Wha'ts the world's biggest transistor?

[ejeffrey]

Turns out that's not the case, and transistors get quite large.

The really large devices are just banks of transistors in parallel and/or series for more current and voltage handling.  I don't know what the largest single die transtor is but once you can stack them you can basically make any size you want and the cost is mostly linear.

[T3sl4co1l]

IGBT dies are typically a couple hundred amperes; if scaling keeps up for the higher rated ones, I would guess 6kV ones are rated... maybe pretty low like 50 or 100A, and also as big as dies can get (20, maaaaybe 30mm?), and then stacked in parallel in modules.

Oh here we go, actual dice: https://www.5scomponents.com/pdf/Chip-Flyer_0409_Web.pdf so you can see what ratings and dimensions are like.  Smaller than I thought.

Mind, there may be bigger ones out there; this should be read as a typical example.  For further reading you could well contact Infineon, ABB or others and see if they'll entertain your curiosity.

90%+ of the finished unit is packaging, necessary to array chips together, dissipate heat, and insulate between them at such voltages where a spark can jump several mm through air already.  Sometimes the best insulation is simply distance.

---

There are also SiC products available, and coming out, in various ratings; I don't think they're quite advanced enough with it to be taking over this kind of power range -- it's devilishly difficult to make low-defect SiC wafers, limiting scalability per cost -- but I have seen papers on 10-12kV MOSFETs and IGBTs in SiC.  They are not generally available, I think probably because of ITAR? (They're much faster than Si parts, so you can make a wicked EMP generator with them, or perhaps initiate explosives, etc..)

Tim

[Fried Chicken]

Start again. What are you trying to do?

Find out what the world's biggest transistor is.  If it even is a transistor in the traditional sense.

And what the heck does that have to do with a lawn mower motor?

Let's say I wanted to control a lawnmower motor with a transistor...  What's the world's biggest transistor?  I know power amplifiers have some big transistors, but what if we get into the stupid range?

I guess I should have explained... I was looking at someone building a lawnmower, and I wasn't sure transistors even got large enough to conceivably do that, so I thought a lawnmower was a ridiculous application for a transistor, like you would use something else to power it.  Turns out that's not the case, and transistors get quite large.

You were quite mistaken. There are transistors capable of switching far greater loads than a lawn mower motor, which is a fairly small in the scheme of things.

You received some irrelevant replies because you didn't say exactly what you're doing. What's the world's biggest transistor? is a meaningless question. Are you talking about physical size, current rating, voltage rating, power dissipation?

I had a quick look through Digi Key and found some transistors which are more than capable of switching a lawn mower motor. As you made no mention of voltage, or whether it's DC or AC, I assumed smoothened DC, from a bridge rectifier powering a universal motor, drawing around 1kW or so.

https://www.digikey.co.uk/en/products/detail/stmicroelectronics/STP45N40DM2AG/5724369
https://www.st.com/content/ccc/resource/technical/document/datasheet/ff/da/51/30/5f/70/4e/77/DM00213649.pdf/files/DM00213649.pdf/jcr:content/translations/en.DM00213649.pdf

https://www.digikey.co.uk/en/products/detail/infineon-technologies/IRFR4620TRLPBF/2021282
https://www.infineon.com/dgdl/irfr4620pbf.pdf?fileId=5546d462533600a4015356323ae920f7

They're not that expensive considering. Anther option is to use thyristors for part of the bridge rectifier, which would probably be cheaper but not so flexible.

Those look tiny for the power they are rated for.

I had some idea that production errors or something made larger transistors impossible or very difficult/expensive.  I guess not.  Just add silicon.

[Wolfram]

If you disqualify multi-die parts like the most common IGBT modules, and SCRs which can't be gated off, the most powerful single-die transistors I know of are IGCTs like this one: https://library.e.abb.com/public/0545534831d6421cb2d188bb713661ae/5SHY%2065L4521_5SYA1261-01%20Sep%2020.pdf

One interesting aspect of these is that the gate drive current is as large as or larger than the commutated current, so in the high thousands of amps. This is the reason for the tight integration of the driver with symmetrical planar busbars (Lstray less than 2 nH typically) and the large bank of electrolytics on the driver board.

There are also some pretty big presspack IGBTs available these days: https://www.infineon.com/dgdl/Infineon-P3000ZL45X168-DataSheet-v03_00-EN.pdf

[T3sl4co1l]

Hmm, doesn't work.  Oh-- leave the ?fileId on there, it's part of the link:
https://www.infineon.com/dgdl/Infineon-P3000ZL45X168-DataSheet-v03_00-EN.pdf?fileId=8ac78c8c7f2a768a017facf5116e40b2

[NiHaoMike]

You were quite mistaken. There are transistors capable of switching far greater loads than a lawn mower motor, which is a fairly small in the scheme of things.

