Fast PNPs--suggestions?

More on this subject 5 years ago:

It is discussed at length here:

and IIRC here

Authors seem to be TEK veterans. Both books are _must_have_! Also cover things like T-coils.

regards, Gerhard

I have that one--it's a great book. Lots of nice math. ;)

I ordered this one as well. Thanks for the steer!

The fancy noise canceller is interesting because it needs a bunch of moderate-speed elements that have to be very precise--gain and phase errors at least 60 dB down (0.1% / 0.06 degrees) over the whole operating range of photocurrent, current balance, and frequency. So in a way it's like microwave design, except that inductance isn't a problem for the most part. You have to watch the layout very carefully, and put in a few strategically-placed small varactors.

Things like current mirrors become a real challenge too. Fun!

Cheers

Phil Hobbs

as the

the BFT92 or

milliamps, nothing specified

side--output

capacitance <~

Cob is 0.5p

started my

that I

to pay not

calibration and

guys look

(3) I

than

computing

everything else

gonna

redesign?

get paid

cert, no inventory

business.

too slow

Where does this say that non-trivial applications do not exist? Do you know what the phrase "to a degree" means?

That was an example. I thought you'd understand that. But it seems not.

Utter nonsense. I have never said that.

right,

Nonsense. You put words in people's mouth that they didn't say. I expected better.

As for choice, there usually is one.

to another

rational

everywhere.

has it

dust,

written.

You say that I wrote something (that wasn't true), I ask for a quote, and you can't furnish such quote.

It is certainly not me who should stick more to the words written in this thread.

Not true.

Phil, when you have one of these boards ready for sale, let me know.

Steve

Oh, I'll be shouting from the housetops, don't worry. ;) Thanks for the interest.

We just decided to bail out on the ST micro because the development support is so putrid, in favour of an LPC1752 or maybe an LPC1343. The Code Red tools are so much nicer than IAR, and you don't have to read megabytes of cryptic headers to find out how to turn on a GPIO pin.

Cheers

Phil Hobbs

Haha, BTDT. Using the STM32F2 for a new project. So far I scrapped ST's "peripheral library", just kept the 0.5MB header file for the register address and peripheral structure definitions. They are actually quite nice chips, so far anyway. Not used the M3 LPCs, although I have used the ARM7 parts a few times.

I've been using vanilla gcc + insight + openocd for the last decade or so. Perhaps I should give some of the more "packaged" tools another try (the gcc based ones that is).

specified

Coming back from holidays and I just see that...

Can't you go with a NPN and folded cascode?

specified

Thanks. That won't work because I need the current balance to be very accurate from DC to 10 MHz (it's the current I care about here). So far I'm using a pair of biased BFT92s with a BF862 and emitter follower wrapped round it to reduce the noise and impedance at the emitters. It looks like it'll give me a worst-case cancellation of -46 dB out to 10 MHz, and a few AC tweaks should get that down to -60 dB, which would be terrific. Hanging random 0.5 pF and 5 nH here and there doesn't do much to it, either, so hopefully it'll survive the transition to a real board relatively painlessly.

I just realized that the reason that my PNP current mirror is working better than I expected is that it has three feedback loops in it--the two local ones that control Q1 and Q4, plus the big one that wraps round the whole thing.

It also cancels the effect of base current to pretty good accuracy, and that includes the AC base current, which is what limits the performance otherwise. It's also well overdamped, even though it's pretty fast, which is nice. (I rarely use current mirrors in real life, and this one had my full attention.)

Was this one your invention, Jim? It has to be pretty ancient. ;)

Cheers

Phil Hobbs

either

road

nice

is

Now you know why i don't read Keith's posts any more.

?-)