A SOT-89 with emitter in the middle

Am 21.12.2017 um 21:07 schrieb LM:

Yes.

Infineon BFQ790.

regards, Gerhard

Good. Digikey has it. Then read long "Absolute Maximum Ratings" list and began to wonder. BFQ790 is a bit delicate and too fast for my applicationas as

And Mini-Circuits wideband RF amplifiers, that come in 3-pin SOT-89 packages, with an inductor to the supply bus, 50-ohm IN and 50-ohm OUT, have their tab and middle pin grounded.

Not me. But phemts usually have the center pin/tab as source, which is really handy.

The heat apparently comes out the collector of transistors but the source of phemts.

It is the substrate that naturally ends up on the tab.

That's why with JFETs, it is usually the gate. The channel is in the middle between the top and bottom gate regions, and the chip thickness plays a role in the channel height, which is not so closely controlled. Hence the variation in Vgs in JFETs for the "same effect".

The BFQ790 seems to be somewhat unusual with the emitter on the substrate side, without that usual +++ doped collector well that kills voltage ratings.

Note the 18 V Vcb for a 20 GHz SiGe transistor. That could be interesting for large output voltage in a cascode. And in a cascode, the emitter is relatively quiet, voltage-wise.

OK, the base on the tab would be even better for that. :-)

It seems, they did a lot of work to characterize that thing. Much more than just abs. max. ratings and s-parameters as it is common today.

cheers, Gerhard

Unfortunately, SOT89 RF transistors get discontinued a lot faster than they get introduced. I designed a wonderful circuit recently, that uses BFQ149s (PNP) parts, and then they were discontinued. Bummer. We bought 11,000 to tide us over for a while, but I'm going to have to redesign eventually.

NEC and NXP have killed most of their good stuff.

The actual breakdown voltages on RF parts, bipolars and phemts and schottky diodes, seem to be at least twice the data sheet value.

Seems like they rate them for DC, expecting the AC peaks to be of that order. Not that I've designed a Class C amp using SC-70s lately. ;)

Cheers

Phil Hobbs

Those can make an easy RF amplifier, true.

Having the isolation between emitter and collector on the chip or having a large capacitance and area in the collector because of heatsinking. Or using some virus size chip no one will solder.

Did you look into Triquint? (Qorvo) Maybe they have something.

What is wrong with grounded base?

Maybe "too fast" means the data was received before it was sent...

If the CRC comes in first, you can check it before the data arrives and save time.

OK, I'll bite. What's the wonderful design that'll swallow up 11,000 BFQ149s, and come back for more?

It's the output stage of a pulse generator.

One of my nearly lifelong hobbies is designing pulse gen output stages. Logic level in, 50 ohm out, programmable Vhi and Vlo, fast, clean, tough, low power dissipation. When I filnally had a really good one, the BFQ149 went EOL.

It annoyed me enough that I bought $5000 worth of PNPs so I don't need to redesign it soon.

As a Christmas present from Broadcom, I got an EOL notice for the ATF55143, a very nice small E-pHEMT. It's sort of an enhancement-mode version of the ATF38143.

Thanks a bunch, Broadcom.

Cheers

Phil Hobbs

The ephemts are wonderful, and I worry about them too.

Gabe, an Avago ephemt guy, lives across the street. I'll ask him for an honest opinion about the status of those parts.

There are a couple of new players in the fast bipolar biz, like Rohm and Toshiba and ON. But the newish stuff is mostly SOT23 and NPN. So we're back to all-NPN design.

I checked, there's virtually no BFQ149 stock available. When the inventory disappears that fast, it means, or it should have meant to NXP, the demand really wasn't low enough to justify a sharp product discontinuation. (There are other solutions, they could have increased the price and waited a few years, or made a mass of product and deposited it with Rochester, etc.)

I think that a bean counter, imagining greater profits by radically trimming a product line, isn't taking into account the ultimate losses that will result from the bad effects of his breaking faith with the engineering community designing in his products in the first place.

Or perhaps the fab the part was run on was being phased out and it wasn't worth the cost of moving it to another fab. Such things happen, particularly when companies are trying to figure out what to do with acquisitions/spin-offs.