Aren't the modern BLDC mowers only on the order of 300-400W? Not even that big by consumer equipment standards, similar to a Walmart special hybrid bicycle motor. (And a good hybrid bicycle goes well into the kW range!)

[coppercone2]

I don't like all this parallel crap someone needs to go out there and make a giant single transistor

[Fried Chicken]

You were quite mistaken. There are transistors capable of switching far greater loads than a lawn mower motor, which is a fairly small in the scheme of things.

Aren't the modern BLDC mowers only on the order of 300-400W? Not even that big by consumer equipment standards, similar to a Walmart special hybrid bicycle motor. (And a good hybrid bicycle goes well into the kW range!)

He's right, I was completely mistaken... hence why I made this post lol.

The biggest transistors I've encountered are audio transistors, and often they're distributed across like 10 of them for a 250wpc amplifier or something.

But I have no clue, I'm not an electronics guy.  I thought there would be some limit having to do with the purity of silicon manufacturing or something, similar to binning CPUs.  There also aren't many youtube torture tests of massive transistors, unlike other electric components.... like the 5000 amp fuse that guy popped lol

[T3sl4co1l]

Heh, well the first problem with (linear) audio amps is power dissipation.  You need a ton of metal to carry that heat out.  If you don't dissipate it in the first place, you need less silicon (though, not wildly so), and get more output for given input -- efficiency goes way up.  Hence, class D amps.

The actual chips used inside those amplifier transistors, is pretty small -- a few mm on a side -- again, packaging dominates overall component size -- it's just bringing all that heat out to the air that's the hard part.  Switching power supplies avoid that by dissipating full power for extremely small moments in time, averaging low power overall.

Tim

[Zero999]

Start again. What are you trying to do?

Find out what the world's biggest transistor is.  If it even is a transistor in the traditional sense.

And what the heck does that have to do with a lawn mower motor?

Let's say I wanted to control a lawnmower motor with a transistor...  What's the world's biggest transistor?  I know power amplifiers have some big transistors, but what if we get into the stupid range?

I guess I should have explained... I was looking at someone building a lawnmower, and I wasn't sure transistors even got large enough to conceivably do that, so I thought a lawnmower was a ridiculous application for a transistor, like you would use something else to power it.  Turns out that's not the case, and transistors get quite large.

You were quite mistaken. There are transistors capable of switching far greater loads than a lawn mower motor, which is a fairly small in the scheme of things.

You received some irrelevant replies because you didn't say exactly what you're doing. What's the world's biggest transistor? is a meaningless question. Are you talking about physical size, current rating, voltage rating, power dissipation?

I had a quick look through Digi Key and found some transistors which are more than capable of switching a lawn mower motor. As you made no mention of voltage, or whether it's DC or AC, I assumed smoothened DC, from a bridge rectifier powering a universal motor, drawing around 1kW or so.

https://www.digikey.co.uk/en/products/detail/stmicroelectronics/STP45N40DM2AG/5724369
https://www.st.com/content/ccc/resource/technical/document/datasheet/ff/da/51/30/5f/70/4e/77/DM00213649.pdf/files/DM00213649.pdf/jcr:content/translations/en.DM00213649.pdf

https://www.digikey.co.uk/en/products/detail/infineon-technologies/IRFR4620TRLPBF/2021282
https://www.infineon.com/dgdl/irfr4620pbf.pdf?fileId=5546d462533600a4015356323ae920f7

They're not that expensive considering. Anther option is to use thyristors for part of the bridge rectifier, which would probably be cheaper but not so flexible.

Those look tiny for the power they are rated for.

I had some idea that production errors or something made larger transistors impossible or very difficult/expensive.  I guess not.  Just add silicon.

Whilst it's true manufactures often exaggerate when it comes to power ratings, a transistor doesn't have to be that big, to switch a relatively large motor. Remember most of the power goes to the motor. The amount of power lost in the transistor should be as little as possible.

Take the STP45N40DM2AG I linked to in my previous post.
https://www.st.com/content/ccc/resource/technical/document/datasheet/ff/da/51/30/5f/70/4e/77/DM00213649.pdf/files/DM00213649.pdf/jcr:content/translations/en.DM00213649.pdf

Suppose you use it to switch a 2400W motor, running off 240V at 10A. Calculate the power lost in the transistor.

According to the data sheet, it has an on resistance of 0.072Ω.

