new dual NPN and PNP transistors

[Jim Thompson]

I read in sci.electronics.design that Winfield Hill
about 'new dual NPN and PNP transistors', on Sun, 3 Oct 2004: [snip]

But I'm puzzled by the beta matching offered by some
manufacturers - what's the appeal? If a 10% hfe spec corresponded
to a 10% mirror or offset match, that would be useful, but there's
no indication that there's any correspondence.

I thought so, too. we are always warned not to accept designs that
depend on even a 2: 1 beta range.

My guess is that if your device has a 10% beta match, flaunt it, even
though it's irrelevant. Next week, dual devices in which the package is
****exactly the same colour for both devices****!!!!.

Gee! I don't have a bit of trouble making better than 1% mirrors,
it's as simple as just plunking them down on the schematic ;-)

...Jim Thompson

 

[Ken Smith]

My guess is that if your device has a 10% beta match, flaunt it, even
though it's irrelevant. Next week, dual devices in which the package is
****exactly the same colour for both devices****!!!!.

... except in the US they will leave the "U" out.

When the spec goes from 10% HFE to 10% on colo(u)r, you can expect them to
call the new part improved.

Improved: Changed in some random way.

Obsolete: The guy who knew how to make them retired.

Low noise: Less than rail to rail noise.

Compatible: May work in some selected applications.

Direct cross: Double crossed by marketing.

 

[Stefan Heinzmann]

Joerg wrote...



http://www.semiconductors.philips.com/pip/BCM847BS.html
Is this interesting? Most low-cost dual parts these days aren't
monolithic, and have no matching spec at all. Some of the "mirror"
parts have 40% mirror specs, not very good, but better than nothing
(this corresponds to a 10mV Vbe offset for matched transistors, see
AoE fig 2.53). But I'm puzzled by the beta matching offered by some
manufacturers - what's the appeal? If a 10% hfe spec corresponded
to a 10% mirror or offset match, that would be useful, but there's
no indication that there's any correspondence.

Wouldn't it be enough to just put two dice from the same wafer into the
same package for achieving such levels of matching?

 

[Jim Thompson]

On Sun, 03 Oct 2004 23:32:34 +0200, Stefan Heinzmann

[snip]

Wouldn't it be enough to just put two dice from the same wafer into the
same package for achieving such levels of matching?

That's what is done in, for example, the MPQ2222.

...Jim Thompson

 

[Joerg]

Hi Kevin,

Yep the classic, LM394 pair, is 50uv matching and 2% hfe match (0.5%
typ), with very low 1/f noise. It costs a bit though.

Nice but.... over $3. Ouch. That's cost prohibitive in most of my
designs. A lot of tasks can also be solved by moving from relative
precision or match to absolute precision. Things such as the reference
LMV431buys you better than 2% for around 15 cents. So that one has
become one of my staples.

Sometimes I envy chip designers like Jim Thompson who can add matched
circuitry with a few mouse strokes and it costs them next to nothing.

Regards, Joerg

 

[Joerg]

Hi Winfield,

Is this interesting? Most low-cost dual parts these days aren't
monolithic, and have no matching spec at all. ...

Well, as Stefan pointed out there is some matching if they are cut from
the same wafer or even better from the same area on a wafer.

But all that reminds me of guys who bought oodles of parts and then sat
there with a meter all day long and matched up stuff. Transistors,
inductors, caps, whatever they could find a market for. A few decades
ago that was an easy way to make money. Remember all the 1% caps needed
for the allpass and the anti-alias in a Doppler? The absolute values
didn't matter much but the deviation did. Then came the DSP and all
these folks had to think about something else, I believe one of them
opened a restaurant.

Regards, Joerg

 

[Robert Baer]

Hi Kevin,

Nice but.... over $3. Ouch. That's cost prohibitive in most of my
designs. A lot of tasks can also be solved by moving from relative
precision or match to absolute precision. Things such as the reference
LMV431buys you better than 2% for around 15 cents. So that one has
become one of my staples.

Sometimes I envy chip designers like Jim Thompson who can add matched
circuitry with a few mouse strokes and it costs them next to nothing.

Regards, Joerg

Diodes Inc makes a lot of NPN, PNP, FET and NPN/PNP pairs in rather
dinky SMT packages, at a decent price.

 

[Joerg]

Hi Robert,

Diodes Inc makes a lot of NPN, PNP, FET and NPN/PNP pairs in rather
dinky SMT packages, at a decent price.

There are also some from Japanese manufacturers. However, when they are
single sourced that is a large obstacle in most designs.

Regards, Joerg

 

[Tim Wescott]

Hi Robert,



There are also some from Japanese manufacturers. However, when they are
single sourced that is a large obstacle in most designs.

Regards, Joerg

It continues to boggle my mind that there aren't more second-sourced
stuff. I think it's the whole cell phone/motherboard phenomenon:
Manufacturers will design one, make a run of 10 or 100k of them, then go
on to the next design -- so as long as they have their parts lined up
from day 1 they don't have to worry about obsolescence or price jacking
down the road or any other long-term thing.

It's just those manufacturers who want to make something over a span of
years that have to worry, and they're getting to be a smaller and
smaller segment of the market.

 

[Winfield Hill]

Jim Thompson wrote...

