IGBT's in a solid state multi KW load bank. In that mode they are working as the power handling variable resistors being controlled by a simple feed back loop. I just set what amps load I need and let them do the rest.

So yes they can in fact work well in linear power applications if properly used of course!

Great work!

RHABE

Hi, You are correct. Use some current balancing technique - series resistors with positive temperature coefficient - iron or steel wire is good. Look up http://www.allaboutcircuits.com/vol_1/chpt_12/6.html

Here is just a thought.

Why cant you switch all the FETS on while driving a small load (low current) until the FETS are stable and then switch in the bigger load.

I had a problem once with a COTS power supply that provided +/- 12VDC had a HUGE time delay between the + and - supplies.

I simply built a circuit that waited until both supplies were stable and then switched both of them into the load with a relay.

Just a thought...Maybe it will help.

bill12780