P = I²R = 10²*0.072 = 100*0.072 = 7.2W

Oh, that's a little optimistic, reading down the data sheet, it also says the on resistance increases with temperature and it's specified at 25°C, but that's no good, since it's going to be near hot motor. Assuming the transistor might get very hot, say 125°C, the on resistance and thus the power loss will double. Refer to the graph Figure 10 Normalized on-resistance vs temperature on page 7 of the data sheet.

P = 14.4W

A heat sink is used to dissipate the power lost in the transistor, which is normally much larger than the transistor.

According to the transistor's data sheet, it has a thermal resistance of 0.5°C/W. This means for every W of power dissipated the temperature of the transistor itself, inside the case will increase by half a degree, when on a perfect heat sink. The maximum operating temperature of 150°C, but it's never a good idea to work at the maximum temperature and the power losses increase, so lets limit it to 125°C. Assuming it needs to work on a hot day, next to a hot motor, say an ambient temperature of 65°C.

Maximum temperature rise
T_r = T_max - T_ambient =  125-65 = 60°C

Maximum thermal resistance to ambient
TR = T_r/P_d = 60/14.4W = 4.17°C/W

The thermal resistance of the transistor is 0.5°C/W so that needs to be subtracted.
4.17-0.5 = 3.67°C/W.

So your heat sink needs to have a lower thermal resistance than that.

[shapirus]

Meh.

Connecting devices in parallel is cheating, it's not interesting.

Let's rephrase the question this way:

what is the world's highest power-rated single-die BJT?

[NiHaoMike]

The biggest transistors I've encountered are audio transistors, and often they're distributed across like 10 of them for a 250wpc amplifier or something.

Modern ones use much less silicon, TI does 600W on a chip the size of a sunflower seed! (As far as I'm aware, that's still the record for a single chip amplifier, anything bigger actually has multiple chips inside the module.)

But I have no clue, I'm not an electronics guy.  I thought there would be some limit having to do with the purity of silicon manufacturing or something, similar to binning CPUs.  There also aren't many youtube torture tests of massive transistors, unlike other electric components.... like the 5000 amp fuse that guy popped lol

Test of a Prius inverter assembly:

[T3sl4co1l]

Meh.

Connecting devices in parallel is cheating, it's not interesting.

Let's rephrase the question this way:

what is the world's highest power-rated single-die BJT?

Not much, considering they're obsolete for power applications

Back in the heyday, I suppose one of those 100A+ triple darlington types would've been up there.  No SOA to speak of though; switching application.

Tim

[CaptDon]

Just an experience observation from the speak of IGBT's and such. Our older AC4400 locomotives used G.T.O. technology in the inverter stacks. Gate Turn Off had a lot of drawbacks for high power inverters that were being switched at various loads and frequencies. It was hard to optomize for such a broad operating range. When we switched to IGBT's life got less complicated. Early on we had incredibly destructive shoot-through blowouts. If you can imagine 12 capacitors 1700uf@2.5Kv all charged to around 1400vdc and being fed from a 4500 h.p. output alternator with the rotating mass of the crankshaft and alternator weighing in at around 40,000 pounds and suddenly feeding a shoot-through dead short!!! Far worse was when one of the oil filled capacitors went dead short!!!! That particular field failure would blow the aux cab doors right off the locomotive and set the insides of the aux cab on fire with burning oil vapors. Our IGBT drivers had to have various fail-safe interlocking to insure one side of the half bridge could not be commanded on until the other side was tested to be off and that off condition included a dead-time timeout. There are six IGBT's per inverter stack and each IGBT is about 6X9 inches although internally consist of dozens of individual power 'transistors'.

[Zero999]

Meh.

Connecting devices in parallel is cheating, it's not interesting.

If the dies were all manufactured on the same wafer and have good thermal and electrical connections, i.e. mounted in the same package, then there's not much difference between a large single die and many small ones. A large die still has the problems associated with non-uniform construction, due to imperfections in the process and one part heating more than the other, causing its characteristics to be different to the rest.

[shapirus]

A large die still has the problems associated with non-uniform construction, due to imperfections in the process and one part heating more than the other, causing its characteristics to be different to the rest.

Yeah, the reasons why parallelling multiple smaller dies has practical advantages and is the way to go for mass production are clear. And that is precisely what directs one's curiosity to the largest single-die devices. It might be a one-off product made for a specific application, an experiment, or whatever that's not as boring as mass-produced stuff to be used in everyday life.

[Fried Chicken]

Those look tiny for the power they are rated for.

I had some idea that production errors or something made larger transistors impossible or very difficult/expensive.  I guess not.  Just add silicon.

Whilst it's true manufactures often exaggerate when it comes to power ratings, a transistor doesn't have to be that big, to switch a relatively large motor. Remember most of the power goes to the motor. The amount of power lost in the transistor should be as little as possible.