Gee! I don't have a bit of trouble making better than 1% mirrors,
it's as simple as just plunking them down on the schematic ;-)

That's because it's not simply you who's making the accurate
mirrors, it's the fab house with their fine processes. :>)
Naturally you do get to provide some embellishment in the exact
configuration you plunk down.

 

[Joerg]

Hi Tim,

It continues to boggle my mind that there aren't more second-sourced
stuff. I think it's the whole cell phone/motherboard phenomenon:
Manufacturers will design one, make a run of 10 or 100k of them, then
go on to the next design -- so as long as they have their parts lined
up from day 1 they don't have to worry about obsolescence or price
jacking down the road or any other long-term thing.

This started as a TV/VCR phenomenon and then migrated to cell phones and
other stuff. There are indeed fewer companies that place their core
business in the usual electronics market. Luckily some big players in
the US keep up in this "old" market but one can already see some
spinning off of these business units, as was the case with Motorola.

In order not to get burned it is more important than ever to always
check who else makes the part. In the 80's it was ok if there was just
one other main player. Nowadays I look for at least two.

Regards, Joerg

 

[Jim Thompson]

On Sun, 03 Oct 2004 22:42:43 GMT, Joerg

[snip]

Sometimes I envy chip designers like Jim Thompson who can add matched
circuitry with a few mouse strokes and it costs them next to nothing.

Regards, Joerg

I'm sitting here doing just that right now. Replacing a couple of
pages of kludge (clearly designed by a discrete designer ;-) with a
few current mirrors, and improving the system accuracy an order of
magnitude.

...Jim Thompson

 

[Joerg]

Hi Spehro,

http://www.diodes.com/datasheets/ds30311.pdf
http://www.diodes.com/datasheets/ds30312.pdf

"Built with adjacent die from a single wafer (!!!!)
2% Matched Tolerance, hFE, VCE(SAT), VBE(SAT)"

Mouser has them in stock and they are only 16 cents US in 1K.

Now that's a deal. Also, if something is available at places like
Mouser, Digikey or Newark it must have some mainstream market. I'd be
worried about single source though. Been burned just one too many times.
The phrase "back order" has the same appeal to me as "pea soup". I don't
like pea soup...

Regards, Joerg

 

[Spehro Pefhany]

Hi Robert,


There are also some from Japanese manufacturers. However, when they are
single sourced that is a large obstacle in most designs.

Here's another one with even worse matching (hFE within 10%, Vbe match
within 5 mV)

http://www.centralsemi.com/pdf/CMKT5089M10.PDF

No sign of any availability, either.


And here's one that's more specified for current mirror applications:

http://www.infineon.com/cmc_upload/documents/044/192/bcm846s.pdf
http://www.infineon.com/cmc_upload/documents/044/193/bcm856s.pdf

Again, no sign of availability.


But, the most interesting one to me is this one from Diodes Inc.:

http://www.diodes.com/datasheets/ds30311.pdf
http://www.diodes.com/datasheets/ds30312.pdf

"Built with adjacent die from a single wafer (!!!!)
2% Matched Tolerance, hFE, VCE(SAT), VBE(SAT)"

Mouser has them in stock and they are only 16 cents US in 1K.


Best regards,
Spehro Pefhany

 

[Robert Baer]

Hi Spehro,

Now that's a deal. Also, if something is available at places like
Mouser, Digikey or Newark it must have some mainstream market. I'd be
worried about single source though. Been burned just one too many times.
The phrase "back order" has the same appeal to me as "pea soup". I don't
like pea soup...

Regards, Joerg

Somebody peed in your soup???

 

[Rich Grise]

Somebody peed in your soup???

You can cream asparagus, but you can't pea soup.

 

[Kevin Aylward]

Win, You believe too much in what you read in books. Most ANALOG is
NOT CELLS.

He wasn't referring to cells, this is obvious.

I roll my own, specifying device sizes, etc.

And sometimes you blow as much as me. Win was correct, its not *you*
that makes the matching. Its the process. Getting the size right is
nothing more than choosing error = K/sqrt(WL). K is set by the process
design.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Tim Shoppa]

Getting useful information through most any of the Japanese web sites
is a harrowing experience. It would be interesting to see if
Japanese-speaking engineers have the same problems, or if the problems
are just with Gaijin users.

We've been spoiled to some extent by the TI and National Semiconductor
and Microchip and Motorola(OnSemi) web sites, which are wonderfully
good and show a lot of coordination
between the folks who do the web site and the engineering and sales staff.
The Fairchild Semi and Agilent sites are OK but not quite top-notch.

Speaking generally, the Japanese and European websites are nearly useless.
I've had some limited success with NEC and Toshiba. My guess is that
they went out and spent money on a pretty website that is pretty much
useless.

Tim.

 

[Spehro Pefhany]

On Mon, 04 Oct 2004 06:59:11 GMT, the renowned "Kevin Aylward"

And sometimes you blow as much as me.

Is this that self-deprecating thing?


Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

You can cream asparagus, but you can't pea soup.

There's no need to inspissate pea soup.

But hey, pea soup can be pretty good. Especially with a few other
vegetables and lots of smoked ham or bacon.


Best regards,
Spehro Pefhany