Siliup MOSFETs and their datasheets a tale of two fets

[ballsystemlord]

Hello,
So I was browsing for MOSFETs on LCSC and came upon Siliup's SP85N02AGHTF (direct datasheet link). I downloaded the datasheet, and began to read through it. 85V... 280A... 270W... SOA curve... at DC it does 30V at 30A which is 900W... 900W!?!?

So I searched online for the same MOSFET and found other distributors with the identical datasheet.

A bit confused, I went to siliup's offical website and got a totally different datasheet for the same part (there is no product page that I could find.)

The SOA curve is much more conservative with 30V at 0.3A or 9W... which is equally strange for a (now) 380W part...


So I'm a bit confused.

Can anyone shed some light on what's going on?

LCSC is (supposedly) a reputable distributor. So, they shouldn't have outdated datasheets. Likewise, they shouldn't have companies that produce "Mystery MOSFETs" on their site.

Thanks!

[Whales]

30A@30V DC for 1USD each?  Woah.

LCSC is (supposedly) a reputable distributor. So, they shouldn't have outdated datasheets.

Other way around, check the version numbers in the lower left corner the PDF:

V1.0 - OEM website datasheet, lower SOA
V1.3 - LCSC hosted datasheet, incredible SOA

All I can suggest is trying to contact Silup and ask why their SOA spec has changed to be stratospheric.  Unless this is common now and other manufacturers do the same for similar prices?  I'm guessing it's an error.

[paul cotter]

Unsafe operating area, ie UOA Late addition: it is a long time since I did any semiconductor design work and this question may/may not be relevant- why do manufacturers not supply both forward and reverse SOAs? I seem to remember that for SMPS design the reverse one is needed?(current crowding on switch off??)

[Kean]

The only way to get 30V across the FET when drawing 30A DC is to have an R_DS of 1Ω.  That would only happen if you have it barely turned on, and not surprisingly it will burn if left in that state.
Common sense says that is definitely not in the SOA for continuous operation.

If you have it fully turned on with V_GS at 10V and drawing 30A DC through it, then R_DS should be around 1.7mΩ.  So V_DS will be about 50mV, thus it would be dissipating only about 1.5W.

At the full 280A DC rating, V_DS will be about 476mV, thus dissipating 133W.  With R_θJC of 0.46 that will give a temperature differential of 61°C between junction and case.  You'll need a decent heatsink with forced air cooling to keep that in the SOA (i.e. junction below 150°C, so heatsink below 89°).

The v1.3 datasheet also has pulse lengths measured in Siemens.  Why is this so common? 

[wraper]

Having 0.46^oC/W Junction-to-Case thermal resistance, 900W is certainly not physically possible at Tc=25^oC and Tj(max)=150^oC as it would result in 439^oC Tj. P_D=270W in V 1.3 LCSC datasheet is correct in this regard. But it has lower 0.33^oC/W and higher P_D=380W in V1.0 datasheet.

[Whales]

Oof, so lots of things are changing.  I don't like that. 

They should change the part number if the part has changed, or, if they got their original measurements wrong then they should provide an equivalent of an apology (a changelog).

EDIT: Has the pinout been mirrored too?  Hopefully the V1.0 (left) is just vaguely drawn 

[wraper]

EDIT: Has the pinout been mirrored too?

No, it's just recesses drawn a bit weird. They're drawn correct in packaging drawing.

[mtwieg]

The V1.3 datasheet has massive red flags:
1. Charge specs (Qg, Qgs, Qgd) are exactly the same, despite the capacitance specs changing significantly. Clearly impossible.
2. The "output characteristics" curves don't match with a device with Rdson of 1.7mohm. More like 5mohm.
3. "Capacitance characteristics" curves are completely different, overall a factor of two greater than V1.0.
4. Gate charge curve is completely different, and suggests a Qg of ~63nC, not 143nC. Also the plateau voltage is way higher.
5. SOA curves are obviously BS.

Based on this alone I wouldn't bother considering this company's devices for a serious project. But I do wonder why LCSC has a higher rev of the datasheet. The V1.3 datasheet doesn't actually have a date, so maybe it's something LCSC invented?

Late addition: it is a long time since I did any semiconductor design work and this question may/may not be relevant- why do manufacturers not supply both forward and reverse SOAs? I seem to remember that for SMPS design the reverse one is needed?(current crowding on switch off??)

I've never seen RBSOA specified for a MOSFET, pretty sure it's only relevant for IGBTs.

[David Hess]

The v1.3 datasheet also has pulse lengths measured in Siemens.  Why is this so common? 

They never should have deprecated mhos (℧).

In practice they expect you to distinguish siemens (S) from seconds (s) using context instead of capitalization.

Unsafe operating area, ie UOA Late addition: it is a long time since I did any semiconductor design work and this question may/may not be relevant- why do manufacturers not supply both forward and reverse SOAs? I seem to remember that for SMPS design the reverse one is needed?(current crowding on switch off??)

Instead of a graph, the RBSOA is covered by the avalanche specifications intended for switching an inductive load.  Not all devices have this specification.

So for instance the 100V 200A 1040W (at 25C) IXTK200N10L2 avalanche rated Linear L2 MOSFET from IXYS has a full FBSOA graph, and specifies an avalanche current (I_A) of 100 amps and 5 joules, at 25C of course.

[Whales]

I would be fascinated to learn about the business structure of these small chinese IC companies.  Presumably the silicon manufacture is all outsourced to a foundry?  Do they have internal knowledge & design, or rely on what the foundry tells them?  What do they do for testing?  Are they a whitelabel?

[oreo99]

I too have found issues with the SOA curves on some China brand Fets.
Also, just the specifications in general.

If you are looking for a rugged, high current mosfet in a TO-247 package, I would recommend the
Huayi HY5608W/HY5110W/HY5012W series.  (depending on voltage requirements).

It is easy to see why this family of devices is so good.  Below is a picture of the dies of three devices to compare.
Huaya HY5012w (300A 125v 500w)
Siliup SP012N02AGHTF (270A 120v 310w)
IR IRFP4227 (65A 200v 330w)

[Kleinstein]

It is not just transistors from China. It is rather common to find wrong SOA curves. This happens when they are derived from the transient thermal resistance - this way the thermal instability is ignored and thus missing the main point with the SAO curve. Infineon usually has realistic SOA curves in there data-sheets. Modern low voltage FETs can have a pretty poor SOA. Still only 0.3 A at 30 V looks like a bit on the conservative side.