Take the STP45N40DM2AG I linked to in my previous post.
https://www.st.com/content/ccc/resource/technical/document/datasheet/ff/da/51/30/5f/70/4e/77/DM00213649.pdf/files/DM00213649.pdf/jcr:content/translations/en.DM00213649.pdf

Suppose you use it to switch a 2400W motor, running off 240V at 10A. Calculate the power lost in the transistor.

According to the data sheet, it has an on resistance of 0.072Ω.

P = I²R = 10²*0.072 = 100*0.072 = 7.2W

Oh, that's a little optimistic, reading down the data sheet, it also says the on resistance increases with temperature and it's specified at 25°C, but that's no good, since it's going to be near hot motor. Assuming the transistor might get very hot, say 125°C, the on resistance and thus the power loss will double. Refer to the graph Figure 10 Normalized on-resistance vs temperature on page 7 of the data sheet.

P = 14.4W

A heat sink is used to dissipate the power lost in the transistor, which is normally much larger than the transistor.

According to the transistor's data sheet, it has a thermal resistance of 0.5°C/W. This means for every W of power dissipated the temperature of the transistor itself, inside the case will increase by half a degree, when on a perfect heat sink. The maximum operating temperature of 150°C, but it's never a good idea to work at the maximum temperature and the power losses increase, so lets limit it to 125°C. Assuming it needs to work on a hot day, next to a hot motor, say an ambient temperature of 65°C.

Maximum temperature rise
T_r = T_max - T_ambient =  125-65 = 60°C

Maximum thermal resistance to ambient
TR = T_r/P_d = 60/14.4W = 4.17°C/W

The thermal resistance of the transistor is 0.5°C/W so that needs to be subtracted.
4.17-0.5 = 3.67°C/W.

So your heat sink needs to have a lower thermal resistance than that.
[/quote]

So the gate voltage on one of these, Vgs is +/- 25v, and the threhsold is from 3-5v?  So you would need up to 25 volts to drive this transistor?
I think it's finally clicked with me how a transistor works... it breaks the rules of electrical circuits.  This was never taught in schools b/c nobody actually knows anything about them.

[Zero999]

So the gate voltage on one of these, Vgs is +/- 25v, and the threhsold is from 3-5v?  So you would need up to 25 volts to drive this transistor?

No.

The on resistance is specified at different gate drive voltages. 25V is just the maximum rating. Read the data sheet.

I think it's finally clicked with me how a transistor works... it breaks the rules of electrical circuits.  This was never taught in schools b/c nobody actually knows anything about them.

No.

A transistor doesn't beak and rules. Why do you believe this to be the case?

[Fried Chicken]

No.

A transistor doesn't beak and rules. Why do you believe this to be the case?

Well it doesn't follow the rules of a traditional circuit.  Same with Op-Amps

[MarkT]

AlN is aluminum nitride.  I didn't look it up, but it must be a non-conductive solid or a semiconductor.

In undoped form its a ceramic insulator with extremely good heat-conductivity (though not as good as the very poisonous BeO ceramic).  You can make it semiconducting through doping.  Its basically a great material for heatsinking.

[MarkT]

I don't like all this parallel crap someone needs to go out there and make a giant single transistor

Well there are some nice large thyristors (SCRs) like: https://www.infineon.com/dgdl/Infineon-T2600N-DataSheet-v03_01-EN.pdf?fileId=5546d46277fc7439017836a6f5ff7deb, basically a 65mm wafer in a hockey-puck package.

[Vovk_Z]

the overhead Xiangjiaba-Shanghai interconnector in China and the Rio Madeira HVDC system in
Brazil. The Xiangjiaba-Shanghai line is the world’s first UHVDC connection. It operates at ±800 kV
and transfers 7200 MW from the Xiangjiaba hydropower plant in southwest China to Shanghai,
which is 2000 km further away. It is a single overhead line. The losses are rated at less than 7%.
The Rio Madeira HVDC system is the longest transmission link in the world. It carries 6300 MW

7200 MW?
Will it fit my Civic? (c)

[tggzzz]

No.

A transistor doesn't beak and rules. Why do you believe this to be the case?

Well it doesn't follow the rules of a traditional circuit.  Same with Op-Amps

There's no such thing as a "traditional circuit"; the only things are electromagnetic fields and their interaction with matter.

There are useful models that are based on simplification of those, and which allow easier "higher level" predictions that are valid in some circumstances but not others.

One example of that is "Ohm's Law". You occasionally come across people that believe V=IR, for example.

[tooki]

No.

A transistor doesn't beak and rules. Why do you believe this to be the case?

Well it doesn't follow the rules of a traditional circuit.  Same with Op-Amps

Please, explain which rules an op-amp breaks and how it breaks them.